Article

Anatomically distinct fibroblast subsets

determine skin autoimmune patterns

https://doi.org/10.1038/s41586-021-04221-8 Zijian Xu1,9, Daoming Chen1,2,9, Yucheng Hu3, Kaiju Jiang1, Huanwei Huang1, Yingxue Du1,

Wenbo Wu1, Jiawen Wang1, Jianhua Sui1, Wenhui Wang4, Long Zhang4, Shuli Li5, Chunying Li5,
Received: 2 March 2020
Yong Yang6, Jianmin Chang7 ✉ & Ting Chen1,8 ✉
Accepted: 5 November 2021

Published online: 15 December 2021

The skin serves as a physical barrier and an immunological interface that protects the
Check for updates body from the external environment1–3. Aberrant activation of immune cells can
induce common skin autoimmune diseases such as vitiligo, which are often
characterized by bilateral symmetric lesions in certain anatomic regions of the
body4–6. Understanding what orchestrates the activities of cutaneous immune cells at
an organ level is necessary for the treatment of autoimmune diseases. Here we
identify subsets of dermal fibroblasts that are responsible for driving patterned
autoimmune activity, by using a robust mouse model of vitiligo that is based on the
activation of endogenous auto-reactive CD8+ T cells that target epidermal
melanocytes. Using a combination of single-cell analysis of skin samples from
patients with vitiligo, cell-type-specific genetic knockouts and engraftment
experiments, we find that among multiple interferon-γ (IFNγ)-responsive cell types in
vitiligo-affected skin, dermal fibroblasts are uniquely required to recruit and activate
CD8+ cytotoxic T cells through secreted chemokines. Anatomically distinct human
dermal fibroblasts exhibit intrinsic differences in the expression of chemokines in
response to IFNγ. In mouse models of vitiligo, regional IFNγ-resistant fibroblasts
determine the autoimmune pattern of depigmentation in the skin. Our study
identifies anatomically distinct fibroblasts with permissive or repressive IFNγ
responses as the key determinant of body-level patterns of lesions in vitiligo, and
highlights mesenchymal subpopulations as therapeutic targets for treating
autoimmune diseases.

Vitiligo is an acquired polygenetic autoimmune disease9. Its defining fea-

ture is skin depigmentation, in which affected skin is unable to spontane-
ously recover owing to the activity of auto-reactive CD8+ T cells, which
causes a loss of epidermal melanocytes7,8,10–15. Vitiligo affects 0.5% to 2%
of the population worldwide16–18. More than 80% of patients with vitiligo
have bilateral symmetric patterns of depigmentation across the central
body axis, which is classified as non-segmental vitiligo4–6. Previous stud-
ies using a mouse model of vitiligo have shown that IFNγ is important for
T cell-induced depigmentation in skin19–21. Although various studies have
proposed underlying mechanisms22–27, the identity of IFNγ-responsive
cells that mediate this function is largely unknown. Hypotheses about the
reason behind the bilateral symmetric patterns that are observed in viti-
ligo pathology include regional variations in microbiota distribution; the
pattern of antigen expression in melanocytes; and different distributions
of neuropeptides released by nerve endings28–32. Despite these studies,
the cellular and molecular mechanisms that orchestrate the patterned
activities of immune cells in the skin of patients with vitiligo remain elusive.
Cellular IFNγ responses in vitiligo-affected skin
To understand the cellular and molecular mechanisms that lead to pat-
terned autoimmune activity in vitiligo-affected skin (Fig. 1a, Extended
Data Fig. 1a), we first analysed skin biopsies from patients with vitiligo.
Immunofluorescent staining showed that melanocytes were absent
within the lesion but evenly distributed in the basal epidermis of the
perilesion region (Fig. 1b). Notably, the majority of the infiltrated CD8+
T cells were concentrated at the junction area between the lesion and
the perilesion skin, which implied a local recruitment mechanism for
CD8+ T cells.
Next we used single-cell RNA sequencing (scRNA-seq) to analyse
all cell types present in patient skin (Extended Data Fig. 1b, c). Unsu-
pervised clustering of more than 50,000 cells from 10 patients with
vitiligo and 5 healthy donors yielded 8 main cell types defined by
signature genes (Extended Data Fig. 1d–h, Supplementary Table 1).
Further analysis of melanocytes revealed two distinct sub-clusters

National Institute of Biological Sciences, Beijing, China. 2Peking University–Tsinghua University–National Institute of Biological Sciences Joint Graduate Program, School of Life Sciences,
1

Peking University, Beijing, China. 3Academy for Multidisciplinary Studies, Beijing National Center for Applied Mathematics, Beijing Advanced Innovation Center for Imaging Theory and
Technology, Capital Normal University, Beijing, China. 4Peking University Third Hospital, Beijing, China. 5Department of Dermatology, Xijing Hospital, Xi’an, China. 6Institute of Dermatology,
Chinese Academy of Medical Sciences and Peking Union Medical College, Nanjing, China. 7Department of Dermatology, Beijing Hospital, National Center of Gerontology, Institute of Geriatric
Medicine, Chinese Academy of Medical Sciences, Beijing, China. 8Tsinghua Institute of Multidisciplinary Biomedical Research, Tsinghua University, Beijing, China. 9These authors contributed
equally: Zijian Xu, Daoming Chen. ✉e-mail: changjm0417@126.com; chenting@nibs.ac.cn

118 | Nature | Vol 601 | 6 January 2022

 a b Bright field DAPI CD8 TYR K14 c d
40 100

80

Melanocytes (%)
20
60

t-SNE 2
0
40

−20 20
g M1
M2 0
−40 1 2 3 4 5 6 7 8 9101 2 3 4 5
Perilesion T cell infiltration region Lesion −50 −25 0 25
t-SNE 1 Patient HD

ll
l
e f g h

el

ce
Tcyt in immune cells (%)

ce a te
K14 PDGFRA

Ke cl e i al c

ng go e
T erh cy
20

En obl yte
40

ll ns
La pha cyt
**

M oth t
d as
us el
br c

. o
Fi ano

M tin
15

DAPI pSTAT1
ra
el
NS

M
10
30
in immune cells (%)

Antigen presentation
5
T cell subtypes

IFNγ signalling
0
20 T cell migration
Per cent of pSTAT1+ cells
Vasculature development 0 20 40 60 80 100
3 **
IFNG normalized

10 mRNA catabolic Melanocyte

NS
expression

2 Oxidative stress Fibroblast

Endothelial cell
1 Lymphocyte immunity Muscle
0
Cell killing Keratinocyte
0
M. phagocyte
t

g
ef
cy

re
re
T

T
T

Langerhans cell
4 8 12 >14 1 10 20
Progressive Quiescent Healthy Enrichment −log10(P value) T cell

Fig. 1 | scRNA-seq analysis reveals distinct IFNγ-responsive cell types in the
skin of patients with vitiligo. a, Schematic illustration of bilateral symmetric
lesions in non-segmental vitiligo. b, Representative bright-field and
immunofluorescent images of vitiligo lesion skin. c, t-distributed stochastic
neighbour embedding (t-SNE) plot of melanocytes from the skin of patients
with vitiligo and healthy donor skin. d, Melanocyte composition in each patient
and healthy donor (HD). e, Analysis of T cell subtype composition. Tcyt,
cytotoxic T cells; Tres, resident T cells; Teff, effector T cells; Treg, regulatory
T cells. f, Percentage of CD8+ cytotoxic T cells and expression of IFNG in CD8+
cytotoxic T cells from the skin of patients with vitiligo and healthy donors.
g, GO categories of genes enriched in each cell type from the skin of patients
in the progressive state. M. phagocyte; mononuclear phagocyte.
h, Representative immunofluorescent images and quantification of pSTAT1
signal in different cell types in the skin of patients with vitiligo. Scale bars,
50 μm (b, h). For exact P values, see Source Data. For statistics, P summary and
sample sizes, see Methods.

(termed M1 and M2) (Fig. 1c). Upregulated genes (more than twofold;
P < 0.01) in melanocytes of the M1 cluster compared to the M2 cluster
showed strong immune responses, especially in terms of the response
to IFNγ (Extended Data Fig. 1i, Supplementary Table 2). Melanocytes
from patients 1–5 were predominantly composed of M1 cluster cells,
whereas patients 6–10 had a similar composition of melanocytes to that
of healthy donors (Fig. 1d). On the basis of the expression profile and
composition of melanocytes, we classified patients 1–5 as in a progres-
sive state, and patients 6–10 as in a quiescent state.
To validate the disease state classifications, we next analysed T cells.
Four major sub-clusters of T cells were identified by their expression
pattern: CD8+ cytotoxic; CD8+ resident; CD4+ effector; and CD4+ reg-
ulatory T cells (Extended Data Fig. 1j, k, Supplementary Table 3), all
of which were enriched in skin from patients compared to skin from
healthy donors (Fig. 1e, Extended Data Fig. 1l–m). Skin depigmenta-
tion in vitiligo is mediated by antigen-specific CD8+ cytotoxic T cells
targeting melanocytes13,33,34. In patients in the progressive state, there
were significantly more CD8+ cytotoxic T cells, which also expressed
significantly higher levels of IFNG (Fig. 1f).
Previous studies showed that IFNγ signalling is important for the pro-
gression of vitiligo in mouse models19–21. Gene ontology (GO) analysis
of enriched genes (more than 1.5-fold; P < 0.01) in the skin of patients
in the progressive state showed that melanocytes and fibroblasts were
overrepresented in categories related to the IFNγ signalling pathway
(Fig. 1g, Extended Data Fig. 1n), which was further confirmed by analy-
sis of signature IFNγ response genes in all cell types (Extended Data
Fig. 1o). Although IFNG was predominantly expressed by cytotoxic
T cells (Extended Data Fig. 1p), T cells in the skin of patients in the
progressive state did not show an increased IFNγ response compared
to healthy skin. This could be related to the low expression level of
IFNGR1 and IFNGR2 in T cells compared to other cell types (Extended
Data Fig. 1q).
To further validate the IFNγ-responsive cell types, we analysed phos-
phorylated (p)STAT1 in different cell types in skin with vitiligo lesions.
Spatially, the density of CD8+ T cells positively correlated with the den-
sity of pSTAT1+ cells (Extended Data Fig. 1r). Quantification showed
that among eight different cell types examined, dermal fibroblasts
accounted for most (around 80%) of the pSTAT1+ cells in patient skin
(Fig. 1h, Extended Data Fig. 1s).
Skin stromal IFNγ response drives vitiligo
To study the functional relevance of our analysis of patient samples, we
developed a mouse model in which vitiligo was induced by inoculation
with B16F10 melanoma coupled with depletion of regulatory T (Treg)
cells35, which recapitulated the pathologies of patients with vitiligo
(Extended Data Figs. 2, 3). After the same vitiligo induction procedure,
wild-type mice developed vitiligo—as shown by tail skin depigmenta-
tion, significant loss of melanocytes and abundant infiltration and
aggregation of CD8+ T cells—but Ifngr1 knockout (KO) mice did not
develop vitiligo (Fig. 2a, b, Extended Data Fig. 4a–h). This phenomenon
was not due to delayed cytotoxic activity, because even at four months
after vitiligo induction, melanocytes in Ifngr1 KO tail skin were still
intact (Extended Data Fig. 4i). RNA-seq and quantitative PCR (qPCR)

Nature | Vol 601 | 6 January 2022 | 119

Article
a WT Ifngr1 KO
b c 104 d WT Pmel
****

WT CD8+ T cell FPKM

100 Ccl5 50 ***
T cells

per 102 CD45+ cells

No. of CD8+ T cells
per 102 CD45+ cells
Gzmb

No. of CD8+ T cells

80 103 40
Gzma hPMEL-AAV
60 30
102 Prf1
40 20
20 101 10
0 0
100
CD8+ T cell Melanocyte 100 101 102 103 104
5 Gy γ Recipient

T
KO

KO
KO CD8+ T cell FPKM

KO

KO
T
W

T
irradiated

W
Vit. induc.

W
Vit. induc. – – + + – – + +

NS f KO KO KO WT
e 2.4 ***

(×104 per cm2)

Skin graft

Melanocytes
KO skin WT skin PDGFRA CD31
1.8
1.2
KO T cell

pSTAT1
0.6

DAPI
0

NS
12 *** Per cent of pSTAT1+ cells

(×104 per cm2)

CD8+ T cells
8 0 20 40 60 80
WT T cell

Fibroblast
4 Endothelial cell
Smooth muscle
0 Immune cell
WT KO T cell
Skin: Melanocyte

KO

T
KO

T
W

W
Melanocyte DAPI DCT CD8 Keratinocyte
T cell: KO WT Langerhans cell

Fig. 2 | Paracrine IFNγ signalling mediated by skin stromal cells drives CD8+
T cell cytotoxicity. a, b, Representative density plot images (a) and FACS
quantification of CD8+ T cells (b) in tail skin epidermis of wild-type (WT) and
Ifngr1 KO mice at day 33 after vitiligo induction (vit. induc.). c, Scatter plot of
differentially expressed genes in skin CD8+ T cells from WT and Ifngr1 KO mice
after vitiligo induction. FPKM, fragments per kilobase of transcript per million
mapped reads. d, Scheme and FACS quantification of CD8+ T cells in
adoptive-transfer-based vitiligo model in wild-type and Ifngr1 KO tail skin
epidermis. e, Schematic diagram, representative whole-mount images and
quantifications of melanocytes and CD8+ T cells in skin graft experiment.
f, Representative immunofluorescent images and quantifications of pSTAT1
signal in different cell types in grafted skin after vitiligo induction in the host
mouse. Scale bars, 500 µm (e), 50 µm (f). For exact P values, see Source Data.
For statistics, P summary and sample sizes, see Methods.

analysis revealed that the cytotoxic genes Gzma, Gzmb, Prf1 and Ccl5
were significantly downregulated in CD8+ T cells isolated from Ifngr1
KO skin compared to wild-type skin (Fig. 2c, Extended Data Fig. 4j).
These results indicated that IFNγ signalling is essential for the skin
recruitment, local aggregation and cytotoxicity of CD8+ T cells that
target epidermal melanocytes.
To further confirm this finding, we used an intervention-free, sponta-
neous model of vitiligo using Pmel TCR transgenic mice; this model gen-
erates antigen-specific CD8+ T cells that recognize melanocytes36,37. The
Pmel;wild-type (Pmel;WT) mice spontaneously developed vitiligo-like
epidermal depigmentation, melanocyte loss and CD8+ T cell infiltration,
whereas the Pmel;Ifngr1 KO mice did not (Extended Data Fig. 4k–p).
To distinguish paracrine versus autocrine IFNγ signalling in orches-
trating the local activity of T cells in vitiligo-affected skin, we used a
model of vitiligo based on the adoptive transfer of T cells20,38, which
relies on sub-lethally irradiated host mice receiving wild-type Pmel
CD8+ T cells and then hPMEL-AAV for the activation of Pmel T cells
(Extended Data Fig. 4q–s). After receiving the same number of trans-
ferred wild-type Pmel T cells (Extended Data Fig. 4t), wild-type recipient
mice developed vitiligo, as shown by skin depigmentation, significant
melanocyte loss and CD8+ T cell infiltration, whereas Ifngr1 KO recipi-
ent mice did not develop vitiligo (Fig. 2d, Extended Data Fig. 4u–x). In
all three different models of vitiligo induction, only significantly more
CD8+ T cells were detected in wild-type skin compared to Ifngr1 KO skin,
whereas the other five main immune cell types showed no difference
(Extended Data Fig. 5a–c).
The adoptive T  cell transfer experiment revealed that parac-
rine IFNγ-signalling is essential for recruiting and orchestrating
auto-reactive CD8+ T cells in skin. To further narrow down whether
IFNγ-responsive skin stromal cells are required for vitiligo progression,
we used a skin graft experiment: full-thickness tail skin was grafted onto
the back of the host mouse, and then host-derived CD8+ T cells that
target melanocytes were activated using the melanoma–Treg-induced
vitiligo method (Extended Data Fig. 5d–g). When wild-type and Ifngr1
KO full-thickness tail skin were grafted onto the same Ifngr1 KO host
mice, the activated Ifngr1 KO CD8+ T cells were only recruited into the
wild-type skin, leading to melanocyte loss, but not into the Ifngr1 KO
skin (Fig. 2e). Thus, given a local environment of wild-type skin, the
Ifngr1 KO auto-reactive CD8+ T cells can be efficiently recruited and
activated, leading to vitiligo development.
When we carried out the reverse graft experiment, the results were
unexpected. When wild-type and Ifngr1 KO full-thickness tail skin were
grafted onto the same wild-type mice, after induction the host-derived
wild-type CD8+ T cells were robustly recruited into both grafted
wild-type and grafted Ifngr1 KO tail skin equivalently, causing a loss of
both wild-type and Ifngr1 KO melanocytes (Fig. 2e). This result directly
contradicted the findings of the adoptive T cell transfer experiment, in
which the transferred wild-type Pmel T cells did not infiltrate the Ifngr1
KO tail skin or lead to melanocyte loss. One possible explanation for
this discrepancy is that wild-type cells from the host skin migrated into
the grafted Ifngr1 KO tail skin, and functionally rescued the KO skin in
terms of driving vitiligo pathogenesis.
A lineage-tracing experiment using C57 tail skin grafted onto
Rosa-mT-host mice showed that host dermal cells (mT+) did indeed
migrate into the grafted skin dermis, but neither host keratinocytes
nor melanocytes did (Extended Data Fig. 5h). After induction of viti-
ligo in the host mice, whereas no pSTAT1 signal was present in the
grafted skin of the KO-to-KO graft, in the KO-to-wild-type graft there
were abundant nuclear pSTAT1 signals in the grafted Ifngr1 KO skin;
quantification showed that most of the pSTAT1+ cells were host-derived
fibroblasts (around 70%) (Fig. 2f, Extended Data Fig. 5i). On the basis
of these results we concluded that IFNγ-responsive skin stromal cells
are required to drive auto-reactive CD8+ T cell-mediated vitiligo in a
paracrine manner.

120 | Nature | Vol 601 | 6 January 2022

 a b c RFP-labelled 8 weeks
PdgfraCreER; Vitiligo model
WT Ifngr1fl/fl fibroblasts
induction

Ifngr1 KO fibroblasts WT fibroblasts

DAPI CD8 DCT

Ifngr1 KO Ifngr1 KO

+ Vitiligo induction
Keratinocyte Melanocyte
Endothelial cell Fibroblast
Myeloid immune cell T cell
DAPI CD8 RFP
No No. of CD8+ T cells per scale
induction + Vitiligo induction No. of CD8+ T cells per mm2
0 100 200 300 400 Fibroblast 0 20 40 60 80 100 120
WT WT TekCre K14Cre 120
cKO cKO WT
Vit.

No. of melanocytes
100 cKO WT
Melanocyte

induc.
80

per scale
+ KO
60 R = –0.85
Csf1rCre TyrCreER Cd4Cre PdgfraCreER P = 2.1 × 10–44
cKO cKO cKO cKO 40 R = –0.14 WT
20 P = 0.06 RFP+ region
–
KO RFP– region
0

d Activated OT-1 Fibroblast-conditioned e f CXCL9+ CXCL10+

CD8+ T cell medium Cells per mm

C CL 0

G A

P
X 3
C CL1

G CL9

LY BP
M 15
IS 5R
G P2
G P4
IL P5
C L5
C L8
C 74

PB
TA 6E
C M
IFNγ

2
0 100 200 300

1
C
C

B
D

B
B
B2

X
3

X
2
C
C
2 250
Progressive
80
Migration ratio (%)

Quiescent 0 200

T cells per mm
60 1× 5× 10× 25× Healthy donor –2 150
40

x 0

p
Ly bp
100

C cl3

Is 5ra
M 15
G cl9
C 74
C cl1

Il1 p5
G p2
G p4

Ta e
pb
C m

C l5
C l8

6
2
Vit. induc.

g
D
B2

b
b
b
c
2
3

x
x
C
C

20 1
– WT 50
+ WT 0
0
IFNγ – + – + – + – + – + – + – + – + + Ifngr1 KO 0
–1
Fibroblast KO WT R = 0.86 R = 0.94
P = 9.4 × 10–7 P = 9.9 × 10–10
g Per cent of CXCL9+ h No. of dermal CD8+ T cells per scale
or CXCL10+ cells
0 20 40 60 80 100 Vit. 0 100 200 300
shRNA H2BRFP induc.
Melanocyte P1 Scrambled – NS
Fibroblast Scrambled NS
+

Endothelial
shCXCL9_1 *
+

Muscle P56
shCXCL9_2 ****
+

Keratinocyte
Vitiligo
M. phagocyte shCXCL10_1 ***
+

CXCL9 induction
Langerhans CXCL10 shCXCL10_2 ****
+

P89
T cell Whole mount RFP+ region RFP– region

Fig. 3 | IFNγ-responsive dermal fibroblasts are uniquely required to fibroblast-conditioned medium. e, Heat map analysis of commonly
orchestrate autoimmune CD8+ T cells through secreted chemokines. a, upregulated secreted factors from vitiligo fibroblasts in human and mice.
Skin illustration and representative density plot images of melanocytes in tail f, Correlation analysis of densities of CD3+ T cells and CXCL9+ and CXCL10+ cells
skin epidermis of mice with the indicated genotypes at day 33 after vitiligo from the skin of patients with vitiligo. g, Quantification of the CXCL9+ and
induction. b, Representative whole-mount images and correlation analysis of CXCL10+ signal in different cell types in the skin of patients with vitiligo.
melanocytes versus CD8+ T cells in tail skin epidermis of wild-type and h, Experimental workflow and quantification of dermal CD8+ T cells from
Pdgfra CreER;Ifngr1 fl/fl cKO mice at day 33 after vitiligo induction. c, Experimental in vivo fibroblast knockdown experiment. P, postnatal day. See Extended Data
workflow, representative whole-mount images and quantification of CD8+ Fig. 4g for scale definition.  Scale bars, 500 µm (b, c). For exact P values, see
T cells in the fibroblast transplant experiment. d, Schematic diagram and Source Data. For statistics, P summary and sample sizes, see Methods.
quantification of the T cell transwell migration assay using

At day 33 of the melanoma–Treg-induced vitiligo model, similar

Vitiligo requires IFNγ-responsive fibroblasts to wild-type mice, TekCre;Ifngr1fl/fl, K14Cre;Ifngr1fl/fl, Csf1rCre;Ifngr1fl/fl,
Next, we used a cell-type-specific knockout strategy to identify the Cd4Cre;Ifngr1fl/fl and TyrCreER;Ifngr1fl/fl mice all developed vitiligo, as indi-
functionally important cell type in skin that drives vitiligo progres- cated by significant loss of melanocytes and CD8+ T cell infiltration in
sion. We ablated Ifngr1 in six different candidate cell types using the skin (Fig. 3a, Extended Data Fig. 6b–e). Only the PdgfraCreER;Ifngr1fl/fl
the cell-type-specific knockout strategy: TekCre;Ifngr1fl/fl mice to tar- mice did not develop vitiligo: skin pigmentation and melanocyte were
get endothelial cells; K14Cre;Ifngr1fl/fl mice to target keratinocytes; not affected, and the number of CD8+ T cells was significantly lower
Csf1rCre;Ifngr1fl/fl mice to target myeloid immune cells; Cd4Cre;Ifngr1fl/fl than that in wild-type and the other conditional knockout (cKO) mice
mice to target T cells; TyrCreER;Ifngr1fl/fl mice to target melanocytes; (Fig. 3b, Extended Data Fig. 6b–f).
and PdgfraCreER;Ifngr1fl/fl mice to target fibroblasts (Fig. 3a). qPCR con- To further validate this discovery, we used two more methods of
firmed the efficient ablation of Ifngr1 in each intended target cell type vitiligo induction. First, the spontaneous vitiligo development in
(Extended Data Fig. 6a). Pmel;WT mice did not occur in Pmel;PdgfraCreER;Ifngr1fl/fl cKO mice,

Nature | Vol 601 | 6 January 2022 | 121

Article
which exhibited no skin depigmentation, melanocyte loss or CD8+ T cell
infiltration (Extended Data Fig. 7a–c). Second, after adoptive-transfer of
an equivalent amount of wild-type Pmel T cells (Extended Data Fig. 7d),
wild-type recipient mice developed vitiligo but PdgfraCreER;Ifngr1fl/fl cKO
recipient mice did not, as shown by the lack of epidermal depigmenta-
tion, melanocyte loss or CD8+ T cell infiltration in skin (Extended Data
Fig. 7e–g). Thus, regardless of the vitiligo induction method used,
PdgfraCreER;Ifngr1fl/fl cKO mice phenocopied the Ifngr1 KO mice and did
not develop vitiligo.
To test whether IFNγ-responsive dermal fibroblasts are not only
required, but also sufficient to induce local aggregation of CD8+ T cells,
we used a fibroblast transplantation assay (Extended Data Fig. 7h).
Intradermal injection of RFP-labelled wild-type—but not Ifngr1 KO—
fibroblasts into the tail skin of Ifngr1 KO mice resulted in significant local
aggregation of CD8+ T cells after the induction of vitiligo (Fig. 3c). This
finding shows that in a local environment that is otherwise completely
non-responsive to IFNγ, IFNγ-responsive fibroblasts alone are sufficient
to mediate the local recruitment of CD8+ T cells.
Fibroblasts recruit T cells through CXCL9/CXCL10
To distinguish whether this essential fibroblast function is depend-
ent on cell–cell contact or secreted factors, we used an in vitro T cell
transwell migration assay using IFNγ-treated fibroblast-conditioned
medium (Fig. 3d, Extended Data Fig. 7i); quantifications showed that
medium from IFNγ-treated wild-type but not Ifngr1 KO fibroblasts could
induce the migration of CD8+ T cells in a concentration-dependent
manner. To identify the secreted factors involved in this process, we
compared the transcriptome analysis of human scRNA-seq data and
mouse RNA-seq data of fibroblasts isolated from vitiligo skin (Extended
Data Fig. 7j, Supplementary Tables 4–6). We identified 28 overlapping
genes that encode fibroblast-specific secreted proteins, 11 of which
are conserved major histocompatibility complex genes and 5 encode
chemokines—namely CCL5, CCL8, CXCL3, CXCL9 and CXCL10 (Fig. 3e).
A T cell transwell migration assay using these secreted chemokines
showed that CXCL9 and CXCL10 induced the migration of CD8+ T cells
in a dosage-dependent manner (Extended Data Fig. 7k).
CXCL9 and CXCL10 have been reported to be involved in vitiligo
and can be expressed by many different cell types24,25,39,40. We found
spatially that the density of CD3+ T cells positively correlated with the
density of CXCL9+ and CXCL10+ cells in patient skin (Fig. 3f, Extended
Data Fig. 7l). Co-localization immunofluorescent analysis revealed that
fibroblasts accounted for around 80% of the total number of CXCL9-
and CXCL10-expressing cells in the skin of patients with vitiligo (Fig,
3g, Extended Data Fig. 7m, n). These results showed that fibroblasts
are the main source of both chemokines during vitiligo progression.
To functionally test whether these fibroblast-derived chemokines are
involved in vitiligo pathogenesis, we used in vivo fibroblast knockdown
experiments41 (Fig. 3h). Intradermal injection of lentivirus expressing
short hairpin RNA (shRNA) resulted in efficient and specific knockdown
of Cxcl9 or Cxcl10 in dermal fibroblasts (Extended Data Fig. 8a–c).
After vitiligo induction, CD8+ T cells were increased in skin expressing
scrambled shRNA, but not in skin expressing Cxcl9 and Cxcl10 shRNA
(Fig. 3h). Correspondingly, melanocytes in the overlaying epidermis
of the Cxcl9- and Cxcl10-shRNA-expressing dermis were also intact
(Extended Data Fig. 8d).
Fluorescence-activated cell sorting (FACS) quantifications of six
types of skin immune cell from wild-type and PdgfraCreER;Ifngr1fl/fl cKO
mice with all three different methods of vitiligo induction showed that
only significantly fewer CD8+ T cells were detected in the cKO compared
to the wild-type skin, whereas all of the other immune cells did not
show a significant difference (Extended Data Fig. 8e). CXCR3 is the
confirmed receptor for CXCL9 and for CXCL1042. Quantifications in
skin from both patients with vitiligo and vitiligo-affected mice showed
that CD8+ T cells constitute the majority of the CXCR3+ immune cells
122 | Nature | Vol 601 | 6 January 2022
(Extended Data Fig. 8f, g). Together, these in vitro and in vivo results
show that IFNγ-responsive fibroblasts mediate local aggregation of
CD8+ T cells through the CXCL9/CXCL10–CXCR3 axis.
Regional fibroblasts differ in IFNγ response
At the whole-organ level, dermal fibroblasts are not homogeneous:
fibroblasts from different anatomic skin regions are intrinsically dif-
ferent in their expression of HOX genes41,43. However, whether skin
fibroblasts from different anatomic regions are different in regulating
the activity of cutaneous immune cells is unknown.
To investigate this, we first quantified the frequencies of viti-
ligo lesions at different anatomic regions of 2,265 patients with
non-segmental vitiligo (Supplementary Table 7). We quantified eight
main body regions and found large variations in the frequency of vitiligo
among them, with the back of the hand, the chest and the back being
the most susceptible to vitiligo, and the palm and arm the least (Fig. 4a).
Next, we conducted an RNA-seq analysis of in vitro IFNγ-treated
primary human dermal fibroblasts isolated from these eight anatomic
regions. The patterns of HOX gene expression indicated that fibro-
blasts retained their positional information during in vitro culture
(Extended Data Fig. 9a–c, Supplementary Table 8). Heat map and
qPCR analysis revealed that the numbers and types of upregulated
genes in response to IFNγ varied significantly among region-specific
fibroblasts (Fig. 4b). Although fibroblasts from all anatomic regions
expressed similar levels of the IFNγ signalling pathway genes IFNGR1,
JAK1, JAK2 and STAT1 (Extended Data Fig. 9d), multiple IFNγ-induced
chemokine genes—including CXCL9, CXCL10 and their human-specific
variant CXCL11—were significantly upregulated in fibroblasts that were
isolated from anatomic regions with a higher incidence of vitiligo, such
as the back of the hand, the back of the foot, the chest and the back
(Fig. 4c, Extended Data Fig. 9e). Further correlation analysis showed
that the percentage of vitiligo incidence positively correlated with the
intrinsic IFNγ response of dermal fibroblasts at different anatomic
regions (Fig. 4d).
Distinct fibroblasts drive vitiligo patterns
To study the functional relevance of this correlation, we turned to the
melanoma–Treg-induced mouse model of vitiligo. In different anatomic
regions of mouse skin, melanocytes are ubiquitously present in hair
follicles but not in epidermis, so we only quantified melanocytes in
hair follicles at four anatomic regions with a similar hair follicle density
(Extended Data Fig. 9f): dorsal, ventral, paw dorsal and paw ventral. In
human vitiligo, hair depigmentation tends to occur after prolonged
pathogenesis6,44, which was also observed in the skin of vitiligo-affected
mice (Extended Data Fig. 2b). We therefore quantified the number of mel-
anocytes in hair follicles after long-term vitiligo (at day 300). Melanocytes
in dorsal and ventral skin hair follicles were depleted in vitiligo-induced
mice compared to control mice, but melanocytes in paw dorsal and paw
ventral skin hair follicles remained intact; correspondingly, increases of
CD3+CD8+ T cells were significantly higher in the dorsal and ventral skin
than in paw dorsal and paw ventral skin (Fig. 4e, Extended Data Fig. 9g,
h). These results revealed that there are differences in vitiligo occurrence
among different anatomic regions in mice, as in humans.
Notably, we found that Cxcl9 and Cxcl10 showed higher expression
levels in dorsal and ventral fibroblasts compared to paw dorsal and
paw ventral fibroblasts either from in-vivo-isolated cells in long-term
vitiligo mice, or in IFNγ-treated cultured cells in vitro (Fig. 4f, Extended
Data Fig. 9i). To study the functional relevance of this expression level
difference, we conducted a T cell transwell migration assay with con-
ditioned medium. The results showed that medium from IFNγ-treated
dorsal skin fibroblasts induced significantly more (more than two times
as much) CD8+ T cell migration than did medium from paw ventral skin
fibroblasts (Fig. 4g, Extended Data Fig. 9j). Our in vivo data showed that
a Vitiligo b
Site No. incidence CXCL10 CXCL11 CXCL9 c 1.5 CXCL9
CXCL10

relative to GAPDH
Hand back 558 24.64% Hand back
Chest 465 20.53%

Expression
Foot back 1.0
Back 460 20.31% Chest
Leg 405 17.88% Back
Foot back 298 13.16% 0.5
Leg
Head 296 13.07%
Arm 201 8.87% Arm
Palm 107 4.72% Head 0
Palm
Total 2,265

ba k
C ck
Ba st
Le k
Ar g
H m
Pa d
lm
ot ac

ea
he
Fo d b
2 0 –2

an
H
T cells (%) per hair follicle
d e NS f

Melanocytes
DAPI DCT P-cad 8 **** 8 Cxcl9 Cxcl10
0.905 Foot back 6

Control
R = 0.62 Chest
Enrichment score

realtive to Ppib
0.900 2 6
Back

Expression
0
0.895 Hand back 4
Arm 15 ** ***

Day 300
Leg 10

CD8+
0.890 Head Vitiligo 2
5
0.885 Palm 0 0
5 10 15 20 25 Paw Dorsal Vit. induc. – + – +

d o aw

ve Paw

al

l
ra
Vitiligo incidence (%)

al

Do l
Dorsal

rs
ra

nt
P
rs

nt

Ve
g h DAPI CD8 RFP i j
60 * Cxcl9 Cxcl10
1× 2× 5× 20 **** T0
Ifngr1 KO

25 6
Enrichment score
Migration ratio

Dorsal

15

realtive to Ppib
40 20 5
T200

Expression
(%)

10 15 4
20 3
5 10 T350
Paw dorsal

2
Fibroblast

5 1
0 0 Normal chemotactic fibroblast
0 0
IFNγ – + – + – + – + – + – + Weak chemotactic fibroblast
Paw dorsal Dorsal Activated fibroblast
Fibrolast Paw ventral Dorsal CD8+ T cell Melanocyte
+ Vitiligo induction

k Scrambled shRNA JAK1 shRNA

300
CD8+ T cells per scale

Vitiligo
induction shRNA
+ Vitiligo induction

No. of dermal

200 – Scrambled
+ Scrambled
100 + Ifngr1
+ Jak1
+ Stat1
0
0 25 50 75 100
DAPI CD8 CD8+ T cell DAPI CD8 CD8+ T cell Infected fibroblasts per scale (%)
RFP RFP+ fibroblast RFP RFP+ fibroblast

Fig. 4 | Anatomically distinct fibroblast subsets with differential IFNγ
responses determine organ-level skin autoimmune patterns.
a, Quantification of lesion frequencies at anatomically distinct skin regions in
patients with non-segmental vitiligo. b, Heat map of 92 upregulated genes in
IFNγ-treated human dermal fibroblasts from eight anatomical regions. c, qPCR of
CXCL9 and CXCL10 from IFNγ-treated human dermal fibroblasts in culture.
d, Correlation analysis of fibroblasts IFNγ response enrichment score versus
vitiligo incidence frequency. e, Representative immunofluorescent images and
quantifications of melanocytes and CD8+ T cells in different skin regions in
control and vitiligo-affected mice at day 300. f, qPCR of Cxcl9 and Cxcl10 in
IFNγ-treated-region-specific mouse dermal fibroblasts in culture.
g, Quantification of migration ratio from T cell transwell migration assay using
IFNγ-treated fibroblast-conditioned medium. h, Representative whole-mount
images and quantification of enrichment score of CD8+ T cells in fibroblast-
injected area of Ifngr1 KO mice after vitiligo induction. i, qPCR of Cxcl9 and Cxcl10
in re-isolated fibroblasts from h. j, Mathematical model predicts the regional
distribution pattern of T cells with mosaic IFNγ-responsive fibroblasts.
k, Representative whole-mount and density plot images and correlation analysis
of the number of T cells versus the percentage of infected fibroblasts in the in vivo
mosaic fibroblast knockdown experiment. See Extended Data Fig. 4g for scale
definition. Scale bars, 50 µm (e), 500 µm (h, k). For exact P values, see Source
Data. For statistics, P summary and sample sizes, see Methods.

a twofold increase in the number of T cells could result in significant To further test whether regional variation in fibroblasts could result
differences in melanocyte loss (Extended Data Fig. 9k). in patterned T cell activity, we built a mathematical model to study the
Next to directly compare the region-specific fibroblasts in CD8+ T cell effect of mosaic fibroblasts on T cell activity (Extended Data Fig. 10a–c,
recruitment in vivo, we intradermally injected RFP-labelled dorsal and paw Supplementary Videos 1–6). In a field of fibroblasts composed of nor-
dorsal wild-type fibroblasts into Ifngr1 KO mice tail skin, then conducted mal and weak chemotactic regions (Fig. 4j, Extended Data Fig. 10d),
vitiligo model induction in the host mice. Quantification showed that there T cells were competitively recruited into the normal chemotactic
was a significantly stronger enrichment of CD8+ T cells in dorsal fibro- region, leading to preferential expansion of T cell clusters in those
blast area than in paw fibroblast area (Fig. 4h). In addition, the re-isolated regions and corresponding melanocyte loss (Supplementary Video 7).
dorsal fibroblasts showed 18 times higher expression of Cxcl9 and 4 times To functionally test this model, we used the fibroblast knockdown
higher expression of Cxcl10 than did paw dorsal fibroblasts (Fig. 4i). approach to create a mosaic field containing fibroblasts with different

Nature | Vol 601 | 6 January 2022 | 123

Article
IFNγ-responsive abilities. By using shRNA against Ifngr1, Jak1 or Stat1
with high knockdown efficiency we could efficiently block IFNγ down-
stream signal in fibroblasts in vitro (Extended Data Fig. 10e, f). Quan-
tification showed that in dermis expressing scrambled shRNA, CD8+
T cells infiltrated both the RFP+ and RFP− regions equivalently, but in
dermis expressing shRNAs against IFNGR1, JAK1 or STAT1, the number
of CD8+ T cells negatively correlated with the percentage of infected
fibroblasts. Correspondingly, in the overlaying epidermis, the number
of melanocytes positively correlated with the percentage of infected
fibroblasts (Fig. 4k, Extended Data Fig. 10g–i). The fibroblast mosaic
knockdown experiments validated the modelling prediction that in a
field with uneven fibroblast responses to IFNγ, CD8+ T cells preferen-
tially aggregate towards a region with fibroblasts capable of stronger
IFNγ responses, hence generating a patterned loss of melanocytes.
Discussion
Here we have identified a feedforward system between mobile CD8+
T cells and immobile fibroblasts, which functions to prioritize the allo-
cation of collective cytotoxicity and efficient elimination of target cells.
The cyclical nature of the feedforward loop leads to the progressive
aggregation of CD8+ T cells in areas with IFNγ-responsive fibroblasts,
as well as sharply declined CD8+ T cells in areas with IFNγ-resistant
fibroblasts. Many other autoimmune skin diseases in addition to viti-
ligo exhibit regional preference with symmetric distributions45–50. The
immune cell activity regulated by regional fibroblasts could provide a
general paradigm through which the progression and regional pattern
of skin autoimmune diseases are regulated. Besides the skin, fibro-
blasts are ubiquitously present in essentially all organs of our body;
our study also provides avenues for investigation of the potentially
broad application of this cellular cross-talk.
Online content
Any methods, additional references, Nature Research reporting sum-
maries, source data, extended data, supplementary information,
acknowledgements, peer review information; details of author con-
tributions and competing interests; and statements of data and code
availability are available at https://doi.org/10.1038/s41586-021-04221-8.
1. Larsen, S. B., Cowley, C. J. & Fuchs, E. Epithelial cells: liaisons of immunity. Curr. Opin.
Immunol. 62, 45–53 (2019).
2. Rognoni, E. & Watt, F. M. Skin cell heterogeneity in development, wound healing, and
cancer. Trends Cell Biol. 28, 709–722 (2018).
3. Afshar, M. & Gallo, R. L. Innate immune defense system of the skin. Vet. Dermatol. 24,
32–38 (2013).
4. Taieb, A. & Picardo, M. Vitiligo. N. Engl. J. Med. 360, 160–169 (2009).
5. Alikhan, A., Felsten, L. M., Daly, M. & Petronic-Rosic, V. Vitiligo: a comprehensive overview
Part I. Introduction, epidemiology, quality of life, diagnosis, differential diagnosis,
associations, histopathology, etiology, and work-up. J Am. Acad. Dermatol. 65, 473–491
(2011).
6. Picardo, M. et al. Vitiligo. Nat. Rev. Dis. Primers 1, 15011 (2015).
7. Ezzedine, K., Eleftheriadou, V., Whitton, M. & van Geel, N. Vitiligo. Lancet 386, 74–84
(2015).
8. Taieb, A. Intrinsic and extrinsic pathomechanisms in vitiligo. Pigment Cell Res. 13, 41–47
(2000).
9. Spritz, R. A. & Santorico, S. A. The genetic basis of vitiligo. J. Invest. Dermatol. 141,
265–273 (2021).
10. Frisoli, M. L., Essien, K. & Harris, J. E. Vitiligo: mechanisms of pathogenesis and treatment.
Annu. Rev. Immunol. 38, 621–648 (2020).
11. Le Poole, I. C., van den Wijngaard, R. M., Westerhof, W. & Das, P. K. Presence of T cells and
macrophages in inflammatory vitiligo skin parallels melanocyte disappearance. Am. J.
Pathol. 148, 1219–1228 (1996).
12. Ogg, G. S., Rod Dunbar, P., Romero, P., Chen, J. L. & Cerundolo, V. High frequency of
skin-homing melanocyte-specific cytotoxic T lymphocytes in autoimmune vitiligo. J. Exp.
Med. 188, 1203–1208 (1998).
13. Van den Boorn, J. G. et al. Autoimmune destruction of skin melanocytes by perilesional
T cells from vitiligo patients. J. Invest. Dermatol 129, 2220–2232 (2009).
14. Maeda, Y. et al. Detection of self-reactive CD8+ T cells with an anergic phenotype in
healthy individuals. Science 346, 1536–1540 (2014).
124 | Nature | Vol 601 | 6 January 2022
15. Cheuk, S. et al. CD49a expression defines tissue-resident CD8+ T cells poised for
cytotoxic function in human skin. Immunity 46, 287–300 (2017).
16. Iannella, G. et al. Vitiligo: pathogenesis, clinical variants and treatment approaches.
Autoimmun. Rev. 15, 335–343 (2016).
17. Rodrigues, M. et al. Current and emerging treatments for vitiligo. J. Am. Acad. Dermatol.
77, 17–29 (2017).
18. Kovacs, S. O. Vitiligo. J. Am. Acad. Dermatol. 38, 647–666 (1998).
19. Gregg, R. K., Nichols, L., Chen, Y., Lu, B. & Engelhard, V. H. Mechanisms of spatial and
temporal development of autoimmune vitiligo in tyrosinase-specific TCR transgenic
mice. J. Immunol. 184, 1909–1917 (2010).
20. Harris, J. E. et al. A mouse model of vitiligo with focused epidermal depigmentation
requires IFN-gamma for autoreactive CD8+ T-cell accumulation in the skin. J. Invest.
Dermatol. 132, 1869–1876 (2012).
21. Chatterjee, S. et al. A quantitative increase in regulatory T cells controls development
of vitiligo. J. Invest. Dermatol. 134, 1285–1294, (2014).
22. Yang, L. et al. Interferon-gamma inhibits melanogenesis and induces apoptosis in
melanocytes: a pivotal role of CD8+ cytotoxic T lymphocytes in vitiligo. Acta Derm.
Venereol. 95, 664–670 (2015).
23. Wang, S. Q. et al. Interferon-γ induces senescence in normal human melanocytes.
PLoS One 9, e93232 (2014).
24. Tulic, M. K. et al. Innate lymphocyte-induced CXCR3B-mediated melanocyte apoptosis
is a potential initiator of T-cell autoreactivity in vitiligo. Nat. Commun. 10, 2178 (2019).
25. Richmond, J. M. et al. Keratinocyte-derived chemokines orchestrate T-cell positioning in
the epidermis during vitiligo and may serve as biomarkers of disease. J. Invest. Dermatol.
137, 350–358 (2017).
26. Biswas, K. B. et al. GPNMB is expressed in human epidermal keratinocytes but disappears
in the vitiligo lesional skin. Sci. Rep. 10, 4930 (2020).
27. Boukhedouni, N. et al. Type-1 cytokines regulate matrix metalloprotease-9 production
and E-cadherin disruption to promote melanocyte loss in vitiligo. JCI Insight 4, e133772
(2020).
28. Ganju, P. et al. Microbial community profiling shows dysbiosis in the lesional skin of
vitiligo subjects. Sci Rep. 6, 18761 (2016).
29. Yuan, X. et al. The impact of NBUVB on microbial community profiling in the lesional skin
of vitiligo subjects. Microb. Pathog. 140, 103943 (2020).
30. Lang, K. S. et al. HLA-A2 restricted, melanocyte-specific CD8+ T lymphocytes detected in
vitiligo patients are related to disease activity and are predominantly directed against
MelanA/MART1. J. Invest. Dermatol. 116, 891–897 (2001).
31. Namazi, M. R. Neurogenic dysregulation, oxidative stress, autoimmunity, and
melanocytorrhagy in vitiligo: can they be interconnected? Pigment Cell Res. 20,
360–363 (2007).
32. Lotti, T., D’Erme, A. M. & Hercogova, J. The role of neuropeptides in the control of regional
immunity. Clin. Dermatol. 32, 633–645 (2014).
33. Palermo, B. et al. Specific cytotoxic T lymphocyte responses against Melan-A/MART1,
tyrosinase and gp100 in vitiligo by the use of major histocompatibility complex/peptide
tetramers: the role of cellular immunity in the etiopathogenesis of vitiligo. J. Invest.
Dermatol. 117, 326–332 (2001).
34. Mantovani, S. et al. Dominant TCR-α requirements for a self antigen recognition in
humans. J. Immunol. 169, 6253–6260 (2002).
35. Byrne, K. T., Zhang, P., Steinberg, S. M. & Turk, M. Autoimmune vitiligo does not require
the ongoing priming of naive CD8 T cells for disease progression or associated
protection against melanoma. J. Immunol. 192, 1433–1439 (2014).
36. Mosenson, J. A. et al. Mutant HSP70 reverses autoimmune depigmentation in vitiligo.
Sci. Transl. Med. 5, 174ra128 (2013).
37. Miao, X. et al. PD-L1 reverses depigmentation in Pmel-1 vitiligo mice by increasing the
abundance of Tregs in the skin. Sci Rep. 8, 1605 (2018).
38. Riding, R. L., Richmond, J. M. & Harris, J. E. Mouse model for human vitiligo. Curr. Protoc.
Immunol. 124, e63 (2019).
39. Richmond, J. M. et al. Resident memory and recirculating memory T cells cooperate
to maintain disease in a mouse model of vitiligo. J. Invest. Dermatol. 139, 769–778
(2019).
40. Rashighi, M. et al. CXCL10 is critical for the progression and maintenance of
depigmentation in a mouse model of vitiligo. Sci. Transl. Med. 6, 223ra223 (2014).
41. Yu, Z. et al. Hoxc-dependent mesenchymal niche heterogeneity drives regional hair
follicle regeneration. Cell Stem Cell 23, 487–500 (2018).
42. Groom, J. R. & Luster, A. D. CXCR3 in T cell function. Exp. Cell. Res. 317, 620–631
(2011).
43. Chang, H. Y. et al. Diversity, topographic differentiation, and positional memory in
human fibroblasts. Proc. Natl Acad. Sci. USA. 99, 12877–12882 (2002).
44. Paus, R., Ito, N., Takigawa, M. & Ito, T. The hair follicle and immune privilege. J. Investig.
Dermatol. Symp. Proc. 8, 188–194 (2003).
45. Farber, E. M., Nickoloff, B. J., Recht, B. & Fraki, J. E. Stress, symmetry, and psoriasis:
possible role of neuropeptides. J. Am. Acad. Dermatol. 14, 305–311 (1986).
46. Boehncke, W.-H. & Schön, M. P. Psoriasis. Lancet 386, 983–994 (2015).
47. Greb, J. E. et al. Psoriasis. Nat. Rev. Dis. Primers 2, 16082 (2016).
48. Gabrielli, A., Avvedimento, E. V. & Krieg, T. Scleroderma. N. Engl. J. Med. 360, 1989–2003,
(2009).
49. Kaul, A. et al. Systemic lupus erythematosus. Nat. Rev. Dis. Primers 2, 16039 (2016).
50. Boyd, A. S. & Neldner, K. H. Lichen planus. J. Am. Acad. Dermatol. 25, 593–619 (1991).
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
© The Author(s), under exclusive licence to Springer Nature Limited 2021
Methods
Ethics statement
The scRNA-seq and immunofluorescent analysis of human vitiligo skin
biopsies study were reviewed and approved by the Ethics Committee
of Beijing Hospital. All donors signed the written informed consent for
sample collection and data analysis. For minors, informed consent was
obtained from the legally authorized representatives of the minors.
Before signing the consent, necessary information including the goals
and related experimental procedures for the study were provided to
the donors. In this study, we collected 3 × 3-mm2 skin biopsies from
treatment-naive patients with vitiligo and control healthy donors who
had surgical operations. The biopsies were then digested into single
cells for further scRNA-seq analysis. For the representative skin lesion
pictures shown in Extended Data Fig. 1a, patients provided written
informed consent and agreed to allow photographs of the selected
area to be taken and used.
The study of anatomically distinct human dermal fibroblasts was
reviewed and approved by the Peking University Third Hospital Medi-
cal Science Research Ethics Committee. All donors signed the written
informed consent for sample collection. In this study, we collected skin
biopsies from human embryos between 22 to 24 weeks of gestation
(weeks after estimated fertilization time). Skin biopsies from anatomi-
cally distinct regions were collected, including the back of the hand,
back of the foot, palm, head, chest, back, leg and arm. Fibroblasts were
isolated from skin biopsies. Informed consent was obtained from all
couples. Before giving consent, necessary information including the
goals and related experimental procedures for the study were provided
to the couples.
Samples from patients with vitiligo
Seven male and three female patients clinically diagnosed with vitiligo
were enrolled in this study. Their ages ranged from 6 to 55, with a median
age of 24. Among these patients, four were diagnosed as being in the
active stage, and the disease stage was uncertain for the other six. The
clinical information regarding disease state is empirically based on
observations of whether the white macules or patches have blurred
boundaries. In addition to this, it relies on feedback from patients about
whether or not the depigmentation areas were progressively expanding
in the last month prior to hospital visits. All of the patients were newly
diagnosed with vitiligo and none of the patients had received therapy
or UVB treatment. A 3 × 3-mm2 biopsy was taken from each patient in
the junction region between the depigmented lesion region and pig-
mented perilesional region. A biopsy of the same size was also taken
from five healthy donors with other surgical operations as a control.
The clinical characteristics of these patients and healthy donors are
summarized in Extended Data Fig. 1c.
Mice
Mice were bred and maintained in the National Institute of Biological
Sciences (NIBS) specific-pathogen-free facility in accordance with the
Guide for the Care and Use of Laboratory Animals of NIBS. Procedures
were approved by the Laboratory Animal Management Committee of
NIBS and were in compliance with all relevant ethical regulations. All
mice used in experiments were socially housed under a 12-h light–dark
cycle with free access to food and water, with 23 °C–25 °C and 50%–56%
humidity.
C57BL/6 mice were purchased from Charles River Laboratories. K14Cre
(MGI: 1926500) and K14-H2BGFP (MGI:5286142) mice were provided
by E. Fuchs. Ifngr1 KO mice ( Jackson stock no. 003288) and TekCre mice
( Jackson stock no. 004128) were provided by F. Shao. OT-1 mice (stock
no. 003831) were provided by L. Chen. Pmel TCR transgenic mice (stock
no. 005023), PdgfraCreER (stock no. 018280), Csf1rCre (stock no. 021024),
TyrCreER (stock no. 012328), Cd4Cre (stock no. 017336), Foxp3Cre (stock no.
008694), Ifngr1 flox (stock no. 025394), Ai14 (stock no. 007914) Ai6
(stock no. 007906) and Rosa-stop-mTmG (stock no. 007676) mice were
obtained from The Jackson Laboratory. For CreER activation, Tamoxifen
was dissolved in sunflower oil with 10% ethanol. PdgfraCreER;Ifngr1fl/fl
and TyrCreER;Ifngr1fl/fl mice received an intraperitoneal injection of
100 μl of 20 mg ml−1 tamoxifen solution every other day from P42 to
P56. Knockout efficiency was validated by cell-type-specific qPCR. It
needs to be mentioned here that TekCre labels both endothelial cells
and haematopoietic cells51, so Ifngr1 expression was ablated in both
cell types in the TekCre;Ifngr1fl/fl mice. Although Csf1r mainly expresses
and functions in myeloid cells, the Cre activity of Csf1rCre is detected
in multiple immune cell types including macrophages, dendritic cells,
monocytes, granulocytes and T and B cells52,53. Ifngr1 expression was
therefore ablated in the immune cells including macrophages and
dendritic cells in Csf1rCre;Ifngr1fl/fl mice.
Cell culture
All of the cell lines were cultured at 37 °C in a cell incubator with 5% CO2.
B16F10 cells (ATCC, CRL-6475) were maintained in DMEM medium sup-
plemented with 10% (v/v) FBS and 1% (v/v) pen–strep. Primary mouse
fibroblasts were maintained in DMEM supplemented with 10% (v/v)
FBS, 1% (v/v) pen–strep, and 1× antibiotic–antimycotic. The 293FT
cells (Thermo Fisher Scientific, R70007) used for virus packaging were
maintained in DMEM medium supplemented with 10% (v/v) FBS, 1% (v/v)
pen–strep, 1% (v/v) l-glutamine, 1% (v/v) 100 mM sodium pyruvate, 1%
(v/v) 7.5% sodium bicarbonate and 500 µg ml−1 G418. All cell lines tested
negative for mycoplasma contamination.
Single-cell collection from human skin biopsies
Tissue processing and the single-cell purification process started within
3 h after surgical skin biopsy collection. After removal of subcutaneous
fat, skin tissue was placed dermis side down in 6 ml of 2.4 U ml−1 disease
solution and incubated at 37 °C with shaking at 80 rpm for 70 min. Epi-
dermis with hair follicles was carefully separated from the dermis using
forceps. The separated epidermis was then placed dermis side down
in 6 ml of 0.25% trypsin at 37 °C for 10 min, and afterwards 6 ml of 5%
FBS medium was added to neutralize trypsin. The epidermal single-cell
suspension was obtained by repeatedly aspirating and dispensing
the solution with a 1-ml pipette 10 times, and then filtered through
two strainers (70 μm followed by 40 μm). Cells were then stained with
anti-human CD45 (BD, cat: 11-9459-41, clone: 2D1, dilution:1:300) and
anti-human CD117 (BD, cat 555714, clone: YB5.B8, dilution:1:300) for
15 min to detect immune cells and melanocytes, and then washed. The
dermis was placed in 10 ml of 2 mg ml−1 collagenase (Sigma-Aldrich,
C2674) and incubated at 37 °C with shaking at 80 rpm for 1 h; after-
wards, 10 ml of 5% FBS medium was added. Single-cell suspensions
were obtained by repeatedly aspirating and dispensing with a pipette
for 10 times. The cells were then filtered through two strainers (70 μm
followed by 40 μm), and then stained for 15 min with anti-human CD45
(BD, 11-9459-41, clone: 2D1, dilution:1:300) for immune cell detection
and then washed.
On the basis of FACS analysis, single cells of different subtypes,
including immune cells (CD45+) from both the epidermis and dermis,
melanocytes (CD117+) from the epidermis, dermal cells (CD45−) from
the dermis and keratinocytes (CD45−, CD117−) from the epidermis were
sorted into collection buffer, which contains 0.04% BSA in PBS to mini-
mize cell losses and aggregation. For each skin biopsy sample, all of
the immune cells, melanocytes and dermal cells, as well as an equal
number of keratinocytes to all of the other three cell types combined,
were pooled together for the next step of single-cell analysis.
Vitiligo induction in mice
Human vitiligo pathogenesis is driven by diverse CD8+ cytotoxic T cells
recognizing different endogenous melanocyte-specific antigens,
including MART-126-35 and TYR369-37754. Clinically it was discovered that
melanoma immunotherapy activates CD8+ T cells recognizing these
Article
antigens shared by melanoma cells and melanocytes55, which results in
vitiligo occurrence in 13.5%–25.7% of patients with melanoma receiving
immunotherapy56–58. In addition, loss of Treg cells that can induce anergy
of autoimmune CD8+ T cells is likely to contribute to vitiligo occurrence
in humans. In mouse studies, vitiligo-like hair coat depigmentation has
been reported after B16 melanoma inoculation followed by CD4+ T cell
depletion, but epidermal melanocyte loss—which is the hallmark of
human vitiligo pathology—has not been studied. On the basis of these
previous reports, we developed a melanoma–Treg-induced protocol by
using B16F10 melanoma inoculation coupled with Treg cell depletion to
activate endogenous auto-reactive CD8+ T cells targeting epidermal
melanocytes, and induced vitiligo occurrence in mice. First the dorsal
skin in the right flank of 8–9-week old C57 mice was intradermally inocu-
lated with 2 × 105 B16F10 melanoma cells (day 0), then CD4 depletion
antibodies were injected on days 4 and 10. Only mice that developed
primary tumours were used for further analysis. Primary tumours were
surgically removed on day 12. Spontaneous tumour metastases were
not observed with this B16F10 cell line, and mice with recurrent primary
tumours after surgery were not used for further study.
Melanocytes in mouse dorsal skin are located in hair follicles but
not in the epidermis; only in mouse tail skin are melanocytes located
in both the hair follicle and epidermis similar to human skin. Skin epi-
dermis depigmentation is the defining feature of vitiligo pathology,
therefore only mouse tail skin was used for vitiligo analysis, unless
otherwise specified. The turnover rate of skin epidermis is about 30
days59. Epidermal melanocyte loss peaked at day 30 after induction,
and epidermis depigmentation was visually apparent by day 60 after
induction. Tail skin epidermal depigmentation was initially patchy
and then progressed to eventually cover the entire epidermal surface,
which did not recover even more than a year later. Similar to patients
with vitiligo, there were no symptoms of skin inflammation in this mela-
noma–Treg-induced vitiligo mouse model, including redness, swelling
or epidermal scaliness.
The adoptive-transfer-based vitiligo mouse model was generated
according to the protocol described previously. In brief, naive Pmel
TCR transgenic CD8+ T cells from spleen were magnetically isolated by
CD8 selection according to the manufacturer’s instructions (Miltenyi).
Purified CD8+ T cells were then transferred intravenously at 106 cells per
mouse. Recipient mice were sub-lethally irradiated (5 Gy, Gammacell
1000) at 24 h prior to the transfer and received AAV-expressing cog-
nate antigen hPMEL (106 pfu) at 30 min after adoptive T cell transfer.
Whole-mount immunofluorescent staining of tail skin epidermis and
FACS analysis of both tail skin and spleen CD8+ T cells were performed
at day 26 after adoptive transfer. Mice that were sub-lethally irradiated
but did not receive any treatment or received only AVV-hPMEL or Pmel
CD8+ T cell transfer were used as controls.
We used three different vitiligo induction models to validate the
functional importance of fibroblasts. It should be pointed out here
that there are some key differences in these models. In the melanoma–
Treg-induced vitiligo mouse model, Treg cells were initially depleted,
whereas in the adoptive-transfer-based vitiligo model, the number of
functional CCR5+ Treg cells was reported to be increased during vitiligo
development60. Future studies using mouse models about the function
of Treg cells and other cell types in driving vitiligo initiation and progres-
sion should take these differences into consideration.
Administration of CD4 and CD8 depletion antibodies
CD4 antibody for cell depletion was purchased from BioXcell. CD8
antibody for cell depletion was developed by the laboratory of J.S.
Anti-CD4 (BioXcell, clone: GK1.5) and anti-CD8 (clone: 2.43) were admin-
istered by intraperitoneal injection in doses of 10 μg per g mouse body
weight. More than 99% depletion of target populations was confirmed
by FACS. CD4 antibody was administered on day 4 and day 10 after
B16F10 inoculation. CD8 antibody was administered every fourth day
after the tumour was removed.
Tail skin graft
Full-thickness tail skins from 4–6-week-old female mice were removed,
flattened and cut into 1 × 1-cm squares. Then two 1 × 1-cm square skin
pieces were cut off from the back of anaesthetized 8-week-old female
recipient mice with the indicated genotype. Donor skin pieces were
placed onto the backs of anaesthetized 8-week-old female recipient
mice with the indicated genotypes, with each recipient receiving a
wild-type and KO graft. Grafts were secured by sterile gauze and elastic
bandages, which were removed at 8–10 days later. Note that we only
used female mice for both donor and recipient, because sex-unmatched
graft would result in skin necrosis within 20 days. For C57 to mTmG or
CD4Cre;mTmG grafts, the transgenic host mice were crossed to C57 to
avoid immune rejection.
Immunofluorescence staining
For section staining, tissues were embedded in OCT compound, fro-
zen, cryosectioned (20–30 µm), and fixed for 10 min in 4% paraform-
aldehyde in PBS. Sections were washed in PBS overnight and then
permeabilized for 20 min in 0.3% H2O2 in methanol at −20 °C and
blocked for 1 h in a solution of 2% normal donkey serum, 1% BSA and
0.3% Triton in PBS at room temperature. The signal of human pSTAT1,
human CD8, human PDGFRA, human CD11c, human langerin, human
CD3 and mouse pSTAT1 was amplified by the ABC Kit (Vector) and TSA
Kit (Thermo Fisher Scientific). The following antibodies were used:
anti-human TYR (made by the T.C. laboratory; 1:500), anti-human DCT
(made by the T.C. laboratory; 1:500), anti-mouse KRT14 (made by the
T.C. laboratory; 1:1,000), anti-human pSTAT1 (CST, cat: 9167, clone:
58D6, dilution: 1:500), anti-human PDGFRA (Abcam, cat: ab203491,
clone: EPR22059-270, dilution: 1:500), anti-human CD31 (Abcam,
cat: ab24590, clone: P2B1, dilution: 1:500), anti-human aSMA (Sigma,
cat: C6198, clone: 1A4, dilution: 1:500), anti-human CD11c (Abcam,
cat: ab 52632, clone: EP1347Y, dilution: 1:500), anti-human langerin
(Abcam, cat: ab192027, clone: EPR15863, dilution: 1:500), anti-human
CD3 (BD, cat: 563423, clone: UCHT1, dilution: 1:300), anti-human CD8
(BD, cat: 561952, clone: RPA-T8, dilution: 1:300). anti-human FOXP3
(Abcam, cat: ab20034, dilution: 1:300), anti-human KLRB1 (Biosource,
cat: MBS2013141, dilution: 1:300), anti-human CD40LG (Biosource,
cat: MBS4381431, dilution: 1:300), anti-human GZMA (Abcam, cat:
ab10870, dilution:1:300), anti-mouse aSMA (Sigma, cat: C6198, clone:
1A4, dilution: 1:500), anti-mouse pSTAT1 (CST, cat: 9167, clone:58D6,
dilution: 1:500), anti-mouse CD31 (eBioscience, cat: 25-0311-81, clone:
390, dilution: 1:500). anti-mouse PDGFRA (R&D, cat: AF-307, clone:
APA5, dilution: 1:300) and anti-mouse CD45 (BD, cat: 559864, clone:
10-F11, dilution: 1:300).
For whole-mount staining, hair shafts that would interfere with
immunofluorescent imaging process were first removed using Nair
treatment. Then the entire tail skin was collected and flattened. Only
the middle one-third section along the long axis of tail skin was used
for whole-mount staining for the purpose of consistency. Flattened
skins were cut into 1 × 1-cm squares and placed in 20 mM EDTA solution
and incubated at 37 °C with shaking at 80 rpm for 1 h. The epidermis
was separated from the dermis with a fine-tipped tweezer. As hair fol-
licles obscure the view of epidermis during whole-mount analysis,
to leave hair follicles in the dermis, the epidermis was peeled from
the dermis along the posterior–anterior direction in a switch motion.
Occasional hair follicles still attached to the epidermis were removed
using a tweezer. The epidermis was then flattened and fixed for 10 min in
4% paraformaldehyde in PBS. Whole-mount skin samples were washed
in PBS overnight and then permeabilized for 20 min in 0.3% H2O2 in
methanol at −20 °C and then blocked for 1 h in a solution of 2% normal
donkey serum, 1% BSA and 0.3% Triton in PBS at room temperature.
The following antibodies were used: anti-mouse DCT (made by the
T.C. laboratory; 1:500), anti-mouse CD8a (eBioscience, cat: 17-0081-
83, clone: 53-6.7, dilution: 1:300), anti-mouse Ki67 (eBioscience, cat:
14-5698-82, clone: SolA15, dilution: 1:500), anti-mouse CD4 (Biolegend,
cat: 100451, dilution: 1:300) and anti-mouse CD11c (Biolegend, cat:
117316, clone: N418, dilution: 1:300).
Whole-mount and section samples were imaged on a Nikon A1-R
confocal microscope. Images were acquired using a 20 × 0.75 objective
lens for representative pictures and a 10 × 0.5 objective lens for quanti-
tative pictures. Z-stacks were acquired at a resolution of 1,024 × 1,024
or 512 × 512. Microscopy data were analysed using Imaris software with
the 3D visualization module (Bitplane). RBG images were assembled
and were labelled panel information with Adobe Illustrator CS6.
To describe the distribution and density of CD8+ T cells and melano-
cytes, fluorescent signals of T cells and melanocytes were converted
to digital signals using the ‘spot’ function in Imaris software. Position
coordinates of each spot were analysed with the ‘smoothScatter’ pack-
age in R to quantify the cell distribution and density. To quantify the
number of RFP+ fibroblasts, melanocytes and CD8+ T cells in each scale
of tail skin, fluorescent signals of these cells were converted to digital
signals using the ‘spot’ function in Imaris software, then cell number
in each scale was quantified using Python.
Lentivirus infection
High-titre lentivirus production was described before61: calcium phos-
phate transfection of 293FT cells was carried out in a 500-cm2 dish with
pLKO.1 and helper plasmids pMD2.G and psPAX2. Viral supernatant
was collected 46–48 h after transfection, filtered through a 0.45-μm
filter and then concentrated. The final titre was determined by FACS
quantification of virus-infected 3T3 cells.
For knockdown of Cxcl9, Cxcl10, Ifngr1, Jak1 and Stat1, shRNA len-
tivirus constructs were obtained from Sigma RNAi Consortium (TRC)
mouse lentivirus library. shRNA was then subcloned into LV-H2BRFP. To
test knockdown efficiency in vitro, lentivirus was produced by calcium
phosphate transfection of 293FT cells in a six-well plate, followed by
lentivirus infection of 3T3 cells by centrifugation at 1,000g for 30 min
at 37 °C with 10 mg ml−1 polybrene. The knockdown efficiency was
determined by qPCR. Target sequences of individual shRNA used in
experiments are listed here:
Ifngr1-shRNA-1: CCACATAGAATATCAGACTTA;
Ifngr1-shRNA-2: GCCAGAGTTAAAGCTAAGGTT;
Jak1-shRNA-1: GCCCTGAGTTACTTGGAAGAT;
Jak1-shRNA-2: CGGTCCAATCTGCACAGAATA;
Stat1-shRNA-1: GCCGAGAACATACCAGAGAAT;
Stat1-shRNA-2: GCTCACTCAGAACACTCTGAT;
Cxcl9-shRNA-1: CTACACTGAAGAACGGAGAT;
Cxcl9-shRNA-2: TCGTCGTTCAAGGAAGACTA;
Cxcl10-shRNA-1: CATTGTATATGGAAGAACTT;
Cxcl10-shRNA-2: ATGTCTGAATCCGGAATCTA;
For in vivo knockdown of target genes in dermis, 10 μl high-titre
lentivirus (>5 × 108 cfu ml−1) was intradermally injected using insulin
syringes into P1 mice tail skin. Vitiligo induction began at 8–9 weeks
after birth. Samples were collected to perform whole-mount staining
at days 26–33 after vitiligo induction.
Fibroblast transplantation assay
To isolate primary mouse fibroblasts, skin from the indicated region of
newborn female mice was collected, flattened and placed dermis-side
down in 6 ml of 2.4 U ml−1 dispase solution and incubated at 37 °C with
shaking at 80 rpm for 60 min. The epidermis along with most of the hair
follicles were removed from dermis. The dermis was placed in 10 ml
of 2 mg ml−1 collagenase (Sigma-Aldrich, C2674) at 37 °C with shaking
at 80 rpm for 1 h, and afterwards 10 ml of 5% FBS medium was added.
Single-cell suspensions were obtained by repeatedly aspirating and
dispensing with pipette 10 times. The cells were then filtered through
two strainers (70 μm followed by 40 μm).
The isolated primary fibroblasts were cultured in DMEM (Gibco)
medium supplemented with 10% (v/v) FBS (Gibco), 1% (v/v) pen–strep,
and 1% antibiotic–antimyotic. Two to three days after seeding, the pri-
mary newborn mouse fibroblasts were infected with lentivirus containing
H2BRFP. At 48 h after infection, the RFP-labelled fibroblasts were then
injected into the tail skin dermis of 8-week-old adult female wild-type
or Ifngr1 KO mice as indicated. A total of 3 × 106 fibroblasts at a concen-
tration of 105 per μl were intradermally injected into three sites per tail
(10 μl per site, 1 cm between sites). Note that the injection procedure can
only deliver the fibroblasts into lower dermis as indicated by our section
staining analysis. Mice were allowed to recover for 3–7 days before vitiligo
induction. Whole-mount analysis of CD8+ T cells was performed at day
33 after vitiligo induction. The density of T cells was analysed by Imaris.
The enrichment score of CD8+ T cells used in Fig. 4h was calculated by
T cell density in the RFP+ region dividing T cell density in RFP− region.
FACS analysis
For FACS quantification and collection of epidermal immune cells
including CD8+ T cells, CD4+ T cells, Langerhans cells and γδ T cells,
skin from the middle section of the tail was collected, flattened, cut into
1 × 1-cm pieces and then floated dermis-side down in 2 ml of 2.4 U ml−1
dispase solution and incubated at 37 °C with shaking at 80 rpm for
50 min. To obtain most of the epithelial cells, the epidermis was then
carefully removed from dermis along the anterior–posterior direc-
tion with a fine-tipped tweezer. The removed epidermis was floated
dermis-side down in 3 ml TrypLE Express Enzyme (Gibco) at 37 °C for
10 min, and afterwards 5% FBS was added to neutralize TrypLE.
For FACS quantification of epidermal melanocyte, melanocytes
located in the hair follicle must be removed. We found out, before skin
collection, that waxing using VEET wax paper facilitates subsequent hair
follicle separation from epidermis. Waxed tail skin was flattened and
floated dermis-side down in 2 ml of 2.4 U ml−1 dispase solution at 37 °C
with shaking at 80 rpm for 50 min. To efficiently obtain hair-follicle-free
epidermis, the epidermis was swiftly peeled from the dermis along the
posterior–anterior direction with a fine-tipped tweezer. The occasional
remaining hair follicles attached to epidermis were removed using
tweezers. The hair-follicle-free epidermis was then floated dermis-side
down in 3 ml Trypsin-EDTA (0.25%) at 37 °C for 10 min, and afterwards
5% FBS was added to neutralize trypsin.
Single-cell suspensions were obtained by repeatedly aspirating and
dispensing with a pipette 20 times followed by filtering through 40-μm
strainers. The cells were then stained with anti-mouse CD117 (eBiosci-
ence, cat: 13-1171-82, clone: 2B8, dilution: 1:200) for melanocyte detec-
tion, anti-mouse CD45 (BD, cat: 559864, clone: 10-F11, dilution: 1:300)
for immune cell detection; anti-mouse CD3 (Biolegend, cat: 100220,
clone: 17A2, dilution: 1:300) and anti-mouse CD8a (eBioscience, cat:
17-0081-83, clone: 53-6.7, dilution: 1:300) for CD8+ T cell detection;
anti-mouse CD3 and anti-mouse CD4 (Biolegend, cat: 100451, dilution:
1:300) for CD4+ T cell detection; anti-mouse MHCII (Biolegend, cat:
107629, dilution: 1:300) and anti-mouse CD207 (Biolegend, cat: 144206,
dilution: 1:300) for Langerhans cell detection; and anti-mouse CD3 and
anti-mouse TCRγδ (Biolegend, cat: 118108, dilution: 1:300) antibodies
for γδ T cell detection.
For FACS quantification and collection of dermal fibroblasts,
endothelial cells and immune cells including macrophage and dendritic
cells from the tail skin, skin from the middle section of the tail was col-
lected, flattened, cut into 1 × 1-cm pieces, and then floated dermis-side
down in 2 ml of 2.4 U ml−1 dispase solution and incubated at 37 °C with
shaking at 80 rpm for 50 min. To get rid of epidermis and hair follicles,
the epidermis was then carefully removed from dermis along the ante-
rior–posterior direction with a fine-tipped tweezer. The dermis was
then placed in 10 ml of 2 mg ml−1 collagenase (Sigma-Aldrich, C2674)
and incubated at 37 °C with shaking at 80 rpm for 1 h; afterwards 5%
FBS solution was used for neutralization. Single-cell suspensions were
obtained by repeatedly aspirating and dispensing with a pipette for 20
times followed by filtering through 70-μm and 40-μm strainers. The
cells were stained with anti-mouse CD31 (eBioscience, cat: 25-0311-81,
Article
clone: 390, dilution: 1:300), anti-mouse-CD45, and anti-mouse PDGFRA
(eBioscience, cat: 17-1401-81, clone: APA5, dilution: 1:300) antibodies
for endothelial cell, total immune cell, and fibroblast detection, respec-
tively; anti-mouse/human CD11b (Biolegend, cat: 101215, dilution:
1:300) and anti-mouse F4/80 (Biolegend, cat: 123116, dilution: 1:300) for
macrophage detection; and anti-mouse MHC-II and anti-mouse CD11c
(Biolegend, cat: 117310, dilution: 1:300) for dendritic cell detection.
For analysis and isolation of immune cells from spleens, the spleen
was mechanically dissociated and the red blood cells were lysed with
RBC lysis buffer according to the manufacturer’s instructions (Beyo-
time). Single-cell suspensions were obtained by filtering with strainers
(70 μm and 40 μm) and then cells were stained with anti-mouse CD3,
anti-mouse CD45 and anti-mouse CD8 antibodies. Anti-mouse VB13
(Biolegend, cat: 140703, clone: MR12-4, dilution:1:300) antibody was
used in Pmel T cell adoptive transfer experiments.
For CXCR3 expression analysis in skin immune cell types of viti-
ligo mice, the entire tail skin was collected, flattened, then floated
dermis-side down in 5 ml trypsin-EDTA (0.25%) at 37 °C for 30 min.
The skin was inverted and scraped with a scalpel and the cell slurry was
filtered with strainers (70 μm and 40 μm), then cells were co-stained
with anti-mouse CXCR3 (Biolegend, cat: 126506, dilution: 1:300) against
the immune cell markers mentioned above for further analysis.
For FACS analysis of CD8+ T cells and melanocytes from mouse dorsal,
ventral, paw dorsal and paw ventral skin, skin biopsies were collected,
flattened and floated dermis-side down in 3 ml of TrypLE and incubated
at 37 °C with shaking at 80 rpm for 30 min. The skin epidermis was gen-
tly scraped with a scalpel and the cell slurry was filtered with strainers
(40 μm), then cells were stained with anti-mouse CD45, CD3, CD8 and
CD117 antibodies for further analysis.
For re-isolation of injected RFP-labelled dorsal or paw fibroblasts
from Ifngr1 KO mouse tail, tail skin was separated into epidermis and
dermis after dispase treatment. After digestion of dermis with colla-
genase and neutralization, single-cell suspensions were obtained by
filtering through 70-μm and 40-μm strainers. Then RFP+ fibroblasts
were isolated for further qPCR analysis.
Antibodies described above were incubated on ice for 15–30 min.
DAPI was used to exclude dead cells. Cell analysis and isolation were
performed on a BD FACSAria sorter equipped with FACSDiva software
(BD Bioscience). Data were analysed with FlowJo software (FlowJo LLC).
IFNγ treatment of fibroblasts in vitro
For isolation of human dermal fibroblasts, skin epidermis and der-
mis of biopsy were mechanically separated after 2.4 U ml−1 dispase
treatment for 1 h at 37 °C with shaking at 80 rpm. Further digestion of
the dermis was performed with 10 mg ml−1 collagenase at 37 °C with
shaking at 80 rpm for 1 h. The fibroblasts were then expanded and
cultured in DMEM supplemented with 10% FBS, 1% pen–strep and 1%
antibiotic–antimycotic. Passage-3 or -4 fibroblasts were treated with
1 U ml−1 recombinant human IFNγ for 3 h followed by RNA-seq. For
each region, the average FPKM value of two individual samples was
used to estimate the gene expression level. Genes were included in
the heat map if they were differentially expressed in at least one of the
eight anatomic regions after IFNγ treatment (log2-transformed fold
change > 1 and P < 0.01).
For isolation of mouse fibroblasts, skin from different regions was
taken from C57 newborn mice. After enzymatic digestion, the single-cell
suspension was stained with anti-mouse PDGFRA. PDGFRA+ fibroblasts
were isolated by FACS and cultured in DMEM supplemented with 10%
FBS, 1% pen–strep and 1% antibiotic–antimycotic. At 72 h after seeding,
primary mouse fibroblasts were treated with 5 U ml−1 recombinant
mouse IFNγ for 3 h and collected for RNA isolation and qPCR.
For 3T3 cells, shRNA-containing lentivirus-infected cells were seeded
in a 6-well plate 24 h before IFNγ treatment. After treatment with 5 U ml−1
recombinant mouse IFNγ for 3 h, cells were collected for RNA isolation
and qPCR.
In vitro transwell T cell migration assay
Transwell migration of lymphocytes was performed with mature cyto-
toxic T lymphocytes (CTLs) and concentrated fibroblast-conditioned
medium. In brief, splenocytes isolated from OT-1 mice were stimulated
with OVA257-264 for 2 days in the presence of 10 ng ml−1 IL-2. Cells were
centrifuged and cultured in fresh medium containing 10 ng ml−1 IL-2 for
two more days, after which most of the cells in the culture were CTLs.
To acquire fibroblast-conditioned medium, primary fibroblasts with
the indicated genotypes or regions were cultured with DMEM with or
without 1,000 U ml−1 IFNγ for 6 h at 37 °C. The medium was then con-
centrated (1×, 2×, 5×, 10× or 25×) for chemo-attractants. In the transwell
cell migration assay, 96-well 5-μm inserts from Corning were used.
Migration medium (200 μl) containing either fibroblast-conditioned
medium with the indicated concentrations or control medium were
loaded into the lower chamber. CTLs (1.5 × 105) suspended in 50 μl
medium were then placed in the upper chamber. At 3 h after incubation
at 37 °C under 5% CO2 conditions, the migrated and unmigrated T cells
were collected from the lower and upper chambers, respectively. To
normalize samples, beads of known concentration were added to all
samples before counting by FACS. The migration ratio was determined
by dividing the number of migrated cells by the number of migrated
plus unmigrated cells.
To test the ability of secreted chemokines to promote CD8+ T cell
migration, CXCL2, CXCL3, B2M, CXCL10 and CXCL9 at concentrations
of 0, 0.2, 0.4, 1 and 2 μg ml−1 were used.
RNA isolation and qPCR
Fibroblast total RNA was isolated from FACS-sorted cells with TRIzol
(Thermo Fisher Scientific) followed by extraction using the Direct-Zol
RNA Miniprep Kit (Zymo Research). CD8+ T cell total RNA was isolated
from FACS-sorted cells using the Picopure RNA Isolation Kit (Thermo
Fisher Scientific) and RiboAmp HS PLUS RNA Amplification Kit (Thermo
Fisher Scientific). For cDNA synthesis, equal amounts of RNA were
reverse-transcribed using Oligo-dT (Vazyme, R222-01). Expression
levels were normalized to the expression of PPIB or GAPDH. qPCR was
conducted using a CFX96TM Real-Time system (Bio-Rad) with Power
SYBRR Green PCR Master Mix (Life Technologies). All primer pairs
were designed for the same cycling conditions: 10 min at 95 °C for ini-
tial denaturation, 40 cycles of 10 s at 95 °C for denaturation, 30 s at
62 °C for annealing, and 10 s at 65 °C for extension. The primers were
designed to produce a product spanning an exon–intron boundary in
each of the target genes.
Single-cell RNA library construction and sequencing
Single-cell cDNA libraries were prepared using the Chromium Single
Cell 3′ Library and Gel Bead Kit v2 according to the manufacturer’s
instructions. In brief, cell suspensions in a chip were loaded on a Chro-
mium Controller (10x Genomics) to generate single-cell GEMs (gel
beads in emulsion). scRNA-seq libraries were then prepared using the
Chromium Single Cell 3′Gel Bead and Library Kit (P/N 120236, 120237,
120262; 10x Genomics). Qualitative analysis of each DNA library
was performed using an Agilent 2100 Bioanalyzer. The concentra-
tion of each DNA library was measured using a Qubit fluorometer
(Invitrogen). Libraries were sequenced on an Illumina NextSeq 500
(2 × 150 paired-end reads).
Processing of single-cell sequencing data
The raw sequence reads were aligned and quantified using Cell Ranger
(v.1.3.1) software, which was obtained from 10x Genomics (https://
support.10xgenomics.com/single-cell-gene-expression/software/pipe
lines/latest/installation). The human hg38 assembly reference was
used for analysis. The raw count matrix data were imported into
R (3.5.1) and R studio (1.1.456) using the Seurat (v.2.3.2) package for
further data analysis. For each of the 15 samples, we initially set up a
first filter of min.cells = 3 and min.genes = 200 per sample. Cells with
more than 5,000 genes or more than 25,000 unique molecular identi-
fiers (UMIs) were removed. We kept cells with less than 1% mitochon-
drial gene expression. The raw counts were normalized by a factor of
10,000 and log-transformed to obtain log(T+ 1) values. Variable genes
were identified by fitting the mean-variance relationship and met the
following criteria: 0.0125 < mean of non-zero values < 3 AND standard
deviation > 0.4.
Cell-type classification using t-SNE
Unsupervised clustering of cells was performed with Seurat. Dimen-
sionality reduction was performed using principal component analysis
(PCA). The first 20 principal components were selected according to
the PCA elbow plot and used for clustering with resolution parameter
0.1. Cell clusters were visualized using t-SNE plots, with all significant
principal components as input. We integrated data from all samples
using canonical correlation analysis. The shared nearest neighbour
graph was constructed on a cell-to-cell distance matrix from the top 30
aligned canonical correlation vectors. The shared nearest neighbour
graph with different resolution was used as input for the smart local
moving algorithm to obtain cell clusters, and visualized with t-SNE
or uniform manifold approximation and projection (UMAP). Cluster
marker genes were identified using the FindAllMarkers function by
Wilcoxon rank-sum test with the following parameters:min. pct = 0.25,
thresh.use = 1. On the basis of previous knowledge and consistency
within the different resolutions, we selected the final number of clus-
ters that included all the major cell types in the skin, resulting in eight
different clusters.
Profiling differentially expressed genes within each cell cluster
To identify differentially expressed genes in two melanocyte subtypes,
we applied the FindMarkers function from Seurat to the normalized
gene expression data, with the following parameters: min.pct > 0.25
and thresh.use = 0.25. The highly expressed genes in the M1 melanocyte
cluster were identified as C1 signature genes (fold change > 2, P < 0.01).
GO analysis of M1 signature genes was performed using the GO web
service (http://geneontology.org). For comparing gene expression
between four T cell clusters, the differentially expressed genes were
selected using the threshold fold change > 0.25, P < 0.01. The gene
expression levels shown in the various charts in the manuscript were
plotted using packages in R.
RNA sequencing, alignment, analysis and visualization
RNA from FACS-purified cells was submitted to Novogene for quanti-
fication and RNA-seq library preparation and sequencing. The library
was sequenced on an Illumina HiSeq platform using the paired-end
150-bp sequencing strategy.
For RNA-seq analysis of mouse tail fibroblasts, raw RNA-seq data
were mapped to the mouse genome (GRCm38/mm10) using TopHat
(v.2.0.13) with default settings to produce a reference-guided transcript
assembly. Cufflinks (v.2.2.1) was used to normalize expression levels
for each sample to obtain FPKM.
Cuffdiff was used to quantify changes in gene expression between
the control wild-type (WT) fibroblasts, vitiligo WT fibroblasts, and
vitiligo Ifngr1 KO fibroblasts. Genes with significantly upregulated
expression levels (P < 0.01, fold change of vitiligo WT/naive WT > 1.5,
vitiligo WT/vitiligo KO > 1.3) were chosen for further analysis. GO analy-
sis of upregulated genes was performed using a GO web service (http://
geneontology.org). Differentially expressed gene data were presented
using the ‘ggplot2’ package in R software.
For in vitro IFNγ-treated human fibroblast RNA-seq analysis, raw
transcriptome sequence data were mapped to the human genome
(hg38) using STAR (v.2.6.1a) with default settings to produce a
reference-guided transcript assembly. FeatureCount (v.2.2.1) was used
to count reads of genes and normalize expression levels for each sample
to obtain FPKM. Differential expression analysis was performed using
the DESeq2 package. Genes with significantly upregulated expres-
sion (P < 0.01, fold change > 2) were chosen for further analysis. Heat
maps showing differentially expressed genes were generated using
the ‘pheatmap’ and ‘ggrepel’ packages in R.
Mathematical model
We developed a mathematical model to predict T cell cluster expan-
sion and vitiligo progression. The skin tissue was modelled as a square
lattice with two layers. The upper epidermal layer contains T cells and
melanocytes. The lower dermal layer contains fibroblasts with different
chemotactic effects. The state of each cell changes over time according
to the state of itself, its neighbour cells and the surrounding signals.
On the upper layer, each lattice site may contain one melanocyte or
one T cell, or it may be empty. At first the melanocytes are uniformly dis-
tributed, with an average spacing of four lattice sites. The melanocyte
does not move, proliferate or die on its own. Only if the surrounding
T cells exceed a threshold number will the melanocyte be killed and the
lattice site it occupied become free. The T cell occupation at each lattice
site is controlled by an independent birth–death stochastic process.
The death rate is constant. The birth rate depends on the concentration
of the T cell-recruiting chemokine produced by fibroblasts on the lower
layer. In addition, the total birth rate of T cells in the system is limited.
On the lower layer, every lattice site contains one fibroblast. The
fibroblast can respond to its nearby activated T cells in the upper
layer and release chemokine. If there is no activated T cell nearby, the
fibroblasts stay in a quiescent state in which they do not produce any
chemokine. By default, all the T cells are inactivated. Only when a mel-
anocyte is killed will the nearby T cells be activated, which in turn will
induce the fibroblast to release chemokine. The production rate of
chemokine per fibroblast is λ. The chemokine diffuses on the lattice
with a rate-constant D and degrades at rate γ. The model details are
stated in Algorithm 1.
At first all fibroblasts are inactive and there is no chemokine in the
system. Progressive melanocyte death is a consequence of two events.
The first event is the randomly local accumulation of T cells near one
melanocyte. If the number of T cells near a melanocyte exceeds the
threshold, the melanocyte will be killed. This melanocyte-killing event
will activate the nearby T cells, which will in turn activate the nearby
fibroblasts beneath them in the lower layer. These fibroblasts will
release chemokine to recruit more T cells to this location. As a result,
the local density of T cells will increase. If the local density of T cells near
a melanocyte exceeds the threshold, subsequent melanocyte-killing
events are to be expected. This process constitutes a positive feedback
mechanism that leads to the extermination of melanocytes in a spa-
tially progressive manner. The melanocyte-free region continues to
expand as more T cells concentrate at the border of the T cell cluster.
However, in the centre of the T cell cluster, after the activated T cells
undergo apoptosis, the fibroblast will become quiescent and hence
the number of T cells will drop.
Algorithm 1 (detailed model implementation)
Initialization.
1. Set chemokine concentration φ(x, t) = 0.
2. Generate melanocytes:
(a).  Seed melanocytes uniformly on the lattice with an interval of four
lattice sites.
(b).  Randomly perturb the location of each melanocyte.
Repeat until t > Tend.
1. Compute T cell arriving rate a(x, t):
(a).  a(x, t) = α(φtiny + φ(x, t)).
(b).  If ∑ x a(x, t) > Amax , then a(x, t) = Amax a(x, t)/ ∑ x a(x, t).
2. For all lattice sites x on the upper layer:
(a).  If x is empty, add a T cell (inactive state) with probability a(x, t)dt.
(b).  If x is occupied by a T cell, remove it with probability βdt.
Article
(c).  If x is occupied by a melanocyte, compute n(x), the number of T cells
within a local patch of size Pc × Pc centred at x. If n(x) > nc, remove this
melanocyte and activate all the n(x) T cells.
3. For all lattice sites x on the lower layer:
(a).  If there exists an activated T cell within a local patch of size 3 × 3
centred at x in the upper layer, activate the fibroblast at x. Otherwise
set the fibroblast to be inactive.
(b).  Update φ(x, t) according to φ(x, t) = φ(x, t) + ∆φ(x, t)dt + λI (x)
dt − γφ(x, t).
Here ∆ is the discretized Laplacian operator on the lattice.
 λ , if the fibroblast at x is active
λ I (x) = 
0, otherwise
4. Set t = t + dt.
Variables. φ(x, t): chemokine concentration at lattice site x at time t.
a(x, t): T cell arriving rate at lattice site x at time t.
Parameters. System settings:
M × N: system size. The size of the lattice is M × N. M × N = 100 × 100
in Extended Data Fig. 10a, b. M × N = 512 × 512 in Extended Data Fig. 10c.
M × N = 400 × 1,024 in Fig. 4j and Extended Data Fig. 10d. The unit of
length (length of one lattice site) is 20 μm.
Tend: total simulation time. Tend = 2,000. The unit of time is 1 h.
dt: time step. A discrete time step dt = 0.005 (hour) is used through-
out our simulation.
T cell propensity:
1. α: birth rate of T cells per lattice site. α controls the base rate of T cell
arrival on the lattice. In our simulation we chose α = 1.2 for the mouse
experiment (elevated T cell amount), and α = 1.1 for the symmetry
experiment (normal T cell amount).
2. β: death rate of T cells per lattice site. The rate of removal of T cells.
For inactive T cells the removal rate corresponds to how fast a T cell
leaves the tissue and possibly re-enters the circulation system. For
active T cells, the removal rate corresponds to its death rate. In prin-
ciple the removal rate for the two different cases may be different.
In our simulation we make them equal at β = 0.04 throughout our
simulation.
3. Amax: total supply of T cells. Controls the total amount of T cells avail-
able to the system. We chose Amax = 0.015 × α × N × N throughout our
simulation.
Remark: α, β and Amax are difficult to measure experimentally. We
chose their values to fit the experimental data. Changes in these param-
eters mainly affect the speed of progression of the melanocyte-killing
region, but have little effect on the characteristic behaviour of the
model, such as the shape of the melanocyte-killing region and the way
it expands.
Chemokine production and distribution:
1. D: diffusion constant of the chemokine. D = 2.
2. λ: production rate of chemokine per active fibroblast. λ = 0.04.
3. γ: decay rate of chemokine. γ = 0.2.
4. φtiny: background chemokine level. φtiny = 0.01. A numerical trick to
avoid division by zero (see a(x, t) in Algorithm 1).
Remark: Together, D, λ, and γ control chemokine dynamics. γ = 0.2
gives a half-life period of 3.5 h for the chemokine. The characteristic
spreading length (the distance over which the steady-state signal from
a point source falls by a factor of 1 − 1/e, or about 63%)62 of the chemokine
is D /γ  = 3.16, or 63 μm (1 unit = 20 μm), which is a reasonable value
for morphogens, growth factors, cytokines and chemokines (less than
100 μm)63. The production rate λ was chosen to be large enough to
activate the melanocyte-killing event.
Melanocyte properties:
1. Pc: local patch size. Pc = 5.
2. nc: critical number of T cells for the killing of a melanocyte. nc = 15. If
the number of T cells in the local Pc × Pc patch centred at a melanocyte
is greater than nc, the melanocyte will be killed. When this happens,
its nearby T cells will be activated.
Remark: The choice of Pc = 5 is based on the estimation that the size
of a melanocyte is approximately 100 μm. The choice of nc = 15 is to fit
the experimental data. A larger nc makes the initial melanocyte-killing
event less likely to occur (a longer waiting time and fewer incidents).
Statistics and reproducibility
Statistical analyses were performed with GraphPad Prism (v.8.00),
R (v.3.5.1), and RStudio (v.1.1.456). For the representative images and
FACS profiles presented in the paper, each experiment was repeated
at least three times independently with similar results. No statistical
methods were used to predetermine sample size.
Figure 1: e, n = 5 individuals for each disease/healthy state. Median,
box: 25/75 percentile, whisker: 5/95 percentile. f, n = 5 individuals for each
disease/healthy state. Two-tailed unpaired t-test. **P < 0.01, NS, not sig-
nificant. Data are mean ± s.d. h, n ≥ 3 for each disease/healthy state. Data
are mean ± s.d. Figure 2: b, d, n ≥ 4 independent mice for each condition.
Two-tailed unpaired t-test. ***P < 0.001, ****P < 0.0001. e, n = 9 independ-
ent host mice for each condition. Two-tailed paired t-test. ***P < 0.001,
NS, not significant. f, n ≥ 3 independent mice. All data are mean ± s.d.
Figure 3: b, n ≥ 150 scales from 3 independent mice for each condition.
Correlation test measured by Pearson coefficient r and two-tailed P-value
analysis. c, n = 3 independent mice for each condition. Data are mean ± s.d.
d, n = 3 independent experiments with at least 2 replicates per experiment.
Data are mean ± s.d. f, n = 20 samples, R, and P from a linear regression
analysis. g, n ≥ 5 samples. Data are mean ± s.d. h, n ≥ 15 scales from at least
3 independent mice for each condition. Median, box: 25/75 percentile,
whisker: 5/95 percentile, dot: outlier. Two-tailed unpaired t-test. *P < 0.05,
***P < 0.001, ****P < 0.0001, NS, not significant. Figure 4: c, n = 3 individuals
with technical triplicates per condition. Data are mean ± s.d. d, Correlation
test measured by Spearman coefficient. e, Top, n ≥ 50 hair follicles from
at least 3 mice. Median, box: 25/75 percentile, whisker: 5/95 percentile,
dot: outlier. Bottom, n = 6 mice for each condition. Two-tailed unpaired
t-test. **P < 0.01, ***P < 0.001, ****P < 0.0001, NS, not significant. Data are
mean ± s.d. f, n = 3 independent mice with technical triplicates per condi-
tion. Data are mean ± s.d. g, n = 3 independent experiments with at least
2 replicates per experiment. Two-tailed unpaired t-test. *P < 0.05. Data
are mean ± s.d. h, n = 16 samples from at least 10 independent mice per
condition. Two-tailed unpaired t-test. ****P < 0.0001. Data are mean ± s.d.
i, n ≥ 2 independent mice with technical triplicates per condition. Data
are mean ± s.d. k, Each dot represent median. Correlation test measured
by Spearman coefficient. Extended Data Fig. 1: m, n ≥ 3 for each disease/
healthy state. Data are mean ± s.d. r, n = 6 samples from at least 3 patients
with vitiligo. Left and middle, two-tailed unpaired t-test. ***P < 0.001. Data
are mean ± s.d. Right, R, and P from a linear regression analysis. Extended
Data Fig. 2: e, n = 10 independent mice for each condition. Two-tailed
unpaired t-test. ****P < 0.0001. f, n = 6 independent mice for each con-
dition. Two-tailed unpaired t-test. ****P < 0.0001. g, n ≥ 6 independent
mice for each condition. Two-tailed unpaired t-test. ****P < 0.0001. h, i,
n = 3 independent mice for each condition. Two-tailed unpaired t-test.
k, Left, n = 5 independent mice for each condition. Right, n = 2 inde-
pendent mice for each condition. All data are mean ± s.d. Extended
Data Fig. 3: f, n = 5 individuals for each disease/healthy state. Two-tailed
unpaired t-test. *P < 0.05, NS, not significant. h, i, n ≥ 4 for each condi-
tion. Two-tailed unpaired t-test. *P < 0.05, **P < 0.01, NS, not significant.
All data are mean ± s.d. Extended Data Fig. 4: d, n = 6 independent mice
for each condition. Two-tailed unpaired t-test. **P < 0.01, ****P < 0.0001.
Data are mean ± s.d. f, o, t, w, n ≥ 4 independent mice for each condition.
Two-tailed unpaired t-test. **P < 0.01, ***P < 0.001, NS, not significant. Data
are mean ± s.d. h, p, x, n ≥ 100 scales from at least 3 independent mice for
each condition. Correlation test measured by Pearson coefficient r and
two-tailed P-value analysis. j, n = 2 independent mice with technical trip-
licates for each condition. Two-tailed unpaired t-test. ***P < 0.001. Data
are mean ± s.d. r, n ≥ 3 independent mice for each condition. Two-tailed
unpaired t-test. ***P < 0.001. Data are mean ± s.d. s, n ≥ 8 independent mice
for each condition. Data are mean ± s.d. Extended Data Fig. 5: a, b, c, n ≥ 4
independent mice for each condition. Two-tailed unpaired t-test. *P < 0.05,
**P < 0.01, NS, not significant. Data are mean ± s.d. f, n ≥ 2 independent
mice for each condition. Data are mean ± s.d. g, n ≥ 6 independent host
mice for each condition. Two-tailed paired t-test. NS, not significant.
Extended Data Fig. 6: a, n = 2 independent mice with technical triplicates
for each condition. Data are mean ± s.d. c, n ≥ 7 independent mice for each
condition. Two-tailed unpaired t-test. ***P < 0.001, ****P < 0.0001, NS, not
significant. Data are mean ± s.d. e, n ≥ 9 independent mice for each condi-
tion. Two-tailed unpaired t-test. *P < 0.05, **P < 0.01, NS, not significant.
Data are mean ± s.d. Extended Data Fig. 7: b, f, n ≥ 110 scales from 3 inde-
pendent mice for each condition. Correlation test measured by Pearson
coefficient r and two-tailed p-value analysis. c, d, g, n ≥ 4 independent mice
for each condition. Two-tailed unpaired t-test. *P < 0.05, **P < 0.01, NS, not
significant. Data are mean ± s.d. i, k, n = 3 independent experiments with
at least 2 replicates per experiment. Data are mean ± s.d. Extended Data
Fig. 8: a, n = 3 independent mice for each condition. Data are mean ± s.d. b,
n = 3 biological replicates for each condition. Data are mean ± s.d. c, n ≥ 3
independent mice with technical triplicates for each condition. Data are
mean ± s.e.m. d, n ≥ 15 scales from at least 3 independent mice for each
condition. Median, box: 25/75 percentile, whisker: 5/95 percentile, dot:
outlier. Two-tailed unpaired t-test. *P < 0.05, **P < 0.01, ***P < 0.001, NS,
not significant. e, n ≥ 4 independent mice for each condition. Two-tailed
unpaired t-test. **P < 0.01, NS, not significant. Data are mean ± s.d. f, n = 5
individuals for each disease/healthy state. Median, box: 25/75 percentile,
whisker: 5/95 percentile. g, n ≥ 3 independent mice for each condition.
Data are mean ± s.d. Extended Data Fig. 9: c, n = 3 individuals with tech-
nical triplicates for each condition. Data are mean ± s.e.m. d, n = 2 indi-
viduals for each condition. Data are mean ± s.d. e, n = 2 individuals with
technical triplicates for each condition. Data are mean ± s.d. f, Left, n ≥ 5
mice for each condition. Right, n ≥ 3 mice. Two-tailed unpaired t-test.
NS, not significant. Data are mean ± s.d. g, n ≥ 50 hair follicles from at
least 3 mice. Median, box: 25/75 percentile, whisker: 5/95 percentile, dot:
outlier. Two-tailed unpaired t-test. ****P < 0.0001, NS, not significant. h,
n ≥ 5 mice for each condition. Two-tailed unpaired t-test. ***P < 0.001,
****P < 0.0001, NS, not significant. Data are mean ± s.d. i, n = 3 independent
mice with technical triplicates for each condition. Data are mean ± s.d. j,
n = 3 independent experiments with at least 2 replicates per experiment.
Two-tailed unpaired t-test. Data are mean ± s.d. k, n ≥ 7 scales for each
condition. Two-tailed unpaired t-test. **P < 0.01, ***P < 0.001. Data are
mean ± s.d. Extended Data Fig. 10: e, f, n = 3 biological replicates for each
condition. Data are mean ± s.d. h, n ≥ 100 scales from at least 3 independ-
ent mice for each condition. Median, box: 25/75 percentile, whisker: 5/95
percentile, dot: outlier. Correlation test measured by Spearman coef-
ficient. i, Each dot represents the median. Correlation test measured by
Spearman coefficient. 
Data reporting. Participants, mice and samples were assigned into
different experimental groups randomly. In the experiments with
specific genotypes, samples were sorted into groups based on their
genotypes. Mice were littermates when available or age-matched
and co-housed to reduce additional variables beyond their genotype
difference. Investigators were blinded to the genotype of mice during
vitiligo induction and data collection.
Reporting summary
Further information on research design is available in the Nature
Research Reporting Summary linked to this paper.
Data availability
The scRNA-seq data, mouse fibroblast RNA-seq data and human fibro-
blast RNA-seq data have been deposited in the Genome Sequence
Archive (GSA) with accession number PRJCA006797. The code for the
vitiligo progression model is available at https://github.com/hydrays/
vitiligo. Source data are provided with this paper.
51. Tang, Y., Harrington, A., Yang, X., Friesel, R. E. & Liaw, L. The contribution of the Tie2+
lineage to primitive and definitive hematopoietic cells. Genesis 48, 563–567 (2010).
52. Deng, L. et al. A novel mouse model of inflammatory bowel disease links mammalian
target of rapamycin-dependent hyperproliferation of colonic epithelium to
inflammation-associated tumorigenesis. Am. J. Pathol. 176, 952–967 (2010).
53. McCubbrey, A. L., Allison, K. C., Lee-Sherick, A. B., Jakubzick, C. V. & Janssen, W. J.
Promoter specificity and efficacy in conditional and inducible transgenic targeting of
lung macrophages. Front. Immunol. 8, 1618 (2017).
54. Le Gal, F. A. et al. Direct evidence to support the role of antigen-specific CD8+ T cells in
melanoma-associated vitiligo. J. Invest. Dermatol. 117, 1464–1470 (2001).
55. Lo, J. A. et al. Epitope spreading toward wild-type melanocyte-lineage antigens rescues
suboptimal immune checkpoint blockade responses. Sci. Transl. Med. 13, eabd8636
(2021).
56. Hua, C. et al. Association of vitiligo with tumor response in patients with metastatic
melanoma treated with pembrolizumab. JAMA Dermatol. 152, 45–51 (2016).
57. Herzberg, B. & Fisher, D. E. Metastatic melanoma and immunotherapy. Clin. Immunol. 172,
105–110 (2016).
58. Nardin, C. et al. Vitiligo under anti-PD1 therapy is associated with increased survival in
melanoma patients. J. Am. Acad. Dermatol. 82, 770–772 (2020).
59. Halprin, K. M. Epidermal "turnover time"–a re-examination. Br. J. Dermatol. 86, 14–19
(1972).
60. Gellatly, K. J. et al. scRNA-seq of human vitiligo reveals complex networks of subclinical
immune activation and a role for CCR5 in Treg function. Sci. Transl. Med. 13, eabd8995
(2021).
61. Chen, T. et al. An RNA interference screen uncovers a new molecule in stem cell
self-renewal and long-term regeneration. Nature 485, 104–108 (2012).
62. Lander, A. D. Morpheus unbound: reimagining the morphogen gradient. Cell 128,
245–256 (2007).
63. Chen, C. C. et al. Organ-level quorum sensing directs regeneration in hair stem cell
populations. Cell 161, 277–290 (2015).
Acknowledgements This work was supported by grants from the National Key R&D Program of
China (2017YFA0103500), the National Basic Research Program of China 973 Programs
(2014CB849602), National Natural Science Foundation of China (11671415) and the Beijing
Municipal Natural Science Foundation (7172192). We thank the NIBS Animal Facility for their
expert handling and care of mice; the NIBS Biological Resource Centre for FACS; the NIBS
imaging facility for assistance with the confocal microscope experiment; the NIBS sequencing
centre for sample processing of 10x genomics RNA-seq; V. Hearing for providing the DCT and
TYR peptide sequences for in-house-made antibodies; F. Shao for help with importing Ifngr1
KO mice; and all members of the laboratory of T.C. for discussions and technical support.
Author contributions T.C. and Z.X. conceived the project, designed the experiments and wrote
the manuscript. Z.X. and D.C. performed most of the in vivo experiments. Y.D. participated in
the lentivirus injection experiment. K.J. performed some of the immunofluorescent staining
experiment. Z.X., H.H. and J.W. conducted bioinformatics analysis of single-cell sequencing
data. W. Wu collected fibroblasts from different body positions and participated in the cell
culture experiment. D.C., H.H., J.W. and Z.X. conducted the bioinformatics analysis of the
RNA-seq data. Y.H. developed the mathematical model. L.Z., W. Wang, Y.Y. and J.C. collected
samples from patients with vitiligo and healthy donors. S.L., C.L. and J.C. collected clinical
data from patients with vitiligo and carried out statistical analysis. J.S. provided the CD8
depletion antibody in the CD8+ T cell depletion experiment.
Competing interests All authors declare no competing interests related to this work.
Additional information
Supplementary information The online version contains supplementary material available at
https://doi.org/10.1038/s41586-021-04221-8.
Correspondence and requests for materials should be addressed to Jianmin Chang or Ting
Chen.
Reprints and permissions information is available at http://www.nature.com/reprints.
Article

Extended Data Fig. 1 | See next page for caption.

Extended Data Fig. 1 | Comprehensive analysis of different cell populations
from patients with vitiligo and healthy donors. a, Representative images of
the skin of patients with vitiligo showing characteristic bilateral symmetric
lesion patterns in different body regions. b, Experimental workflow and
representative FACS profiles to obtain single cells from human skin biopsies
for single cell RNA-seq. c, Clinical information for all donors with the number
of cells sequenced in single-cell analysis. d, t-SNE visualization of all collected
cells showing eight main cell types with distinct clusters. e, Heat map analysis
of differentially expressed genes in each cell type. f–h, violin plot (f), feature
plot (g), and dot plot (h) analysis of signature genes for each cell type.
i, Volcano plot and GO analysis of genes enriched in melanocytes from the M1
cluster compared to M2 cluster. j, T-SNE projection of T cells from patients with
vitiligo and healthy donors. k, Volcano plot showing genes differentially
expressed (>2-fold, t test p < 0.01) in each T cell sub-cluster. l, m, Representative
immunofluorescent images (l) and quantification (m) of T cell subtype markers
GZMA, ZNF683, CD40LG, and FOXP3 in T cells (CD3+) in the skin of patients
with vitiligo. n, GO categories of genes enriched in each cell type from the skin
of patients in the progressive state compared to healthy donor skin. o, Heat
map analysis of 20 IFNγ response genes in eight cell types from the skin of
patients in the progressive state and healthy donor skin. p, Violin plot showing
the expression of IFNG in T cell subtypes. q, Violin plots showing IFNGR1 and
IFNGR2 expression in different skin cell types. r, Representative
immunofluorescent staining images, quantification (left and middle), and
regression analysis (right) between density of CD8+ T cell and pSTAT1+ cell in
the skin of patients with vitiligo. s, Representative immunofluorescent staining
images of pSTAT1 signal in each cell type in the skin of patients with vitiligo.
Scale bars, 50 µm (l, r, s). For exact p values, see Source Data. For statistics,
p summary and sample sizes, see Methods.
Article

Extended Data Fig. 2 | See next page for caption.

Extended Data Fig. 2 | Characterization of the melanoma–Treg-induced
vitiligo mouse model. a, Schematic diagram of the melanoma/Treg-induced
vitiligo mouse model. First the dorsal skin in the right flank of 8–9-week old
C57 mice was inoculated with B16F10 melanoma cells (Day 0), then CD4
depletion antibodies were injected on Days 4 and 10. The tumour was surgically
removed on Day 12 to prevent interference of subsequent analysis.
b, Representative hair coat images of control and vitiligo induced mice at
Day 300 after the vitiligo induction procedure. At 4 weeks after induction,
dorsal skin hair follicles close to the B16F10 tumour cell injection and surgical
removal sites started to show depigmentation as a result of wounding-induced
new hair growth. Then the depigmented hair follicles expanded and
eventually rendered the whole dorsal hair coat depigmented at ~Day 300.
c, Representative tail skin images of mice at Day 0, 45, 60, 90, 120, 240, 360,
480, 720 post melanoma/Treg-induced vitiligo induction procedure.
Melanocytes in mouse dorsal skin are located in hair follicles but not in the
epidermis; only in mouse tail skin are melanocytes located in both the hair
follicle and epidermis similar to human skin. Skin epidermis depigmentation is
the defining feature of vitiligo pathology. Therefore, we only used mouse tail
skin for vitiligo analysis. In tail skin, the depigmentation was initially patchy
and then progressed to eventually cover the entire epidermal surface, which
did not recover even more than a year later. d, Representative whole-mount
immunofluorescent staining and density plot images of DCT+ melanocytes and
CD8+ T cells in mouse tail skin epidermis at Day 0, 19, 26, 33 after the vitiligo
induction procedure. Prior to vitiligo induction, very few if any CD8+ T cells
could be detected in epidermis. Starting from Day 19 after the vitiligo
induction procedure, CD8+ T cells infiltration and small regions of melanocyte
loss could be observed. Note, melanocytes loss only occurred in regions where
CD8+ T cells locally aggregated into clusters; as the CD8+ T cells clusters
continuously expanded so did the regions of melanocyte loss correspondingly.
e, FACS quantifications of CD45+CD3+CD8+ T cells and CD117+ melanocyte in
mice with or without vitiligo induction confirmed the loss of melanocytes and
enrichment of CD8+ T cells after vitiligo induction. f, Representative whole-
mount immunofluorescent staining images and quantifications of DCT+
melanocytes and CD8+ T cells in mouse tail skin epidermis with combined or
individual B16F10 inoculation and CD4 depletion antibody treatment, showing
the melanoma/Treg-induced vitiligo model requires both B16F10 inoculation
and CD4 depletion antibody injection. g, Schematic diagram, representative
whole-mount immunofluorescent staining images and quantification of DCT+
melanocytes and CD8+ T cells in the skin epidermis of melanoma/Treg-induced
vitiligo model with or without CD8 depletion antibody treatment, showing
CD8+ T cells are responsible for melanocyte loss in the melanoma/Treg-
induced vitiligo mouse model. h, i, Representative whole-mount images
(h) and FACS profiles (i) with corresponding quantification of the percentage
of CD8+ T cells that express CD3 or CD11c in mouse epidermis, showing
majority of the CD8+ cells are CD3+ T cells, but not CD11c+ dendritic cells.
j, Representative whole-mount immunofluorescent staining images and
density plot images of CD8+ T cells and DCT+ melanocytes in tail skin epidermis
at Day 19, 26, 33 after vitiligo induction procedure. Note the continuously
expanding CD8+ T cell clusters in skin epidermis. k, Representative whole-
mount immunofluorescent staining images and quantifications of Ki67+, CD3+
T cells, and DCT+ melanocytes in tail skin epidermis at Day 33 after vitiligo
induction procedure. Enlarged image on the left represents border of lesion
skin, the right one represents lesion region. Quantification showed the
proliferation rates of CD8+ T cells at the border versus inside the clusters are
equivalent; and percentage of Ki67+ cells in melanocytes, T cells and
keratinocytes showed majority of the proliferating cells in skin epidermis are
keratinocytes. These data indicated that the continuous expansion of CD8+
T cell clusters mainly results from skin-infiltrated CD8+ T cells being actively
recruited into the border regions. Scale bars, 500 µm (c, d, f, g, j), 100 µm
(h, k). For exact p values, see Source Data. For statistics, p summary and sample
sizes, see Methods.
Article
Extended Data Fig. 3 | The melanoma–Treg-induced vitiligo mouse model
recapitulates hallmarks of human vitiligo. a, UMAP projection of skin
immune cells from patients with vitiligo and healthy donors. We identified 5
major clusters in immune cells, including two adaptive immune cell clusters
(CD4+ T cell and CD8+ T cell) and three innate immune cell clusters
(Langerhans cell, macrophage, dendritic cell 1/2). Other immune cells of very
low abundance in human skin, such as mast cells and γδ T cells, could also be
observed in our single-cell RNA-seq data, but owing to their limited number
these cell types were not analysed individually. b–e, Heat map (b), dot plot
(c), violin plot (d) and feature plot (e) analysis of signature genes in each
immune cell subtype. f, Quantification and comparison of five immune cell
types between progressive-state patients with vitiligo and healthy donors by
scRNA-seq analysis. g, Representative FACS profiles analysing immune cell
types in mice vitiligo skin, including epidermal TCRγδhigh and TCRγδlow γδ T cell
present in mouse. h, i, Quantification and comparison of immune cell types
between vitiligo (at Day 33) and control mice. Data in f-g showed the
predominantly enriched immune cell type is CD8+ T cell in both patients with
vitiligo and the melanoma/Treg-induced vitiligo mouse model, whereas the
other immune cell types did not show significant difference. j, Representative
whole-mount immunofluorescent staining images of CD4+ T cells in WT skin,
and Tregs in Foxp3 Cre;Ai14 skin, indicating at Day 33 of vitiligo analysis, the CD4+
T cell population have recovered from systematical CD4+ T cell depletion.
k, Heat map analysis of cytotoxicity-related genes expressed in skin CD8+
T cells of patients with vitiligo and vitiligo-induced mice, indicating that in
molecular level, this melanoma/Treg-induced vitiligo mouse model elicits
endogenous auto-reactive CD8+ cytotoxic T cells similar to human vitiligo. Ctl:
control, Vit: vitiligo. Scale bars, 200 µm (k). For exact p values, see Source Data.
For statistics, p summary and sample sizes, see Methods.
Extended Data Fig. 4 | See next page for caption.
Article
Extended Data Fig. 4 | Analysis of wild-type and Ifngr1 KO mice with three
different vitiligo induction methods. a–j, Analysis of WT and Ifngr1 KO
mice with melanoma/Treg-induced vitiligo model. a, Schematic diagram.
b, Representative tail skin images of WT and Ifngr1 KO mice at Day 60 after
vitiligo induction. c, d, Representative whole-mount images (c) and
quantification (d) of tail skin epidermal melanocytes and CD8+ T cells at Day 33
after vitiligo induction. e, f, Representative FACS profile of CD117+
melanocytes and CD3+CD8+ T cells (e), and quantification of CD117+
melanocytes (f) in the tail skin epidermis at Day 33 after vitiligo induction.
g, Whole-mount views of scales in tail skin epidermis (dotted line) in control
and vitiligo mice. This is the unit area we used to quantify melanocyte or CD8+
T cell density in wholemount analysis throughout this paper. h. Scatter plots
and correlation analysis of melanocyte number versus CD8+ T cell number in
each scale of WT and Ifngr1 KO tail skin at Day 33 after vitiligo induction.
i, Representative whole-mount images of DCT+ melanocytes in tail skin
epidermis of WT and Ifngr1 KO mice at 4 months after vitiligo induction. j, QPCR
analysis of GZMA, GZMB, PRF1, and CCL5 expression in skin CD8+ T cells isolated
from WT and Ifngr1 KO mice at Day 33 after vitiligo induction. k–p, Analysis of
WT and Ifngr1 KO mice with Pmel transgenic spontaneous vitiligo mouse
model. k, Schematic diagram. l–m, Representative tail skin images (l, P70) and
whole-mount immunofluorescent staining images (m, P42) of Pmel;WT and
Pmel; Ifngr1 KO mice. n, o, Representative FACS profiles (n) and quantification
(o) of CD117+ melanocytes and CD45+CD3+CD8+ T cells in tail skin epidermis of
WT, Ifngr1 KO, Pmel;WT and Pmel;Ifngr1 KO mice at P42. p, Scatter plots and
correlation analysis of melanocyte number versus CD8+ T cell number in each
scale in Pmel;WT and Pmel;Ifngr1 KO tail skin at P42. q–x, Analysis of WT and
Ifngr1 KO mice with adoptive T cell transfer-based vitiligo mouse model.
q, Schematic diagram. r, Quantification of CD117+ melanocytes and
CD45+CD3+CD8+ T cells in tail skin epidermis of WT mice after γ irradiation,
with or without WT Pmel T cell transfer or AVV-hPMEL intraperitoneal
injection. These data indicate sub-lethal irradiation alone, irradiation with
hPMEL-AAV alone, or irradiation with Pmel CD8+ T cell transfer alone did not
result in melanocyte loss or CD8+ T cell infiltration in WT tail skin epidermis.
s, Representative immunofluorescent staining images and FACS quantification
of epidermal melanocyte of mice post γ-irradiation 26 days with controls,
indicating after 5 Gy γ irradiation leads to ~2-fold increase of melanocyte
number in tail skin epidermis compared to untreated mice. t, Representative
FACS profiles and quantification of CD3+CD8+VB13+ Pmel T cells in spleen or
skin of WT and Ifngr1 KO mice at Day 26 with or without adoptive transfer-based
vitiligo model induction. These data showed although both WT and Ifngr1 KO
host mice contained the same number of transferred WT TCR VB13+ Pmel
T cells in the spleen, only WT host mice exhibit robust VB13+ Pmel T cell
infiltration in skin, whereas the Ifngr1 KO host mice did not. u–v, Representative
tail skin images (u, Day 60) and whole-mount immunofluorescent images
(v, Day 26) of WT and Ifngr1 KO mice post vitiligo model-induction. Scale bars,
500 µm. w, Representative FACS profiles and quantification of epidermal
CD117+ melanocytes in WT and Ifngr1 KO mice at Day 26 post vitiligo model-
induction. x, Scatter plots and correlation analysis of melanocyte number
versus CD8+ T cell number in each scale in WT and Ifngr1 KO mice at Day 26 post
vitiligo model-induction. Scale bars, 500 µm (b, c, g, i, l, m, s, u, v). For exact p
values, see Source Data. For statistics, p summary and sample sizes,
see Methods.
Extended Data Fig. 5 | IFNγ-responsive skin stromal cells are required for
vitiligo progression. a–c, Analysis of six main immune cell types in tail skin of
WT and Ifngr1 KO mice with different vitiligo induction methods. a, Day 33 after
melanoma/Treg-induced vitiligo model induction. b, P42 of Pmel;WT and
Pmel;Ifngr1 KO mice. c, Day 26 after adoptive transfer-based vitiligo
model-induction. d, Timeline of the graft and vitiligo induction assay.
Representative whole-mount immunofluorescent staining images of grafted
WT tail skin epidermis on WT host at Day 0 and Day 21 with or without vitiligo
induction were showed. These data indicate tail skin did not spontaneously
develop vitiligo after the graft procedure alone; only after vitiligo induction in
the host mice did the grafted skin develop vitiligo as indicated by melanocyte
loss and large amounts of CD8+ T cells infiltration. e, f, Lineage tracing
experiment of CD8+ T cells: e, Schematic diagram WT tail skin graft on
CD4 Cre;mTmG host and representative immunofluorescent staining images of
CD8+ T cells in the grafted skin epidermis after vitiligo induction; f, FACS
analysis and quantification of spleen cells from CD4Cre;Ai6 mouse to determine
GFP labelling efficiency in CD8+ T cells. Cells were pre-gated on CD45+ live
singlets, indicating that the infiltrated CD8+ T cells were derived from the host
mice rather than the grafted skin. These data indicate the graft infiltrated
CD8+ T cells were derived from the host mice rather than the grafted skin.
g, Quantification of melanocytes and CD8+ T cells number in grafted skin
without vitiligo model induction in host mice. Donor skin pairs grafted onto
the same host mouse are linked by lines. h, Representative immunofluorescent
staining of the junction region between grafted C57 tail skin and host dorsal
skin of a membrane-Tomato (mT) transgenic mouse. After full-thickness C57 tail
skin was grafted onto membrane-tdTomato (mT)-expressing host mice
(Rosa-mT), in which all host cells are genetically labelled as mT+, host derived
dermal cells indeed invaded into the grafted skin dermis. But neither K14+
keratinocytes nor DCT+ melanocytes migrated from the host to grafted skin.
i, Representative immunofluorescent staining image of pSTAT1 signal in
immune cell (CD45+), smooth muscle cell (a-SMA+), Langerhans cells
(Langerin+) of grafted Ifngr1 KO tail skin on WT host after vitiligo induction on a
host mouse. Enlarged image on the right represents Langerin+ cells in the
epidermis. Scale bars, 50 µm (d, e, h, i). For exact p values, see Source Data. For
statistics, p summary and sample sizes, see Methods.
Article

Extended Data Fig. 6 | See next page for caption.

Extended Data Fig. 6 | IFNγ-responsive skin fibroblasts are required for
driving vitiligo pathogenesis. a, Schematic diagram and representative FACS
profiles of skin cells isolation. Detailed gating strategies are described in
method section. QPCR analyses of cell-type-specific signature genes in FACS-
isolated populations include KRT14 for keratinocytes, DCT for melanocytes,
CD45 for immune cells, PDGFRA for fibroblasts, and CD31 for endothelial cells
respectively. QPCR validation of IFNGR1 knockout used FACS-purified
endothelial cells, keratinocytes, immune cells, melanocytes, fibroblasts and
CD8+ T cells from TekCre;IFNGR1 fl/fl, K14 Cre;IFNGR1 fl/fl, Csf1rCre;IFNGR1 fl/fl,
TyrCreER;IFNGR1 fl/fl, CD4 Cre;IFNGR1 fl/fl, and Pdgfra CreER;IFNGR1 fl/fl mice, all with WT
mice as controls. b, Representative whole-mount immunofluorescent staining
images of melanocytes and CD8+ T cells in control and six cell-type-specific
conditional knockout lines at Day 33 after vitiligo induction. c, Quantification
of skin CD8+ T cells in WT and six cell-type-specific knockout lines at Day 33
after vitiligo induction procedure based on wholemount staining. d, e,
Representative FACS profiles (d) and quantification (e) of CD117+ epidermal
melanocytes and CD3+CD8+ T cells in control and six cell-type-specific
conditional knockout lines post vitiligo induction. f, Schematic diagram of
melanoma/Treg-induced vitiligo procedure and representative tail skin images
(Day 60) of WT and PdgfraCreER ;IFNGR 1fl/fl cKO mice after vitiligo induction. Vit.
induc.: Vitiligo induction. Scale bars, 500 µm (b, f). For exact p values, see
Source Data. For statistics, p summary and sample sizes, see Methods.
Article

Extended Data Fig. 7 | See next page for caption.

Extended Data Fig. 7 | IFNγ-responsive fibroblasts are both necessary and
sufficient to mediate local recruitment of CD8+ T cells through secreted
chemokines. a–c, Analysis of WT and Pdgfra CreER;IFNGR1 fl/fl cKO mice with Pmel
transgenic spontaneous vitiligo mouse model. a, Schematic diagram and
representative tail skin images (P70) and representative wholemount images
(P42) of Pmel;WT and Pmel;Pdgfra CreER;IFNGR1 fl/fl cKO mice. b, Scatter plots and
correlation analysis of melanocyte number versus CD8+ T cell number in each
scale from Pmel;WT and Pmel;Pdgfra CreER;IFNGR1 fl/fl cKO mice at P42.
c, Representative FACS profiles and quantification of tail skin epidermal
CD117+ melanocytes and CD3+CD8+ T cells in WT, PdgfraCreER;IFNGR1 fl/fl cKO,
Pmel;WT and Pmel;Pdgfra CreER;IFNGR1 fl/fl cKO mice at P42. d–g, Analysis of WT
and Pdgfra CreER;IFNGR1 fl/fl cKO mice with adoptive transfer-based vitiligo model.
d, Representative FACS profiles and quantification of CD3+CD8+VB13+ Pmel
T cells in spleen or tail skin of WT and PdgfraCreER;IFNGR1 fl/fl cKO mice at Day 26
with or without adoptive transfer-based vitiligo model induction. e, Schematic
diagram and representative tail skin images (Day 60) and representative
wholemount immunofluorescent staining images (Day 26) of WT and
Pdgfra CreER;IFNGR1 fl/fl cKO mice after vitiligo induction. f, Scatter plots and
correlation analysis of melanocyte number versus CD8+ T cell number in each
scale WT and Pdgfra CreER;IFNGR1 fl/fl cKO mice at Day 26 after vitiligo induction.
g, Representative FACS profiles and quantification of tail epidermal CD117+
melanocytes, and quantification of tail epidermal CD117+ melanocytes and
CD3+CD8+ T cells of WT and Pdgfra CreER; IFNGR1 fl/fl cKO mice at Day 26.
h, Section immunofluorescent analysis of intradermally injected RFP-labelled
fibroblast. Note injected RFP+ fibroblasts located in lower dermis.
i, Quantification of T cell transwell migration assay showing T cells migration
ratio using control medium without fibroblast, corresponding to Fig. 3d.
j, Analysis of human and mouse vitiligo fibroblasts up regulated genes. Left,
heat map showing differential expressed genes in tail skin fibroblasts from WT
control, WT vitiligo-induced, and Ifngr1 KO vitiligo-induced mice. Middle,
volcano plot of differentially expressed genes in human fibroblasts from
progressive-state patients with vitiligo compared with healthy donors. Red
dots denote genes >1.5 fold upregulated (t test p < 0.01) in fibroblasts from
progressive-state patients with vitiligo. Right, heat map analysis of the 28
common upregulated secreted factors in fibroblasts from patients with vitiligo
and mice. k, Quantification of migration ratio from the T cell transwell
migration assay with various chemokines. l, Representative
immunofluorescent staining images of CXCL9 and CXCL10 signal with T cells
(CD3+) in the skin of patients with vitiligo. m, n, Representative
immunofluorescent staining images of CXCL9 (m) and CXCL10 (n) signals in
melanocytes (DCT+), fibroblasts (PDGFRA+), endothelial cells (CD31+), smooth
muscle cells (a-SMA+), keratinocytes (K14+), mononuclear phagocytes
(CD11c+), Langerhans cells (Langerin+), and T cells (CD3+) in the skin of patients
with vitiligo. Scale bars, 500 µm (a, e, l), 50 µm (h), 100 µm (m, n). For exact p
values, see Source Data. For statistics, p summary and sample sizes,
see Methods.
Article

Extended Data Fig. 8 | See next page for caption.

Extended Data Fig. 8 | Fibroblasts directly affect CD8+ T cells through the
CXCL9/CXCL10–CXCR3 axis in vitiligo. a, Schematic diagram of the
experimental procedure of lentivirus-infected cell type analysis. FACS profiles
and quantification showing majority of lentivirus-infected cell population in
the skin is fibroblast. b, QPCR analysis validated high knockdown efficiency of
shRNAs targeting CXCL9/CXCL10 in vitro. c, Top, experimental design of
in vivo knockdown efficiency detection in different dermal cell types. Bottom,
qPCR analysis in FACS-isolated virus-infected (RFP+) fibroblasts, endothelial
cells, and immune cells, indicating CXCL9 and CXCL10 expression were
specifically silenced in virus-infected fibroblasts, not in other cell types.
d, Representative whole-mount images showing dermis and overlying
epidermis with injection of indicated shRNA-expressing lentivirus. Box-
whisker plots of epidermal melanocyte number in each scale of mice tail skin
injected with the indicated shRNAs. e, FACS quantification of immune cell
types in tail skin of WT and PdgfraCreER;IFNGR1 fl/fl cKO mice with three different
vitiligo induction methods. f, Feature plot showing CXCR3 expression pattern
and quantification of CXCR3 in all cell clusters from patients with vitiligo and
healthy donors. g, Representative FACS profiles and quantification of CXCR3+
cells composition in vitiligo-induced mice (at Day 33). Scale bars, 500 μm
(d). For exact p values, see Source Data. For statistics, p summary and sample
sizes, see Methods.
Article
Extended Data Fig. 9 | Anatomically distinct dermal fibroblasts have
intrinsic differences in the IFNγ response. a, Heat map of upregulated genes
in human dermal fibroblasts from different anatomical regions after IFNγ
treatment (> 2-fold, t test p < 0.01). b, HOX expression pattern (based on
RNA-seq results) in human dermal fibroblasts from eight different anatomic
regions with or without IFNγ treatment during in vitro culture. c, QPCR analysis
of HOXB8, HOXC8, HOXB13, and HOXD11 in human dermal fibroblasts from
eight anatomic regions. d, JAK1, JAK2, IFNGR1, and STAT1 expression pattern
(based on RNA-seq results) in human dermal fibroblasts from eight different
anatomic regions with or without IFNγ treatment. e, QPCR analysis of TNFSF10,
IL6, and CTSH in human dermal fibroblasts from eight anatomically distinct
regions after IFNγ treatment. f, Left, representative whole-mount images and
quantification of hair follicle density in the dorsal and paw dorsal skin of mice
at P21. Right, representative section image and quantification of skin thickness
in the dorsal and paw dorsal skin of mice at P60. g, Representative
immunofluorescent images, and quantification of hair follicle located
melanocytes in ventral and paw ventral skin of control mice and vitiligo mice
(at Day 300). h, FACS analysis and quantification of skin CD3+CD8+ T cells and
CD117+ melanocytes from paw dorsal, paw ventral, dorsal, and ventral regions
at Day 300 after vitiligo induction. i, Schematic diagram of fibroblast isolation
and qPCR analysis of CXCL9 and CXCL10 in fibroblasts from four distinct
regions of long-term vitiligo mice. j, Quantification of migration ratio using
medium without fibroblast, corresponding to Fig. 4g. k, Relationship between
melanocyte number and infiltrated CD8+ T cell number in each scale after
vitiligo induction based on quantification data in Extended Data Fig. 4h. Scale
bars, 100 µm (f), 50 μm (g). For exact p values, see Source Data. For statistics,
p summary and sample sizes, see Methods.
Extended Data Fig. 10 | See next page for caption.
Article
Extended Data Fig. 10 | Mathematical modelling reveals that fibroblasts
direct collective CD8+ T cell local activity. a, The mathematic model
developed to predict local CD8+ T cell recruitment and clonal expansion
behaviour. The model is a 3D square lattice with two layers. The upper
epidermal layer contains T cells and melanocytes. The lower dermal layer
contains fibroblasts with different chemotactic abilities. The CD8+ T cell
population in skin is considered to be a decentralized system. Each CD8+ T cell
is equipped with the means of sensing a change in density. Over time, the
behaviour of each cell changes according to its state and the states of its
neighbouring cells and the surrounding signals. The collective pattern can be
globally modulated by changing the parameters governing local cell-cell
interactions. Once the T cell surrounding a melanocyte exceeds a threshold
number, melanocyte death and IFNγ secretion occur. The IFNγ signal induce
neighbouring fibroblast chemotactic effect to recruit nearby CD8+ T cells,
reaching the local CD8+ T cell density threshold for adjacent melanocyte
cytotoxicity and IFNγ secretion. This positive feedback loop between CD8+
T cells and fibroblasts could ensure T cell clonal expansion and vitiligo
progression. b, The mathematic model predicts the expansion process of a
single T cell clone. Five representative time points from T0 to T80 show the
initial T cell cluster state and the subsequent T cell expansion process over 80
time units. In model 1 with normal chemotactic fibroblasts, full spectrum of
T cell cluster formation and expansion patterns observed in WT vitiligo mouse
model were reproduced (first row, fibroblasts with normal chemotactic effect,
also shown in Supplementary Videos 1, 2). In model 2, in which the fibroblasts
were incapable of chemotaxis, we obtained patterns observed in
Pdgfra CreER ;IFNGR1 fl/fl cKO mice. In this model, although T cells could still
randomly aggregate in the epidermis, they failed to propagate this effect and
recruit more T cells to the initial site (second row, fibroblasts with no
chemotactic effect, also shown in Supplementary Videos 3, 4). In model 3, in
which the chemotactic effect of fibroblasts was turned down to 1/2 of the
normal value, random CD8+ T cell aggregates only recruited a limited number
of T cells, resulting in slow T cell cluster expansion and melanocyte loss in one
or two directions (third row, fibroblasts have 1/2 normal chemotactic effect,
also shown in Supplementary Videos 5, 6). c, The mathematic model predicts
large-scale T cell clone expansion over the long term. Four representative time
points from T0 to T450 show the initial state and the subsequent T cell cluster
expansion process over 450 time units. CD8+ T cells in the normal chemotaxis
model efficiently coordinate so as to achieve clonal expansion and melanocyte
clearance. CD8+ T cells in the no chemotaxis model fail to undergo clonal
expansion and melanocyte death is detected. CD8+ T cells in the weak
chemotaxis model (migration ability decreases to 1/2) generate small T cell
clones and expand slowly. d, The mathematic model predicts the T cell clone
distribution pattern under regional variant fibroblasts with different levels of
chemotactic effect to T cell. In this model, the fibroblasts marked in blue have a
normal chemotactic effect, and those marked in green have a weak
chemotactic effect. In this region, T cell chemotaxis (migration ability)
decrease to 1/2 of normal value. Six representative time points from T0 to T600
show the initial state and subsequent T cell cluster expansion (also shown in
Supplementary Video 7). The results show that the T cell clones are more likely
to expand and generate white patches on the normal chemotactic region. Both
white patches and T cell clone patterns are highly correlated with the regional
fibroblast variants. e, f, QPCR analysis validated high knockdown efficiency
(e) and the effect blocking IFNγ downstream signal (f) in vitro for shRNAs
targeting IFNGR1, JAK1, or STAT1. g, Representative epidermis whole-mount
immunofluorescent staining images and density plot of shRNA-mediated
knockdown assay, relative to the corresponding dermis in Fig. 4k. h, Box-
whisker plots and correlation analysis of T cell number versus percentage of
infected fibroblasts (upper panels), and melanocyte number versus
percentage of infected fibroblasts (lower panels) in each scale of in vivo mosaic
fibroblast knockdown experiment. i, Scatter plots of median of melanocyte
number versus percentage of infected fibroblasts in each scale of in vivo
mosaic fibroblast knockdown experiment. Scale bars, 500 µm(g). For exact
p values, see Source Data. For statistics, p summary and sample sizes,
see Methods.
 nature portfolio | reporting summary
Corresponding author(s): Ting Chen
Last updated by author(s): Sep 28, 2021

Reporting Summary
Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)

For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.

Software and code

Policy information about availability of computer code
Data collection BD FACSDiva (8.0.1), Nikon-A1R (4.13.00)

Data analysis Imaris (7.4.2), GraphPad Prism (6.0), R (3.5.1), RStudio (1.1.456), GO analysis (http://geneontology.org), FlowJo (v10.0.7), Seurat (V2.3.2),
TopHat (v2.0.13), Cufflinks (v2.2.1), STAR (v 2.6.1a), FeatureCount (v2.2.1). The code for the vitiligo progression model is available at: https://
github.com/hydrays/vitiligo.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and
reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.

Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
March 2021

- A description of any restrictions on data availability

- For clinical datasets or third party data, please ensure that the statement adheres to our policy

The single-cell RNA-sequencing data, mouse fibroblast RNA-seq data, and human fibroblast RNA-seq data have been deposited in the GSA with accession number
PRJCA006797 (https://ngdc.cncb.ac.cn/bioproject/browse/PRJCA006797).

Source data for all figures are provided with the paper.

1
 nature portfolio | reporting summary
Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design

All studies must disclose on these points even when the disclosure is negative.
Sample size No statistical method were used to predetermine sample size. Sample size was determined based on other similarly designed experiments
and results on cellular immunology analyses (Malik et al., 2017, Science Immunology), aiming for 3-10 mice or sample per group.

Data exclusions No data were excluded, with the exception of vitiligo mouse model induction experiment. Mice that fail to developed primary tumors after
intradermally inoculated B16F10 cells were excluded for further analysis. Mice with recurrent primary tumors after surgery were excluded for
further analysis.

Replication For in vivo experiment, biological replicates and independent mice from at least 2 independent litters were used. For in vitro experiments,
biological replicates and technical triplicates were used.

Randomization Participants, mice, and samples were assigned into different experimental groups randomly. In the experiments with specific genotypes,
samples were sorted into groups based on their genotypes. Mice were littermate when available or age matched and co-housed so as to
reduce additional variables beyond their genotype difference.

Blinding Investigator were blinded to the genotype of mice during vitiligo induction and data collection.

Reporting for specific materials, systems and methods

We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.

Materials & experimental systems Methods

n/a Involved in the study n/a Involved in the study
Antibodies ChIP-seq
Eukaryotic cell lines Flow cytometry
Palaeontology and archaeology MRI-based neuroimaging
Animals and other organisms
Human research participants
Clinical data
Dual use research of concern

Antibodies
Antibodies used anti-human Tyr (Ting Chen Lab made, 1:500)
anti-human KRT14 (Ting Chen Lab made, 1:1000)
anti-human DCT (Ting Chen Lab made, 1:500)
anti-human CD45 (BD, Cat: 11-9459-41, Clone: 2D1, Dilution:1:300)
Anti-human CD117 (BD, Cat 555714, Clone: YB5.B8, Dilution: 1:300)
anti-human pSTAT1 (CST, Cat: 9167, Clone: 58D6, Dilution: 1:500)
anti-human pdgfra (abcam, Cat: ab203491, Clone: EPR22059-270, Dilution: 1:500)
anti-human CD31 (abcam, Cat: ab24590, Clone: P2B1, Dilution: 1:500)
anti-human aSMA (Sigma, Cat: C6198, Clone: 1A4, Dilution: 1:500)
anti-human CD11c (abcam, Cat: ab 52632, Clone: EP1347Y, Dilution: 1:500)
March 2021

anti-human Langerin (abcam, Cat: ab192027, Clone: EPR15863, Dilution: 1:500)

anti-human CD3 (BD, Cat: 563423, Clone: UCHT1, Dilution: 1:300)
anti-human CD8 (BD, Cat: 561952, Clone: RPA-T8, Dilution: 1:300)
anti-mouse DCT, (Ting Chen Lab made, Dilution: 1:500)
anti-mouse pdgfra, (R&D, Cat: AF-307, Clone: APA5, Dilution: 1:300)
anti-mouse pdgfra, (ebioscience, Cat: 17-1401-81, Clone: APA5, Dilution: 1:300)
anti-mouse VB13, (BIolegend, Cat: 140703, Clone: MR12-4, Dilution:1:300 )
anti-mouse CD11c (Biolegend, Cat: 117316, Clone: N418, Dilution: 1:300)
anti-mouse CD45, (BD, Cat: 559864, Clone: 10-F11, Dilution: 1:300)

2
 anti-mouse CXCR3, (Biolegend, Cat: 126506, Dilution: 1:300)
anti-mouse CXCR3, (Biolegend, Cat: 126511, Dilution: 1:300)

nature portfolio | reporting summary

anti-mouse MHCII, (Biolegend, Cat: 107629, Dilution: 1:300)
anti-mouse CD4, (Biolegend, Cat: 100451, Dilution: 1:300)
anti-mouse F4/80, (Biolegend, Cat: 123116, Dilution: 1:300)
anti-mouse CD11c, (Biolegend, Cat: 117310, Dilution: 1:300)
anti-mouse/human CD207, (Biolegend, Cat: 144206, Dilution: 1:300)
anti-mouse/human CD11b, (Biolegend, Cat: 101215, Dilution: 1:300)
anti-mouse TCRγδ, (Biolegend, Cat: 118108, Dilution: 1:300)
anti-mouse TCRγδ, (eBioscience, Cat:11-5711-82, Dilution: 1:300)
anti-mouse CD31, (ebioscience, Cat: 25-0311-81, Clone: 390, Dilution: 1:300)
anti-mouse aSMA, (Sigma, Cat: C6198, Clone: 1A4, Dilution: 1:500)
anti-mouse pSTAT1, (CST, Cat: 9167, Clone:58D6, Dilution: 1:500)
anti-mouse KRT14 (Ting Chen Lab made, Dilution: 1:1000)
anti-mouse CD117 (ebioscience, Cat: 13-1171-82, Clone: 2B8, Dilution: 1:200)
anti-mouse CD3 (Biolegend, Cat: 100220, clone: 17A2, Dilution: 1:300)
anti-mouse CD8a (ebioscience, Cat: 17-0081-83, Clone: 53-6.7, Dilution: 1:300)
anti-mouse Ki67 rat (eBioscience, Cat: 14-5698-82, Clone: SolA15, Dilution: 1:500)
Anti-human FOXP3 (abcam, Cat: ab20034, Dilution: 1:300)
Anti-human KLRB1 (Biosource, Cat: MBS2013141, Dilution: 1:300)
Anti-human CD40LG (Biosource, Cat: MBS4381431, Dilution: 1:300)
Anti-human GZMA (abcam Cat: ab10870, Dilution:1:300 )

The signal of human pSTAT1, human CD8a, human pdgfra, human CD11c, human Langerin, human CD3e, human Foxp3, human
KLRB1, human mouse pdgfra, mouse pSTAT1 was amplified by ABC Kit and TSA Kit.

Validation All the commercial antibodies were validated by the suppliers as indicated in the quality assurance literature provided by the supplier
either through flow cytometry or immunofluorescence assays.

The home made antibodies were validated by immunofluorescence assays in this study (Tyr, 1:500) and previous papers (Krt14,
1:1000, Xu et al, eLife, 2015; DCT, 1:500, Lu et al., eLife, 2020).

Eukaryotic cell lines

Policy information about cell lines
Cell line source(s) B16F10 cells (ATCC, CRL-6475) were purchased from ATCC. 293FT cells (Thermo Fisher Scientific, Cat #R70007) were
purchased from Thermo Fisher Scientific.

Authentication All cell lines were kept at low passages in order to maintain their identity.

Mycoplasma contamination All cell lines were tested negative for mycoplasma contamination.

Commonly misidentified lines No commonly misidentified cell lines were used in this work.
(See ICLAC register)

Animals and other organisms

Policy information about studies involving animals; ARRIVE guidelines recommended for reporting animal research
Laboratory animals C57BL/6 mice were purchased from Charles River Laboratories. K14Cre mice (MGI: 1926500) and K14H2BGFP mice (MGI: 5286142)
were kindly provided by Elaine Fuchs. IFNGR1 KO mice (Jackson Stock NO: 003288) and TekCre mice (Jackson Stock NO: 004128)
were kindly provided by Dr. Feng Shao. OT-1 mice (Stock NO: 003831) were kindly provided by Dr. Liang Chen. PMEL TCR
transgenetic mice (Stock NO: 005023) and PdgfraCreER (Stock NO: 018280), Csf1rCre (Stock NO: 021024), TyrCreER (Stock NO:
012328), CD4Cre (Stock NO: 017336), Foxp3Cre (Stock NO: 008694), IFNGR1 flox (Stock NO: 025394), Ai14 (Stock NO: 007914), Ai6
(Stock NO: 007906) and Rosa-stop-mTmG (Stock NO: 007676) mice were obtained from The Jackson Laboratory. Mice were group
housed (up to 5 animals per cage) on a 12:12 hour light-dark cycle, with23˚C-25˚C, 50%-56% humidity. Unless otherwise specified, all
experiments used balanced groups of male and female mice, and all experiments are conduct and compared using mice of the same
age at 8-9 weeks.

Wild animals This study did not involve wild animals.

Field-collected samples This study did not involve field-collected samples.

Ethics oversight Mice were bred and maintained in National Institute of Biological Sciences specific pathogen-free facility in accordance with the
March 2021

Guide for the Care and Use of Laboratory Animals of the NIBS. Procedures were approved by the Laboratory Animal Management
Committee of NIBS and were in compliance with all relevant ethical regulations.
Note that full information on the approval of the study protocol must also be provided in the manuscript.

3
Human research participants

nature portfolio | reporting summary

Policy information about studies involving human research participants
Population characteristics Seven male and three female patients who were pathologically diagnosed with vitiligo were enrolled in this study. Their ages
ranged from 6 to 55, with a median age of 24. Among these patients, four were diagnosed to be in active stage, while the
other six were not sure. All the patients were newly diagnosed with vitiligo and none of the patients had received therapy or
UVB treatment. A 3×3 mm2 biopsy was taken from each patient in the junction region between depigmented lesion region
and pigmented perilesional region. Biopsy with the same size was also taken from 5 healthy donors with other surgical
operation as control. The clinical characteristics of these patients and healthy donors were summarized in Extended Data
Figure 1c.

In this study of anatomically distinct human fibroblasts, we collected skin biopsies from human embryos (female, 22 to 24
weeks of gestation).

Recruitment Vitiligo patients who were scheduled to have surgery for pathological analysis in Beijing hospital were recruited if they
consented to donate tissue. All the patients were newly diagnosed with vitiligo and none of the patients had received therapy
or UVB treatment. All vitiligo skin tissue used are based on their availability, which would not affect our conclusion. Healthy
donors who were scheduled to have surgery for plastic operation in Beijing hospital were recruit if they consented to donate
tissue. All healthy donors’ tissue used are based on their availability, which would not affect our conclusion.

Research donors of human embryo were recruited from Peking University Third Hospital. Before giving consent, donors have
a suitable opportunity to receive proper counseling about the implication of the donation and potential risk. Embryos were
collect with written informed consent from the donors in this study. Both human embryo used are based on their availability,
which would not affect our conclusion.

Ethics oversight The study of single cell RNA-seq and immunofluorescent analysis of human vitiligo skin biopsy was reviewed and approved by
Ethics Committee of Beijing Hospital. For minors, informed consent was obtained from the LAR’s of the minors. The study of
anatomically distinct human fibroblast was reviewed and approved by the Peking University Third Hospital Medical Science
Research Ethics Committee.
Note that full information on the approval of the study protocol must also be provided in the manuscript.

Flow Cytometry
Plots
Confirm that:
The axis labels state the marker and fluorochrome used (e.g. CD4-FITC).
The axis scales are clearly visible. Include numbers along axes only for bottom left plot of group (a 'group' is an analysis of identical markers).
All plots are contour plots with outliers or pseudocolor plots.
A numerical value for number of cells or percentage (with statistics) is provided.

Methodology
Sample preparation For FACS quantification and collection of epidermal immune cells including CD8+ T cells, CD4+ T cells, Langerhans cells, and
γδ T cells, skin from middle section of the tail was harvested, flattened, cut into 1x1 cm pieces, and then floated dermis side
down in 2 mL of 2.4 U/mL dispase solution and incubated at 37˚C with shaking at 80 rpm for 50 min. To obtain most of the
epithelial cells, the epidermis was then carefully removed from dermis along the anterior-posterior direction with a fine-
tipped tweezer. The removed epidermis was floated dermis side down in 3 mL TrypLE™ Express Enzyme (Gibco) at 37˚C for
10 min, afterwards 5% FBS was added to neutralize TrypLE.

For FACS quantification of epidermal melanocyte, melanocytes located in the hair follicle must be removed. We found out
prior to skin collection, wax using VEET wax paper facilitates subsequent hair follicle separation from epidermis. Waxed tail
skin was flattened and floated dermis side down in 2 mL of 2.4 U/mL dispase solution at 37˚C with shaking at 80 rpm for 50
min. To efficiently obtain hair follicle-free epidermis, the epidermis was swiftly peeled from the dermis along the posterior-
anterior direction with a fine-tipped tweezer. The occasional remaining hair follicles attached to epidermis were removed
using tweezers. The hair follicle-free epidermis was then floated dermis side down in 3 mL Trypsin-EDTA (0.25%) at 37˚C for
10 min, afterwards 5% FBS was added to neutralize trypsin.

Single cell suspensions were obtained by repeatedly aspirating and dispensing with a pipette for 20 times followed by
March 2021

filtering through 40 μm strainers. The cells were then stained with anti-mouse CD117 (ebioscience, Cat: 13-1171-82, Clone:
2B8, Dilution: 1:200) for melanocyte detection, anti-mouse CD45 (BD, Cat: 559864, Clone: 10-F11, Dilution: 1:300) for
immune cell detection; anti-mouse CD3 (Biolegend, Cat: 100220, clone: 17A2, Dilution: 1:300) and anti-mouse CD8a
(ebioscience, Cat: 17-0081-83, Clone: 53-6.7, Dilution: 1:300) for CD8+ T cell detection; anti-mouse CD3 and anti-mouse CD4,
(Biolegend, Cat: 100451, Dilution: 1:300) for CD4+ T cell detection; anti-mouse MHCII (Biolegend, Cat: 107629, Dilution:
1:300) and anti-mouse CD207, (Biolegend, Cat: 144206, Dilution: 1:300) for Langerhans cell detection; anti-mouse CD3 and
anti-mouse TCRγδ (Biolegend, Cat: 118108, Dilution: 1:300) antibodies for γδ T cell detection.

For FACS quantification and collection of dermal fibroblasts, endothelial cells, immune cells including macrophage and

4
 dendritic cells from tail skin, skin from middle section of the tail was harvested, flattened, cut into 1x1 cm pieces, and then
floated dermis side down in 2 mL of 2.4 U/mL dispase solution and incubated at 37˚C with shaking at 80 rpm for 50 min. To

nature portfolio | reporting summary

get rid of epidermis and hair follicles, the epidermis was then carefully removed from dermis along the anterior-posterior
direction with a fine-tipped tweezer. The dermis was then placed in 10 mL of 2 mg/mL collagenase (Sigma-Aldrich, C2674)
and incubated at 37˚C with shaking at 80 rpm for 1 h, afterwards 5% FBS solution was used for neutralization. Single cell
suspensions were obtained by repeatedly aspirating and dispensing with a pipette for 20 times followed by filtering through
70 μm and 40 μm strainers. The cells were stained with anti-mouse CD31, (ebioscience, Cat: 25-0311-81, Clone: 390,
Dilution: 1:300), anti-mouse-CD45, and anti-mouse pdgfra, (ebioscience, Cat: 17-1401-81, Clone: APA5, Dilution: 1:300)
antibodies for endothelial cell, total immune cell, and fibroblast detection, respectively; anti-mouse/human CD11b
(Biolegend, Cat: 101215, Dilution: 1:300) and anti-mouse F4/80 (Biolegend, Cat: 123116, Dilution: 1:300) for macrophage
detection; anti-mouse MHC-II and anti-mouse CD11c, (Biolegend, Cat: 117310, Dilution: 1:300) for dendritic cell detection.

For analysis and isolation of immune cells from spleens, the spleen was mechanically dissociated and the red blood cells were
lysed with RBC lysis buffer according to the manufacturer’s instructions (Beyotime). Single cell suspensions were obtained by
filtering with strainers (40 μm) and then cells were stained with anti-mouse CD3, anti-mouse CD45, and anti-mouse CD8
antibodies. anti-mouse VB13 (BIolegend, Cat: 140703, Clone: MR12-4, Dilution:1:300 ) antibody was used in Pmel T cell
adoptive transfer experiments.

For CXCR3 expression analysis in skin immune cell types of vitiligo mice, the entire tail skin was harvested, flattened, then
floated dermis side down in 5 mL Trypsin-EDTA (0.25%) at 37˚C for 30 min. The skin was inverted and scraped with a scalpel
to get total skin cell suspensions. Single cell suspensions were obtained by filtering with strainers (70 μm and 40 μm), then
cells were co-stained with anti-mouse CXCR3, (Biolegend, Cat: 126506, Dilution: 1:300) against immune cell markers
mentioned above for further analysis.

For FACS analysis of CD8+ T cells and melanocytes from mouse dorsal, ventral, paw dorsal, and paw ventral skin, skin biopsies
were harvested, flattened, and floated dermis side down in 3 mL of TrypLE and incubated at 37˚C with shaking at 80 rpm for
30 min. The skin epidermis was gently scraped with a scalpel to get the epithelial cell suspensions. Single cell suspensions
were obtained by filtering with strainers (40 μm), then cells were stained with anti-mouse CD45, CD3, CD8, and CD117
antibodies for further analysis.

For re-isolation of injected RFP-labeled dorsal or paw fibroblasts from IFNGR1 KO mouse tail, tail skin was separated into
epidermis and dermis post dispase treatment. After digestion of dermis with collagenase and neutralization, single cell
suspensions were obtained by filtering through 70 μm and 40 μm strainers. Then RFP+ fibroblasts were isolated for further
qPCR analysis.

Instrument BD FACSAria III, BD FACSAria II

Software BD FACSDiva (8.0.1), FlowJo (v10.0.7)

Cell population abundance For sorting for human single cell RNA sequencing, more than 10,000 immune cells and 2,000 melanocytes were collected.
The purity was more than 95% and was verified by single cell RT-qPCR. For sorting for mouse RNA-seq, at least 200,000
fibroblast were collected. The purity was verified by expression of Pdgfra and Krt14 with RT-qPCR. For immune analysis in
mice, 1,000 CD45+ singlets were collected. For CD117+ melanocytes analysis, 20,000 singlets were collected.

Gating strategy In the FACS for single cell RNA sequencing, cells were gated on live (DAPI-) and FSC/SSC- area. CD45+ and c-Kit+ were used to
enrich immune cells and melanocytes; CD45-, c-Kit- cells are predominantly keratinocytes. In dermis, CD45 was used to
distinguish immune cells from dermal cells (Extended Data Figure 1b).

In the FACS for qPCR validation and mouse fibroblast RNA sequencing, cells were gated on live (DAPI-) and FSC/SSC- area.
CD45+ and c-Kit+ markers were used to enrich immune cells and melanocytes; CD45-, c-Kit- cells are predominantly
keratinocytes. In dermis, CD45, CD31 were used to distinguish immune cells and endothelial cells from fibroblast, PDGFRA
were used to enrich fibroblasts (Extended Data Figure 6a).

For immune analysis, cells were gated on live (DAPI-) and FSC/SSC- area. CD45+ were used to enrich immune cells; CD3
+(medium), CD8+ cells are predominantly CD8+ T cells; CD3+(medium), CD4+ cells are predominantly CD4+ T cells;
F4/80+, CD11b+ cells are predominantly Macrophage; CD207+, MHC-II+ cells are predominantly Langerhans cells; TCR γδ+,
CD3+ cells are predominantly γδ T cells; CD11c+, MHC-II+ cells are predominantly dendritic cells (Extended Data Figure 3f).
Positive and negative gates were set using fluorescence minus one (FMO) background intensity controls. Fluorophores were
chosen to minimize spectral overlap.

Tick this box to confirm that a figure exemplifying the gating strategy is provided in the Supplementary Information.
March 2021