Articles

DOI: 10.1038/s41564-017-0006-5

Gut-homing Δ42PD1+Vδ2 T cells promote

innate mucosal damage via TLR4 during acute
HIV type 1 infection
Allen Ka Loon Cheung1, Hau-yee Kwok1, Yiru Huang1, Min Chen1,2, Yufei Mo1, Xilin Wu1, Ka-shing Lam1,
Hoi-Kuan Kong3, Terrence Chi Kong Lau   3, Jingying Zhou1, Jingjing Li1, Lin Cheng1,4, Boon Kiat Lee1,
Qiaoli Peng4, Xiaofan Lu5, Minghui An6, Hui Wang4, Hong Shang6, Boping Zhou4, Hao Wu5, Aimin Xu7,
Kwok-Yung Yuen1 and Zhiwei Chen   1,4*

The innate immune cells underlying mucosal inflammatory responses and damage during acute HIV-1 infection remain incom-
pletely understood. Here, we report a Vδ​2 subset of gut-homing γ​δ​T cells with significantly upregulated Δ​42PD1 (a PD1 iso-
form) in acute (~20%) HIV-1 patients compared to chronic HIV-1 patients (~11%) and healthy controls (~2%). The frequency of
Δ​42PD1+Vδ​2 cells correlates positively with plasma levels of pro-inflammatory cytokines and fatty-acid-binding protein before
detectable lipopolysaccharide in acute patients. The expression of Δ​42PD1 can be induced by in vitro HIV-1 infection and is
accompanied by high co-expression of gut-homing receptors CCR9/CD103. To investigate the role of Δ​42PD1+Vδ​2 cells in vivo,
they were adoptively transferred into autologous humanized mice, resulting in small intestinal inflammatory damage, probably
due to the interaction of Δ​42PD1 with its cognate receptor Toll-like receptor 4 (TLR4). In addition, blockade of Δ​42PD1 or TLR4
successfully reduced the cytokine effect induced by Δ​42PD1+Vδ​2 cells in vitro, as well as the mucosal pathological effect in
humanized mice. Our findings have therefore uncovered a Δ​42PD1–TLR4 pathway exhibited by virus-induced gut-homing Vδ​2
cells that may contribute to innate immune activation and intestinal pathogenesis during acute HIV-1 infection. Δ​42PD1+Vδ​2
cells may serve as a target for the investigation of diseases with mucosal inflammation.

H
uman Vγ​9Vδ​2 T cells (Vδ​2) are unconventional gamma-delta
T lymphocytes (γ​δ​-T) that are associated with gut infiltration
and inflammation. γ​δ​T cells express T cell receptor rearranged
with the T cell receptor gamma (TCRG) and delta (TCRD) genes and
comprise 3–10% of human peripheral blood lymphocytes, with the Vδ​
2 subset dominant over Vδ​1 in healthy individuals. Vδ​1 cells reside in
the gut epithelium and maintain homeostasis and surveillance, while
Vδ​2 cells have a critical role in important innate and adaptive immune
responses against pathogens and in sterile inflammation, includ-
ing cytolysis of infected cells and rapid production of cytokines1. In
vivo activation of Vδ​2 cells in macaques led to accumulation in the
gut and the onset of antimicrobial responses. This is also reported in
Crohn’s disease and inflammatory bowel disease, where Vδ​2 cells were
found to be gut-homing and might be responsible for the pathology
observed in small intestinal and colonic tissues2,3. Apart from the rec-
ognition of phosphoantigen that leads to their activation, Vδ​2 cells can
crosstalk with B cells or dendritic cells (DCs), migrate to mucosal sites,
and serve as antigen presenting cells (APCs) to α​β​-T cells1,2,4.
During acute human immunodeficiency virus (HIV)-1 infec-
tion, it is well documented that patients’ small intestines undergo
‘HIV enteropathy’ characterized by diarrhoea, increased gastroin-
testinal inflammation, enhanced intestinal permeability and mal-
absorption5. By histological analysis, the intestinal villi are often
seen with atrophy and blunting, epithelial damage and infiltrating
lymphocytes in both patients and acute simian immunodeficiency
virus (SIV)-infected macaques, and this can occur in the absence
of microbial translocation6,7. Inflammatory mediators such as infil-
tration and activation of CD8+ T cells, pro-inflammatory cytokines
and CD4 T cell depletion are believed to contribute to these drastic
changes in the small intestine following local immune activation8–11.
While the Vδ​1 subset of γ​δ​T cells undergoes rapid expansion in
mucosal tissues due to immune activation12, the circulating Vδ​2
cells also increases within days to a few weeks after infection but
begin to diminish by 6 weeks post-infection in an SIV model or
in human intestines13,14. Interestingly, during HIV-1 infection, cir-
culating Vδ​2 cells appear to have potential gut-homing properties,
as demonstrated by high CCR5 expression and upregulated CCR9
and CD10314,15. The depletion of Vδ​2 cells in the gut, however, may
be mediated through gp120, which becomes abundant in the intes-
tines due to replicating virus and infected CD4 T cell death in acute

1
​AIDS Institute, Research Center for Infection and Immunity, Department of Microbiology, Li Ka Shing Faculty of Medicine, University of Hong Kong,
999097 Hong Kong SAR, China. 2​Yunnan Center for Disease Control and Prevention, 650000 Kunming, Yunnan Province, China. 3​Department of
Biomedical Sciences, City University of Hong Kong, 999097 Hong Kong SAR, China. 4​HKU-AIDS Institute Shenzhen Research Laboratory and AIDS Clinical
Research Laboratory, Guangdong Key Laboratory of Emerging Infectious Diseases, Shenzhen Key Laboratory of Infection and Immunity, Shenzhen Third
People’s Hospital, 518000 Shenzhen, China. 5​STD/HIV Research Laboratory and Department of Infectious Diseases, Beijing Key Laboratory of HIV/AIDS
Research, Beijing You-An Hospital, Capital Medical University, 100000 Beijing, China. 6​Key Laboratory of AIDS Immunology of National Health and Family
Planning Commission, Department of Laboratory Medicine, The First Affiliated Hospital of China Medical University, China Medical University, 110000
Shenyang, China. 7​Department of Pharmacology & Pharmacy and Department of Medicine, Research Centre of Heart, Brain, Hormone and Healthy Aging,
Li Ka Shing Faculty of Medicine, University of Hong Kong, 999097 Hong Kong SAR, China. *e-mail: zchenai@hku.hk

Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology 1389

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles
HIV-1 infection, and subsequently acts on CCR5 expressed by these
cells14,16. As a result, an inversion of the Vδ​2:Vδ​1 subsets ratio occurs
shortly after HIV-1 infection in peripheral blood, and is not readily
restored with highly active antiretroviral treatment (HAART)12,14,17.
The functions of Vδ​2 cells are dysregulated during this time, when
they have an increased ability to produce IL-17 but have impaired
cytotoxicity and IFN-γ​ response18–21. In particular, the ability of Vδ​2
cells to crosstalk with DCs is significantly impeded as a result of
HIV-1 infection and aberrant cytokines being produced22. However,
apart from phenotypic or functional changes, little is known of the
mechanism by which Vδ​2 cells perform their immune stimulatory
role once they migrate to the gut during early HIV-1 infection2,23.
Interactions with other cell types, such as DCs, follicular T helper
and conventional T cells that may contribute to HIV pathogenesis,
are proposed2,4, but defining the interacting receptor/ligand respon-
sible for mediating pathogenic effects is critical in understanding
the role of Vδ​2 cells during acute HIV-1 infection.
We recently identified an alternatively spliced isoform of human
programmed death 1 (PD-1) containing a 42-nucleotide deletion—
Δ​42PD1— which is capable of inducing pro-inflammatory cyto-
kines from human peripheral blood mononuclear cells (PBMCs)
when expressed as a recombinant soluble protein24. Using in-house
anti-Δ​42PD1 monoclonal antibodies25, we demonstrate here that Δ​
42PD1 is significantly expressed on up to 20% of peripheral Vδ​2
cells in acute HIV-1 patients, which correlates with the level of
plasma pro-inflammatory cytokines (tumour-necrosis factor
(TNF)-α​, interleukin (IL)-6, IL-1β​, interferon (IFN)-α​). Further
analysis showed that one-third of Δ​42PD1+Vδ​2 cells co-express
the gut-homing receptors CCR9/CD103 in patient samples. Thus,
we hypothesized that Δ​42PD1 plays a role in early immune activa-
tion during acute HIV-1 infection in the gut. To test this, we used
a humanized mouse model where HIV-1-induced Δ​42PD1+Vδ​2
cells transmigrate to the intestines, promoting inflammation dam-
age and increased IL-6 expression. Importantly, we identified that
the innate Toll-like receptor (TLR)4/MD2 complex is probably the
cognate receptor for Δ​42PD1 by which pro-inflammatory cytokines
are induced in monocyte-derived DCs. This study thus discovers
a previously unrecognized mechanism exhibited by gut-homing
Δ​42PD1+Vδ​2 cells that may contribute to innate immune activation
and mucosal damage during acute HIV-1 infection.
Results
Increased Δ42PD1 expression on γδ T cells in acute and chronic
HIV-1-infected patients. To study the biological role of Δ​42PD1,
we recently generated two anti-Δ​42PD1 monoclonal antibodies for
immune assays25. Lymphocytes freshly isolated from healthy indi-
viduals were tested by multicolour flow cytometry and gating on
specific cell markers, CD4+ T, CD8+ T, B, NK, NKT and γ​δ​T cells
(Supplementary Fig. 1a). Δ​42PD1 was found to be expressed mainly
on ~6% of γ​δ​T cells among these cell types (Supplementary Fig. 1b).
We also found that Δ​42PD1 is exclusively expressed on the Vδ​2 but
not the Vδ​1 subset of γ​δ​T cells (Supplementary Fig. 1c).
To determine the role of Δ​42PD1 in HIV/AIDS, we examined
two cohorts of HIV-1 acutely infected patients (n =​ 57) and one
cohort of chronic HIV-1 patients (n =​ 50) for the expression of
Δ​42PD1 on Vδ​2 cells from frozen PBMC samples and compared
with healthy controls (n =​ 22) (Supplementary Table  1). By flow
cytometry, the frequency of CD3+Vδ​2 cells in the PBMCs of the
acute patients was found to be 2.62 ±​  1.99% (mean  ±​  s.d.), which
is significantly higher than in chronic patients (1.61 ±​  1.73%,
P <​  0.001) but comparable to healthy controls (2.65  ±​  1.49%,
P =​  0.5225) (Fig.  1a,b). When we further examined the expres-
sion of Δ​42PD1, the proportions of Δ​42PD1+Vδ​2 cells were sig-
nificantly higher in acute HIV-1 patients (18.53 ±​  10.26%) than
in chronic patients (11.72 ±​  9.95%, P <​ 0.001) and healthy con-
trols (2.18 ±​  1.41%, P <​  0.001) (Fig. 1c and Supplementary Fig. 2).
1390 Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NaTuRe MICRObIOLOgy
However, a higher frequency of CD3+Vδ​1 cells than CD3+Vδ​2 was
observed in acute patients, which contrasts with the predominance
of Vδ​2 over Vδ​1 in healthy individuals (Fig. 1d). This is consistent
with a disproportionate change in Vδ​2 and Vδ​1 cells in acute and
chronic HIV-1 patients, as reported previously12,14,17. This difference
was more dramatic in chronic patients, where an expansion of Vδ​1
cells was observed (Fig. 1d). Increased CD8+ T cell activation was
evident in acute and chronic HIV-1 patients on measuring CD38 and
HLA-DR expression (Fig. 1e). In addition, a comparison between
two stages of acute HIV-1 infection based on Fiebig status showed
significantly higher Δ​42PD1+Vδ​ 2 cells in seronegative patients
(Fiebig I–II) than in seropositive (Fiebig III–V) patients, indicating
that the rapid onset of Δ​42PD1 expression probably occurred early
(<​20 days) after HIV-1 infection (Fig. 1f). We also performed cor-
relation analyses between the frequency of Δ​42PD1+Vδ​2 cells from
acute and chronic patients with viral load, CD4:CD8 ratio, CD4
count, CD8 count, frequency of CD38+HLA-DR+ CD4+ and CD8+
T cells, but only the frequency of Δ​42PD1+Vδ​2 cells and viral load
in acute patients had a statistically significant positive correlation
(Supplementary Fig. 3).
Due to the increased CD8+ T cell activation observed in acute
patients, we then analysed their plasma cytokine levels, because cyto-
kine storm is one of the landmark characteristics of early HIV-1 infec-
tion26. Surprisingly, we found that the frequency of Δ​42PD1+Vδ​2
cells was positively correlated with plasma levels of TNF-α​, IL-6,
IL-1β​and IFN-α​ (Fig. 1g), suggesting a potential role of Δ​42PD1 in
innate immune activation. The endotoxin lipopolysaccharide (LPS)
level in acute patient plasma (0.84 ±​ 1.15 endotoxin unit (EU) ml–1),
however, was similar to healthy controls (0.79 ±​ 0.69 EU ml–1), consis-
tent with the lack of microbial translocation during the acute phase
of infection27. We therefore concluded that Δ​42PD1+Vδ​2 cells might
play a role in immune activation soon after HIV-1 primary infection,
probably before seroconversion and microbial translocation.
Δ42PD1 expression is regulated by cytokines induced by HIV-1
infection. To determine if HIV-1 infection can stimulate the expres-
sion of Δ​42PD1 on Vδ​2 cells, we performed in vitro viral infec-
tion of healthy human PBMCs. Fresh PBMCs were infected with
X4-tropic HIV-1NL4-3, R5-tropic HIV-1Henan or mock. Detection of
proviral DNA by qPCR, long terminal repeat (LTR) transcripts by
quantitative reverse-transcription real-time PCR (qRT–PCR) and
intracellular p24 signals in CD4+ T cells by flow cytometry indi-
cated successful infection (Supplementary Fig.  4). By 3 days after
HIV-1 infection of PBMCs, ~10% of Vδ​2 cells expressed Δ​42PD1,
of which ~50% of cells were CD69+, indicating activation (Fig. 2a).
Other viruses, including human cytomegalovirus (CMV), HIV
pseudoviruses with vesicular stomatitis virus glycoprotein (VSV-
G) or ADA envelope, could also induce Δ​42PD1 expression on Vδ​
2 cells, suggesting a generalized response (Supplementary Fig. 5a).
In a time-course experiment, induction of Δ​42PD1 expression on
Vδ​2 cells reached the highest level by day 3 post-infection (p.i.) as
compared to mock, and remained so at day 6 p.i. for both HIV-1NL4-3
and HIV-1Henan (Fig. 2b). Moreover, cell numbers of CD3+Vδ​2+ and
Δ​42PD1+ CD3+Vδ​2+ cells were increased by an average of two- and
ninefold, respectively, after HIV-1 infection compared to mock
(Supplementary Fig. 5b,c). However, direct application of HIV-1 to
purified γ​δ​T cells resulted in neither induction of Δ​42PD1 nor suc-
cessful infection, due mainly to undetectable intracellular p24. In
contrast, HIV-1 induction resulted in the upregulation of CD4 on
Δ​42PD1+CD3+Vδ​2+ cells (Supplementary Fig.  6a,b) but not CD8
(Supplementary Fig. 6c,d). Therefore, it is possible that certain sol-
uble factors resulting from infected PBMCs probably upregulated
Δ​42PD1 expression on Vδ​2 cells.
Previous studies have shown that γ​δ​T cells can be activated by
cytokines such as IL-12 and IL-1528. Our multiplex cytokine analysis
also revealed IL-12 release in supernatants after HIV-1NL4-3 infection
NaTuRe MICRObIOLOgy Articles
a b c
15 P = 0.5225 80 ***

∆42PD1+ CD3+ Vδ2+ (%)

1.05
*** *** *** ***

CD3+ Vδ2+ (%)

60
10

Vδ2–PE
SSC

40
5
20
88.6
0 0
FSC CD3-PB Acute Chronic Healthy Acute Chronic Healthy

0.313 7.53 d e
∆42PD1-AF647
***

CD38+HLA–DR+ CD8+ (%)

40 80
*** *** ***
Isotype

Vδ1
30 Vδ2 60

CD3+ (%)
**
67.2 20 40
***
CD3-PB CD3-PB 10 20

0 0
Acute Chronic Healthy Acute Chronic Healthy
g
50,000 800
R2 = 0.5251 R2 = 0.2343
40,000 P < 0.0001 P = 0.0122
TNF-α (pg ml–1)

600

IL-6 (pg ml–1)

f 30,000
50 *** 400
20,000
∆42PD1+Vδ2+ (%)

40
10,000 200
30
0 0
20 0 10 20 30 0 10 20 30

10
4,000 6,000
R2 = 0.4714 R2 = 0.1672
0 P = 0.0001 P = 0.0381
3,000

IFN-α (pg ml–1)

IL-1β (pg ml–1)

I-II III-V 4,000

Fiebig status
2,000
2,000
1,000

0 0
0 10 20 30 0 10 20 30
∆42PD1+Vδ2+ (%) ∆42PD1+Vδ2+ (%)

Fig. 1 | Δ42PD1+Vδ2 cells are found in early HIV-1 infection and correlate with immune activation. a, Representative plots for the detection of
Δ​42PD1-expressing CD3+Vδ​2+ cells in PBMCs of acute HIV-1-infected patients by flow cytometry based on isotype control. Numbers on scatter plots
represent percentages. b,c, Acute (n =​ 57, triangles) and chronic (n =​ 50, squares) HIV-1 patients and healthy controls (n =​ 29, circles) were assessed for
frequency of CD3+Vδ​2+ cells (b) and Δ​42PD1-expressing CD3+Vδ​2+ cells (c). d, Proportions of Vδ​1 and Vδ​2 cells among CD3+ cells analysed in HIV-1
acute and chronic patients and healthy controls (n =​ 12 for each group). e, Frequency of CD38+HLA-DR+ CD8 T cells for acute (n =​ 47), chronic (n =​ 17)
and healthy (n =​ 14) individuals. f, Δ​42PD1+Vδ​2+ cells among acute HIV-1 patients who are seronegative (Fiebig I–II, n =​ 17) and seropositive (Fiebig III–V,
n =​ 31) were assessed. Data represent mean ±​ s.e.m. and statistics were calculated based on the Mann–Whitney U-test; *P <​ 0.05, **P <​ 0.01, ***P <​ 0.001.
(a–f). g, Acute HIV-1 patient plasma cytokines were measured in a multiplex assay. Levels of cytokines were correlated with frequency of Δ​42PD1-
expressing CD3+Vδ​2+ cells by linear regression (95% confidence intervals shown as dotted lines) to calculate statistics. Each symbol represents the data
of one individual patient (n =​ 19); n, number of individual patient samples.

of PBMCs (Supplementary Fig.  7). We therefore sought to deter- day 3 p.i. Interestingly, only Vγ​9/Jγ​1.2 transcript was detected in
mine the specific effects of IL-12 and IL-15 on inducing Δ​42PD1 both mock- and HIV-1-induced Vδ​2+ cells (Supplementary Fig. 8a).
expression with or without neutralizing antibodies in PBMC cul- Furthermore, Vγ​9/Jγ​1.2 can be detected in both HIV-1 stimulated
tures infected with HIV-1NL4-3 or HIV-1Henan for 3 days. Compared Δ​42PD1− or Δ​42PD1+ CD3+Vδ​2+ cells (Supplementary Fig. 8b). To
to control, anti-IL-12 and/or anti-IL-15 abrogated the expression of verify if there are differences in TCR arrangement in the Vγ​9/Jγ​1.2
Δ​42PD1 on Vδ​2 cells (Fig. 2c). These data demonstrated that IL-12 transcripts, we performed sequence alignment analysis on TRGV9
or IL-15 cytokines arising from HIV-1 infection could activate Vδ​2 and TRGJP reference regions (Supplementary Fig.  8c,d). In the
cells to express Δ​42PD1. junction CDR3 region, we found one nucleotide change at position
The majority of Vδ​ 2 cells in peripheral blood express the 350 that led to an amino acid change from methionine (M) to lysine
TCR-Vγ​9 chain. Previous reports have shown that phosphoanti- (K) for Δ​42PD1− CD3+Vδ​2+ cells, which might differ in response
gen-responsive Vδ​2 cells employ the Jγ​P (or Jγ​1.2) gene segment to HIV-1. Nonetheless, our analysis shows that Δ​42PD1 may not be
from the junction region and these cells are depleted in HIV-1 related to phosphoantigen recognition for Vδ​2 cells.
patients29–31. To investigate if Δ​42PD1+Vδ​2 cells express TCR-Vγ​9/ The gut-homing property of Δ​42PD1+Vδ​2 cells has been shown
Jγ​1.2 during HIV-1 infection, we performed RT–PCR analysis on in a previous study where CCR9 and CD103 were upregulated
CD3+Vδ​2+ cells isolated from mock or HIV-1-treated PBMCs at during HIV-1 infection, and this might facilitate their migration to the

Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology 1391

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles NaTuRe MICRObIOLOgy

∆42PD1+ ∆42PD1+
a b
28.1
20
47.3

∆42PD1+Vδ2+ (%)
9.52 15 * *
* HIV–1NL4-3

∆42PD1-AF647
10 HIV–1Henan
*
** * *
Vδ2–PE

– –
∆42PD1 ∆42PD1 5
78.9 6.45 Mock

CD103-AF488
0
CD3-PB CD3-PB 0 1 2 3 4 5 6
0.331
Day p.i.

SSC
CD69-PE/Cy7 CCR9-PerCP/
Cy5.5
c d e
80 80 80 ***
40 ** ∆42PD1+
CCR9+CD103+ Vδ2+ (%)

Mock
** ** ** ∆42PD1–

CCR9+CD103+ (%)
∆42PD1+Vδ2+ (%)

30
HIV-1NL4-3 60 *** 60 60
* HIV-1Henan ** ***
20 40 *** 40 ** 40

10 ** *** 20 20 20
** **
* *
0 0 0 0
3 4 5 6 3 4 5 6 ∆42PD1+ ∆42PD1–
pe

5
-1

-1

-1
ty

IL

IL

IL

Day p.i. (HIV-1NL4-3) Day p.i. (HIV-1Henan)

Iso

ti-

ti-

ti-
An

An

an
2/
-1
IL
ti-
An

Fig. 2 | HIV-1 infection induces Δ42PD1 expression on Vδ2 cells. a, Flow cytometry plots for assessing the frequency of Δ​42PD1 expression on
CD3+Vδ​2+ cells from human PBMCs infected with X4-tropic HIV-1NL4-3 at day 3 p.i., representative of four independent experiments. Expression of
activation marker CD69 and gut-homing receptors CCR9/CD103 among Δ​42PD1+ and Δ​42PD1− subpopulations is shown. b, Data from mock (black
circles), HIV-1NL4-3 (white squares) or R5-tropic HIV-1Henan (grey triangles) infected PBMCs in a 6 day time-course experiment (four independent
experiments). c, PBMCs infected with mock, HIV-1NL4-3 or HIV-1Henan were examined at day 3 p.i. for expression of Δ​42PD1 on Vδ​2+ cells in the presence
of neutralizing antibodies against IL-12 and/or IL-15 (four independent experiments). d,e, Expression of CCR9/CD103 on Δ​42PD1+ and Δ​42PD1− CD3+Vδ​
2+ cells, shown in a time-course experiment (four independent experiments) (d) or acute cohorts (n =​ 57) (e). Data were acquired by flow cytometry and
represent mean ±​ s.e.m.. Statistics were calculated based on Student’s paired t-test (b–d) or Mann–Whitney U-test (e); *P <​ 0.05, **P <​ 0.01, ***P <​ 0.001.

intestines in patients14. We therefore examined this in a time-course propria (LP) and intraepithelial lymphocytes (IEL) of the small
experiment of PBMCs infected with either HIV-1NL4-3 or HIV-1Henan intestine in mice receiving HIV-Vδ​2 cells as compared to mock-Vδ​2
by flow cytometry (Fig. 2a). We found that CCR9 and CD103 expres- by flow cytometry among other tissues (Fig. 3b and Supplementary
sion is mainly found in the Δ​42PD1+ subpopulation of Vδ​2 cells Fig.  9b). In comparison, we also found that NSG-HuPBL mice
following either viral infection (Fig. 2d). In contrast, Δ​42PD1−Vδ​2 without adoptive transfer harboured minimal CD3+Vδ​2+
cells had significantly less coexpression of CCR9 and CD103. To cells (Supplementary Fig.  10a). To confirm that transferred Vδ​2
determine the biological relevance of CCR9/CD103 expression on cells migrate to the small intestines, we injected carboxyfluorescein
Δ​42PD1+Vδ​2 cells, we examined the two acute cohorts of HIV- succinimidyl ester (CFSE)-labelled Vδ​2 cells isolated from mock
infected patients and found that Δ​42PD1+Vδ​2 cells significantly and HIV-1 PBMCs and, by day 3 post-adoptive transfer, the highest
(~30%) harboured CCR9/CD103 coexpression compared with number of CFSE+ cells were found in the small intestines (detected
Δ​42PD1−Vδ​2 cells (Fig. 2e). by flow cytometry; Supplementary Fig. 10b). Using a more sensitive
and stable dye, we performed another set of experiment with Qdot
Δ42PD1+Vδ2 cells migrates to the small intestines in a human- nanocrystal-labelled Vδ​2 cells, which also showed migration of
ized mouse model. To determine the in vivo relevance of transferred cells to the small intestine (detected by flow cytometry
Δ​42PD1+Vδ​2 cells during HIV-1 acute infection, we used NOD. and confocal microscopy; Fig. 3e,f and Supplementary Fig. 10c).
Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice humanized with freshly iso- Small intestinal pathology was then examined in mice receiving
lated human PBMCs (NSG-HuPBL), as we recently described32. HIV-Vδ​2 cells (haematoxylin and eosin (H&E) staining, Fig.  3c).
Three weeks after transfusion of 1 ×​  107 PBMCs, immune cells were In contrast to the Mock-Vδ​2 control, which showed intact villi
reconstituted in the NSG mice with detectable human CD45+ cells structures, HIV-Vδ​2 mice demonstrated the apparent development
and subpopulations of CD3+, CD4+ and CD8+ T cells in peripheral of Gruenhagen’s space and vacuolization (black arrows, Fig.  3c)
blood (Supplementary Fig. 9a). Next, to determine if Δ​42PD1+Vδ​2 and progression to detachment of epithelium (asterisk), mucosal
cells are capable of homing to mucosal tissues in vivo, autologous ulceration and disintegration of the LP (red arrow)33–35 (Fig.  3c).
CD3+Vδ​2+ cells were sorted from mock (Mock-Vδ​2) or HIV-1NL4-3- Furthermore, fluorescent immunohistochemistry (IHC) stain-
infected (HIV-Vδ​2) PBMC cultures at day 3 p.i. and intravenously ing of tissue sections showed infiltration of Δ​42PD1+Vδ​2+ cells in
(i.v.) transfused into NSG-HuPBL mice (Fig. 3a). After 3 days, cell the LP (white arrow, Fig.  3d). We therefore concluded that puri-
preparations from different tissues were analysed. We found that a fied HIV-induced Δ​42PD1+Vδ​2+ cells preferentially migrated to
higher frequency of Δ​42PD1+Vδ​2+ cells was detected in both lamina the small intestines after adoptive transfer, leading to inflamma­tory

1392 Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NaTuRe MICRObIOLOgy Articles
a b 60
1 × 107 Mock-Vδ2
2 × 105 CD3+Vδ2+ (i.v.)

Vδ2+CD3+ (%)
PBMCs i.p. Endpoint HIV-Vδ2 *
40

Day –21 Day 0 Day 3 20 *

0
Mock-Vδ2 (n = 4) HIV-Vδ2 (n = 5)
NSG-HuPBL (n = 9)

en
Sp C

LN
SI LP)
LI L)
LI P)
L)
M

(I E

(I E
le

(L
(
PB

SI
c Mock-Vδ2 HIV-Vδ2 d Vδ2 Δ42PD1 Merge
*

Mock-Vδ2
HIV-Vδ2
e f
1,500 Mock-Vδ2 HIV-Vδ2 No transfer
*
/104 lymphocytes
No. Qdot+CD3+

1,000

500
Qdot
Hoechst
0
Mock-Vδ2 HIV-Vδ2

Fig. 3 | Preferential migration of HIV-induced CD3+Vδ2+ cells to the intestines in humanized mice. a, Schematic diagram of humanized mice experiment
using human PBMC reconstituted NSG mice (HuPBL). b, Flow cytometry data showing the detection of CD3+Vδ​2+ cells in different tissue compartments,
at 3 days post-adoptive transfer (i.v.) of FACSorted CD3+Vδ​2+ cells from PBMCs infected with mock (Mock-Vδ​2; n =​ 4) or HIV-1 (HIV-Vδ​2;
n =​ 5) into mice. LN, lymph nodes; LP, lamina propria; IEL, intraepithelial lymphocytes of small intestine (SI) and large intestine (LI); n, number of
individual mice per group. c, Microscopic analysis of small intestine tissue sections 3 days post-transfer, by H&E, representative of ten images taken
per mice. Intestinal pathology show Gruenhagen’s space and vacuolization (black arrows), epithelium detachment (asterisk) and mucosal ulceration
and disintegration of the LP (red arrow). Yellow scale bars, 20 μ​m. d, Fluorescent IHC images for the detection of cells expressing Vδ​2 and Δ​42PD1,
counterstained with Hoechst 33258, and co-localization of Vδ​2 and Δ​42PD1-positive cells (merged), are shown for small intestine tissue sections of
mice receiving HIV-Vδ​2 cells (enlarged inset) and absent in mock. Images are representative of ten images taken per sample. e, Qdot-labelled Vδ​2+ cells
purified from mock or HIV-1 infected PBMCs were i.v. transferred into NSG-HuPBL mice (n =​ 4 each) for 3 days before assessment by flow cytometry.
CD3+Qdot+ cell numbers in the small intestine homogenates are plotted. f, Representative confocal microscopy images showing Qdot+ cells (white
arrows) found in the villi of small intestines. At least ten images were taken per mouse small intestine tissue sample. Two mice without transfer served as
control. White scale bars in d and f, 20 μ​m. Data represent mean ±​ s.e.m. and statistics were calculated based on Mann–Whitney U-test (b,e); *P <​ 0.05.

pathology and tissue damage there. To determine the clinical rel-
evance of this phenomenon, we performed correlation analysis of
HIV-1 patient plasma of markers associated with intestinal muco-
sal damage, that is, soluble (s)CD14, fatty acid binding protein 2
(FABP2) and haptoglobin to the frequency of Δ​42PD1+Vδ​2+ cells
(Supplementary Fig. 11a). Negative and positive correlations were
found for sCD14 and FABP2, respectively. However, the only sig-
nificant difference for FABP2 was found between acute and chronic
patients (Supplementary Fig.  11b). These findings suggest that
Δ​42PD1+Vδ​2+ cells may infiltrate to the intestines and, in part, play
a role in causing damage to the mucosal barrier.
Δ42PD1 induction of cytokines from DCs is dependent on TLR4.
To determine the molecular mechanism of possible Δ​42PD1+Vδ​2+
-mediated production of pro-inflammatory cytokines, we sought to
identify the cognate receptor of Δ​42PD1 on DCs. Because extra-
cellular treatment of Δ​42PD1 mainly induces pro-inflammatory
cytokines in human PBMCs24, we suspect that its receptor is one of
the surface innate membrane activators, such as TLR2, TLR4 and
TLR6, although they have been shown previously only to respond
to endogenous intracellular or soluble proteins36. To test this
hypothesis, monocyte-derived dendritic cells (MoDCs) pre-treated
with blocking antibody against TLR2, TLR4 and TLR6 were used
to monitor cytokine signals after sΔ​42PD1fc protein or LPS treat-
ment. Purified proteins were assessed for endotoxin by the limulous
amoebocyte lysate (LAL) assay or oxidized phospholipid contami-
nants by measuring oxidized low-density lipoproteins (oxLDL), but
no positive results were found. By qRT–PCR analysis, TNF-α​, IL-6
and IL-1β​transcriptional levels were found to be reduced when
TLR4 was inhibited, but not TLR2 or TLR6 (Fig. 4a). Consistently,
MoDCs pretreated with the TLR4-specific inhibitor CLI-095 abro-
gated TNF-α​, IL-6 and IL-1β​release by LPS and sΔ​42PD1fc treat-
ment on MoDCs (Fig.  4b). Furthermore, when we used vectors
encoding short hairpin RNA (shRNA) against TLR4, TNF-α​ and
IL-6 production levels were significantly decreased compared to
control MoDCs following sΔ​42PD1fc protein treatment (Fig.  4c).
The effect of shRNA in downregulating TLR4 and CD14 proteins
was shown in MoDCs using western blotting and flow cytom-
etry (Supplementary Fig.  12a,b). In contrast, shRNA against
another component of the TLR4 complex, CD14, resulted in an

Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology 1393

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles NaTuRe MICRObIOLOgy

a tnfa il6 il1b d *

25 200 100 Isotype 1.0
Anti-TLR2 * * Media

Absorbance (620 nm)

0.8 LPS
Relative expression

20 150 80 Anti-TLR4
Anti-TLR6 sΔ42PD1fc
15 60 0.6
* 100
10 40 0.4
50
5 20 0.2

0 0 0 0.0
a

a
fc

fc
a
S

ly

4
trl
fc
D1

1
i

i
i
LP

LP

LP
PD
ed

LR

D1
ed

PD
ed

on

-C
2P
M

-T

-C
M
M

42

42

NA
lls
4

NA

NA
sΔ

Ce
sΔ

sΔ

iR

iR

iR
ps

ps

ps
b e
150 TLR4 *
* *
TLR4/MD2-wt

Normalized luciferase
8,000 4,000 DMSO
600
CLI-095 * TLR4/MD2-F126A
100 TLR4/MD2-C95Y
TNF-α (pg ml–1)

IL-1β (pg ml–1)

6,000 3,000
IL-6 (pg ml–1)

400
4,000 2,000
* *
* 50
200 **
2,000 1,000
ND ND ND ND
0 0 0 0
ia

ia

ia
c
fc

α
sΔ PS

sΔ PS

fc

ag
c
1

1
1f

1f

LP

F-
LP

PB
ed

PD
PD

PD
ed

ed

1fl
PD
L

PD

TN
M

M
42
42

42

42

42
sΔ

sΔ

sΔ
c f
**
*
** ** *
psiRNA-Ctrl 60 *
2,500 6,000 psiRNA-TLR4
*
Normalized luciferase
psiRNA-CD14
2,000
TNF-α (pg ml–1)

* 40
IL-6 (pg ml–1)

4,000
1,500
*
1,000 * 20
2,000
500
0 0 0
c
ia
ia

D1
g

Ab

D1

Ab

D1
S

1f
1f

PD
PD

Ne
LP

ed
ed

PD

LP

2P

2P

2P
iso

iso
M
M

42
42

Δ4

Δ4

Δ4
+

+
sΔ
sΔ

ti-

ti-
ia

5
-1
293A cells
ed

an

an
IL
M

2/
+

+
ia

5
-1

-1
ed

TLR4 IL

IL
M

TLR4/MD2-WT 2/
-1
IL

TLR4/MD2-F126A γδ T cells
TLR4/MD2-C95Y

Fig. 4 | Δ42PD1 functions via TLR4 for the induction of cytokine production. a,b, Human MoDCs were pre-treated with antibody blockers against
TLR2, TLR4, TLR6 or isotype antibody control (n =​ 4) (a), or TLR4-specific inhibitor CLI-095 or DMSO (n =​ 4) (b) before addition of LPS (100 ng ml–1)
or recombinant purified sΔ​42PD1fc protein (2 μ​g ml–1). MoDCs in media only serve as negative control. TNF-α​, IL-6 and IL-1β​mRNA expression at 6 h
post-treatment or supernatant cytokine were analysed by qRT–PCR and multiplex cytokine bead assay, respectively. ND, not detected. c, MoDCs were
transfected with control vector (Ctrl) or vector encoding shRNA against TLR4 (psiRNA-TLR4) or CD14 (psiRNA-CD14) and analysed for TNF-α​ and
IL-6 production following 24 h treatment with LPS or sΔ​42PD1fc (n =​ 3). d, HEKBlue-hTLR4 cell line was transfected with psiRNA-Ctrl, psiRNA-TLR4 or
psiRNA-CD14 before treatment with LPS or sΔ​42PD1fc. e,f, Transient co-transfection of 293A cells with NF-κ​B-luc, TLR4 and/or MD2-wt, MD2-F126A
or MD2-C95Y, was performed, followed by treatment with sΔ​42PD1fc, sΔ​42PD1flag, LPS or TNF-α​ (100 ng ml–1) (n =​ 3) (e); or co-cultured at 1:1 with cells
293A-Neg, -PD1, -Δ​42PD1; or media or IL-12/IL-15-stimulated γ​δ​T cells that were pre-treated with isotype or α​-Δ​42PD1 antibody (n =​ 3) (f). Data are
presented as mean ±​ s.e.m. of independent experiments (n) and statistics were calculated using Student’s paired t-test (a–f); *P <​ 0.05, **P <​ 0.01.

increased level of cytokines induced by sΔ​42PD1fc, suggesting that experiments was performed. First, HEKBlue-hTLR4 cells were
CD14 probably hinders Δ​ 42PD1–TLR4 signalling (Fig.  4c). In downregulated for CD14 using psiRNA-CD14 and subsequently
addition, when we performed experiments in the TLR4 signalling MD2 using two commercial siRNAs. The effect of downregula-
reporter cell line (HEKBlue-hTLR4), we found that LPS-induced tion was assessed by flow cytometry (Supplementary Fig. 13a). sΔ​
signalling was decreased when using shRNA against TLR4 or CD14. 42PD1fc and sΔ​42PD1flag proteins or LPS stimulation resulted in
However, sΔ​42PD1fc only induced reporter signalling when CD14 reporter activity. We used two different tagged proteins to show
was downregulated (Fig. 4d). These results demonstrate that Δ​42PD1 that the tag did not affect the function of sΔ​42PD1. CD14 down-
can induce cytokines through TLR4 in the absence of CD14. regulation resulted in a higher response by sΔ​42PD1, in contrast
To further illustrate that Δ​ 42PD1 can induce TLR4 down- to LPS, which was decreased (Supplementary Fig. 13b), consistent
stream signalling and to eliminate the possibility of endotoxin or with the above results. Knockdown of MD2 by siRNA resulted
oxidized phospholipids that require MD2, a series of reporter cell in a further reduction in LPS response, but had no significant

1394 Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NaTuRe MICRObIOLOgy Articles
a b
sΔ42PD1fc coated sPD1fc coated
10 10
KD = 6.82 µM
Time
8 0
TLR4/MD2 (µM)

Response unit

Response unit
6 TLR4/MD2 (µM) –10
0.49
44.1
4 –20 1.64
14.7
2 –30 4.9
4.9
1.64 14.7
0 –40 44.1
0.49
Time

sΔ42PD1fc protein

TLR4-His protein
IP: anti-rabbit Fc
TLR4 – + – + + – –
TLR4/MD2 – – + – – + +

Marker
LBP – – – + – + –
HMGB1 – – – – + – +

IB: anti-His
- 70 kDa

IB: anti-rabbit Fc - 70 kDa

d Pre-bleach → Post-bleach FRET 0.3 **

CFP channel

FRET efficiency
0.2
sΔ42PD1-CFP

0.1
YFP channel

130 0

sΔ42PD1-CFP/TLR4-YFP

sΔ42PD1-CFP mixed
110
Grey value

with TLR4-YFP
90

co-transfected
YFP channel
70
TLR4-YFP

50
0 100 200 300 400
Distance (pixels)

Fig. 5 | Direct interaction between Δ42PD1 and TLR4. a,b, SPR of flowing concentrations (0.49 to 44.1 μ​M) of TLR4/MD2 purified protein against
coated sΔ​42PD1fc (a) or sPD1fc (b). c, Co-IP of purified sΔ​42PD1fc and TLR4 or TLR4/MD2 proteins with or without LBP and HMGB1. Detection of
immunoprecipitate (IP) pulldown by anti-rabbit Fc agarose beads and immunoblotting (IB) by western blotting using anti-His or anti-rabbit Fc antibodies.
Dotted line: the same immunoblot from two exposure times to avoid over-developed TLR4 protein control signals. d, FRET analysis of co-transfecting sΔ​
42PD1-CFP and TLR4-YFP in 293T cells. Confocal microscopy was used to detect CFP and YFP channels, and FRET analysis was performed using PixFRET
(grey to white signals, blue background) with values measured against distance. FRET efficiency was calculated by subtracting background regions
from signal regions, and statistics were calculated by Student’s paired t-test of at least 20 fields from two independent experiments. A representative
fluorescence image is shown. Bar graph data represent mean ±​ s.e.m; **P <​ 0.01.

effect on sΔ​ 42PD1 proteins. We also performed co-culture
experiments using previously described stably expressing PD1 or
Δ​42PD1 293A cells24,25. 293A-Δ​42PD1 showed a significant signal-
ling response only when CD14 was downregulated, but was unaf-
fected by a reduction in MD2 (Supplementary Fig. 13c). Furthermore,
when we used IL-12/IL-15-stimulated γ​δ​T cells (with upregulated
Δ​42PD1) for co-culture, similar results were observed. Compared
to media-treated cells, IL-12/IL-15-γ​δ​T cells could induce reporter
activity on CD14-downregulated cells (Supplementary Fig.  13d),
but this was reduced when Δ​42PD1 was blocked using an in-house
anti-Δ​42PD1 antibody (clone CH101, Supplementary Fig.  14). To
verify whether Δ​42PD1 requires MD2 to trigger TLR4 signalling,
site-directed mutagenesis was performed to generate F126A and
C95Y mutations of the pEFBOS-humanMD2 expression plasmid
(provided by K. Miyake) (Supplementary Fig. 15a). These mutations
would lead to MD2 unresponsiveness to endotoxin and TLR4 sig-
nalling37–40. Following transient co-transfection of 293A cells with
expression plasmids encoding NF-κ​B-luc and TLR4, with or without
MD2-WT (wild-type), MD2-F126A or MD2-C95Y (Supplementary
Fig.  15b), cells were treated with sΔ​42PD1fc and sΔ​42PD1flag pro-
teins, LPS or TNF-α​. The results show that, unlike TNF-α​, LPS could
induce NF-κ​B activity with TLR4/MD2-WT transfected cells, but
not TLR4 alone, TLR4/MD2-F126A or TLR4/MD2-C95Y (Fig. 4e).
In contrast, sΔ​42PD1fc and sΔ​42PD1flag proteins could trigger NF-κ​
B activity with only TLR4 being expressed, despite the presence of
wild-type or MD2 mutants (Fig.  4e). Similar findings were found
for 293A-Δ​42PD1 or IL-12/IL-15-γ​δ​T cells co-cultures with these
reporter cells (Fig. 4f). Together, our data illustrate that Δ​42PD1 can
trigger TLR4 downstream signalling independent of MD2, which
eliminates the possibility of contaminants.
Δ42PD1 directly interacts with its cognate receptor TLR4.
To determine if direct interaction occurs between Δ​42PD1 and
TLR4, several experiments were performed. A surface plasmon

Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology 1395

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles
resonance (SPR) experiment was first carried out by flowing dif-
ferent concentrations of TLR4/MD2 to immobilized sΔ​42PD1fc
protein. The results in Fig.  5a show positive binding kinetics of
TLR4/MD2 to sΔ​42PD1fc and the equilibrium dissociation con-
stant (KD) value of the interaction is calculated to be 6.82 μ​M
(Fig.  5a). In contrast, TLR4/MD2 had no interaction with
sPD1fc proteins (Fig.  5b) and LPS showed no binding activity to
sΔ​42PD1fc (Supplementary Fig. 16a). Furthermore, we confirmed that
no binding exists between sΔ​42PD1fc and PD-L1 (Supplementary
Fig.  16b), while a strong interaction was observed for sPD1fc and
PD-L1 (Supplementary Fig.  16c). In another set of experiments
based on fluorescence binding, the interaction between Δ​42PD1
and TLR4/MD2 was confirmed with a KD (4 μ​M) similar to that
of the SPR analysis (Supplementary Fig.  16d). Also, no signifi-
cant interaction of Δ​42PD1 and PD-L1 was observed in this assay
(Supplementary Fig. 16e). Second, co-immunoprecipitation (co-IP)
experiments were conducted. Purified sΔ​42PD1fc protein was
mixed with TLR4-His or TLR4/MD2-His complex proteins in a 1:1
ratio. Because LPS binding protein (LBP) and high-mobility growth
box 1 (HMGB1) could interact with TLR441, they were added to the
mixture in equal amounts to determine if the interaction between
sΔ​42PD1fc-TLR4 was affected. Following western blotting, the anti-
rabbit Fc detected sΔ​42PD1fc and indicated successful pulldown
despite different combinations of proteins being mixed (Fig.  5c).
Anti-His detection demonstrated co-precipitated TLR4-His after
sΔ​42PD1fc was mixed with TLR4/MD2 or TLR4, confirming their
interaction. Interestingly, the presence of HMGB1, but not LBP,
decreased the interaction between sΔ​42PD1fc and TLR4 or TLR4/
MD2 (Fig.  5c), suggesting that HMGB1 might have outcompeted
sΔ​42PD1fc in its interaction with TLR4. Moreover, in a dose-depen-
dent experiment, decreasing the amount of sΔ​42PD1fc reduced the
detection of TLR4 in the co-IP experiment but appeared to saturate
at a ratio of more than 1:1 (Supplementary Fig. 17). Finally, to fur-
ther demonstrate that sΔ​42PD1fc interacts with TLR4, fluorescence
resonance energy transfer (FRET) experiment was performed by
co-transfecting plasmids encoding sΔ​42PD1−CFP and TLR4-YFP
into 293T cells. After one day, confocal microscopy was performed
to assess if FRET had occurred. By comparing pre- and post-
photobleaching of the donor TLR4-YFP fluorescence, an increase
in sΔ​42PD1−CFP fluorescence was observed in the YFP channel
and by software analysis using PixFRET (Fig.  5d, white regions).
In contrast, the negative control pair sΔ​42PD1-CFP and TLR9-
YFP showed no significant fluorescence transfer (Supplementary
Fig.  18). Overall, these experiments demonstrate that sΔ​42PD1fc
binds to TLR4 as its cognate cell surface receptor; this has not been
demonstrated previously.
HIV-induced Δ42PD1+Vδ2 cells with gut-homing capabilities
can crosstalk with DCs via TLR4 to induce robust cytokine pro-
duction. Because Δ​42PD1 interacts directly with the TLR4/MD2
receptor complex, we sought to determine if Vδ​2 cells engage this
Δ​42PD1–TLR4 pathway for cytokine induction. To test this hypoth-
esis, we used HIV-induced Δ​42PD1+Vδ​2+ cells to co-culture with
autologous MoDCs at a 1:1 ratio for 24 h before performing qRT–
PCR for cytokine expression. Vδ​2 cells from mock infection that
express minimal levels of Δ​42PD1 were also sorted and used for
comparison. Antibody blockers against TLR4 and Δ​42PD1 were
included to show the specificity of the Δ​42PD1–TLR4 pathway.
Collectively, there were significantly increased expressions of tnfa,
il1b, il6 and ifna after co-culture of MoDCs with HIV-induced
Δ​42PD1+Vδ​2 cells as compared to Mock-Vδ​2 (Fig.  6a). Blocking
TLR4 or Δ​42PD1 significantly reduced the level of cytokine induc-
tion during co-culture (Fig. 6a), demonstrating a role of cell surface
Δ​42PD1 as an inducer of cytokines in the context of HIV-1 infec-
tion via TLR4 that was previously unrecognized. Moreover, cell–cell
interaction was required, because transwell separating MoDCs and
1396 Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NaTuRe MICRObIOLOgy
HIV-induced Δ​42PD1+Vδ​2 cells abrogated cytokine induction. As
controls, MoDCs treated with sΔ​42PD1fc or LPS led to the expres-
sion of tnfa, il1b and il6. To confirm these findings, an enzyme-linked
immunosorbent assay (ELISA) was performed for IL-6 secretion in
the co-culture experiment between sorted HIV-Vδ​2 and autologous
MoDCs (Fig.  6b), and this correlated with the qRT–PCR results
(Fig.  6a). Also, blocking the Δ​42PD1–TLR4 pathway with TLR4-
specific inhibitor CLI-095 or antibodies significantly decreased the
level of IL-6 release following co-culture (Fig. 6b).
To determine the in vivo relevance of these findings, we inves-
tigated if Δ​42PD1+Vδ​2 cell transmigration could lead to increased
pro-inflammatory cytokines in the gut using our humanized mouse
model. We found that in the small intestines of NSG-HuPBL mice
with adoptively transferred HIV-Vδ​2, Δ​42PD1+ cells were found in
close proximity to TLR4+ cells in the LP of intestinal villi, which
appear to show pathological detachment of the epithelium (Fig. 6c).
Furthermore, Vδ​2+ cells appear to be located in cellular regions of
increased IL-6 expression, and this may suggest that their inter-
action with TLR4-expressing cells induced IL-6, mediating the
inflamed pathology observed in the gut tissue (Fig. 6d).
Inhibition of Δ42PD1–TLR4 alleviates gut pathology in human-
ized mice. Because the Δ​42PD1–TLR4 pathway exhibited by HIV-
induced Vδ​2 cells may confer pathological effects in the gut, we
next sought to determine if blocking this pathway would reduce
inflammation in the small intestine. In another set of NSG-HuPBL
mice experiments, anti-Δ​42PD1 (clone CH101), isotype antibody
or CLI-095 were administered i.p. 2 h before adoptive transfer of
mock- or HIV-induced Vδ​2 cells and by another injection 2 days
after transfer (Fig. 6e). H&E staining of the small intestines showed
that mice receiving Mock-Vδ​2 cells did not present with gut inflam-
mation, whereas transfer of HIV-Vδ​2 into mice treated with isotype
antibody resulted in significant pathological effects (Fig. 6f), consis-
tent with the above results (Fig. 3c). In contrast, mice that received
anti-Δ​42PD1 antibody had markedly reduced gut pathology, similar
to CLI-095-treated mice (Fig. 6f). Flow cytometry analysis showed
that the frequency of CD3+Vδ​2+ cells among human CD45+ cells was
significantly reduced in the small intestines of anti-Δ​42PD1 treated
mice compared to isotype, but not CLI-095 treatment (Fig.  6g).
Therefore, the Δ​42PD1–TLR4 pathway may be an important con-
tributor to gut inflammation following HIV-1 infection, and this
effect can be reduced using the anti-Δ​42PD1 antibody.
Discussion
In this study, to our knowledge, we describe for the first time that
an endogenously surface-expressed protein, Δ​42PD1, uses TLR4/
MD2 as an innate receptor to perform its immune stimulatory func-
tion (Fig. 5). One critical role of this innate Δ​42PD1–TLR4 pathway
was found to be regulating HIV-1 infection through the induction
of Δ​42PD1+Vδ​2 cells, which co-expressed the gut-homing recep-
tors CCR9 and CD103 (Fig. 2a,d). Once these HIV-1-induced Vδ​2
cells transmigrated into the small intestines, they promoted mucosal
inflammation and pathology in a humanized mouse model system by
triggering pro-inflammatory cytokine responses, probably through
TLR4-expressing DCs (Figs.  3 and 6c,d,f). In addition, a high fre-
quency of gut-homing Δ​42PD1+Vδ​2 cells was readily found among
acutely HIV-1-infected individuals, especially before seroconversion
(Fig. 1f), and was correlated positively with plasma cytokine levels
(Fig.  1g). The elevated cytokines may contribute to the polyclonal
activation of CD38+HLA-DR+ CD8 T cells in the early phase of HIV
infection and mucosal damage (Figs. 1e and 6f). Our findings there-
fore reveal a previously unrecognized Δ​42PD1–TLR4 pathway in
regulating immune activation in acute HIV-1 pathogenesis.
Δ​42PD1 is the first endogenous surface protein identified on
human cells that interacts with TLR4. Previous studies have only
identified endogenously secreted molecules that could signal through
NaTuRe MICRObIOLOgy Articles
a 40
tnfa
50
il6 b
30 *** 40 ***
20 30 * 4,000 *
5 *** 15 ***
4 *** * **
10 3,000
Relative gene expression
3

IL-6 (pg ml–1)

2 5
1 2,000
0 0

il1b il1b 1,000

80 80
70 ** 70 ** + ND
60 60 0
50 * 50 *
*** ***

IsoAb

Untreated
DMSO
CLI-095

sΔ42PD1fc
Anti-TLR4

γδ-T only

LPS
Anti-Δ42PD1
30 30
20 20
10 10
ND ND ND ND
0 0
Co-DC DC
IsoAb
Anti-Δ42PD1
Anti-TLR4
IsoAb
Anti-Δ42PD1
Anti-TLR4
Transwell
Untreated
sΔ42PD1fc
LPS

IsoAb
Anti-Δ42PD1
Anti-TLR4
IsoAb
Anti-Δ42PD1
Anti-TLR4
Transwell
Untreated
sΔ42PD1fc
LPS
Mock-Vδ2 HIV-Vδ2 DC Mock-Vδ2 HIV-Vδ2 DC
Co-DC Co-DC

c Δ42PD1 d Vδ2
e 1 × 107
TLR4 IL-6 Hoescht
PBMCs i.p. 1 × 105 CD3+Vδ2+ (i.v.) Endpoint

Day –21 Day 0 Day 3

–2 h Day 2

NSG-HuPBL (n = 1 5) Ab or CLI-095 (i.p.) Ab or CLI-095 (i.p.)

1. Mock-Vδ2 (n = 2) ( )
2. IsoAb + HIV-Vδ2 (n = 5) ( )
Groups -
3. Anti-Δ42PD1 + HIV-Vδ2 (n = 5) ( )
f Mock-Vδ2 IsoAb + HIV-Vδ2 4. CLI-095 + HIV-Vδ2 (n = 3) ( )

** g
4 20
Per cent CD3+Vδ2+ of CD45+

*
3 15
pathology grade
Small intestine

2 10

1 5

0 0
Ab

95
Ab

D1
2

2
1

95
PD
Vδ

Vδ

Iso

I-0
2P
Iso

I-0
2
k-

k-

Δ4
Δ4

CL
CL
oc

oc

ti-
ti-
M

An
An

Anti-D42PD1 CLI-095
+HIV-Vδ2 +HIV-Vδ2

Fig. 6 | HIV-induced Δ42PD1-expressing γδ T cells can induce robust cytokines from autologous DCs via Δ42PD1–TLR4. a, CD3+Vδ​2+ cells were sorted
from mock or HIV-1 infected PBMCs at day 3 p.i. and co-cultured 1:1 with autologous MoDCs (Co-DC) for 8 h (four independent replicates).
Anti-TLR4 or anti-Δ​42PD1, or isotype antibodies or transwell were used. qRT–PCR for the expression of tnfa, il6, il1b and ifna was performed on sorted
CD11c+ MoDCs. LPS or sΔ​42PD1fc protein served as control. ND, not detected. b, ELISA for IL-6 in co-culture experiments between HIV-1-induced sorted
Vδ​2+ cells and autologous MoDCs after 24 h with anti-TLR4 or anti-Δ​42PD1 antibodies, CLI-095 or respective controls (four independent replicates). c,d,
Fluorescent IHC staining of small intestine tissue sections from humanized mice that received sorted CD3+Vδ​2+ cells from mock (Mock-Vδ​2) or HIV-1-
infected PBMC cultures (HIV-Vδ​2) for 3 days. Immunostaining for human Δ​42PD1 and TLR4 (c) or IL-6 and Vδ​2 and counterstained for nucleus (d) was
performed to show close proximity between cells of interest (inset, white arrows). Images represent more than ten images taken per mouse from at least
four mice (per group). White scale bars, 20 μm. e, Schematic representation of NSG-HuPBL mice receiving Mock-Vδ​2 or HIV-Vδ​2 cells with anti-Δ​42PD1
Ab or CLI-095 administered (i.p.) 2 h prior and 2 days after cell transfer. (Iso)type antibody was used as control. f, Representative H&E images of small
intestine showing gut pathology from each group of mice, with grading shown as a graph. Images from at least 20 fields were taken and assessed. Yellow
scale bars, 20 μ​m. g, Flow cytometry detection of CD3+Vδ​2+ cells gated on human CD45+ cells. Graph data represent mean ±​ s.e.m. and statistics were
calculated based on Student’s paired t-test (a,b) or Mann-Whitney U-test (f,g); *P <​ 0.05, **P <​ 0.01, ***P <​ 0.001.

TLR4, including danger-associated molecular patterns (DAMPs) such which are produced as a result of cell death or pathological condi-
as HMGB1, and liver secretory glycoprotein Fetuin-A; but others have tions, we found that cell-surface-expressed Δ​42PD1 can drive a rapid
shown Ni2+ can genuinely trigger TLR441–44. Unlike the danger signals, inflammatory response via TLR4 on DCs (Fig. 4). Other studies have

Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology 1397

© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles
shown that γ​δ​T cells and DCs can crosstalk to achieve dual activa-
tion of both cell types45,46, which is consistent with our finding that
Δ​42PD1+Vδ​2 cells can act on DCs through TLR4 signalling to stimu-
late a pro-inflammatory response.
To present Δ​42PD1 as a genuine TLR4 agonist, three postu-
lates should be met, as proposed by Mancek-Keber and Jerala47 and
should address the possibility of contaminants such as endotoxin
and oxidized phospholipids in experiments. These postulates are as
follows: (1) binding to TLR4 or TLR4/MD2, (2) activation of the
receptor signalling by in situ produced agonist, and (3) structural
identification for molecular interactions with TLR4 or TLR4/MD2.
Because endotoxin and oxidized phospholipids engage MD2 to
achieve TLR4 signalling, in our reporter assays the downregulation
of MD2 or the use of the F126A and C95Y mutants demonstrates
that Δ​42PD1 could still function, which indicates the absence of
such contaminants from these purified proteins. Also, no signifi-
cant differences in downstream signalling between TLR4 or TLR4/
MD2 were found with Δ​42PD1 protein stimulation or Δ​42PD1-
expressing cells, in consistency with the co-IP results, suggesting
that Δ​42PD1 may interact with TLR4 at sites distinguished from
those identified for LPS48. Together with SPR analysis, interaction of
Δ​42PD1 with TLR4/MD2 was confirmed, and Δ​42PD1 is capable
of triggering TLR4 signalling. Furthermore, blocking of Δ​42PD1
in IL-12/IL-15-induced Δ​42PD1+γ​δ​ -T cell co-culture experi-
ments with reporter cell lines or DCs demonstrates that naturally
expressed Δ​42PD1–TLR4 does indeed play a role in the induction
of cellular responses. Determination of the crystal structure of Δ​
42PD1 is currently under way, but we have attempted to gain some
idea as to where the interaction between Δ​42PD1 and TLR4 may
occur by docking wild-type PD-1 (known structure) to TLR4/MD2
in silico (Supplementary Fig.  19). It appears that PD-1 does not
interact with MD2, and the 42-nucleotide deletion may result in a
different structure for Δ​42PD1 to bind to the central N-terminal
region of TLR4. Without the structure of the TLR4/MD2/Δ​42PD1
complex, we cannot, at this moment, definitively draw a conclusion
about how Δ​42PD1 interacts with TLR4/MD2. However, we dem-
onstrate that Δ​42PD1 is an agonist for TLR4.
The observation of upregulated Δ​42PD1 in response to acute
HIV-1 infection suggests that this protein can be an important reg-
ulator of early innate immune responses. Given the rapid-response
property of the γ​δ​T cells and their function as a bridge between
innate and adaptive immune responses, we speculate that Δ​42PD1
might play an essential role in modulating these functions. Indeed,
we found that the highest proportion of these cells arose at the earli-
est stages of acute HIV-1 infection and reduced when the patients
seroconverted, with a further decrease when the chronic state was
reached (Fig. 1c). However, HIV-1 infection induces Δ​42PD1+Vδ​2
cells, which were found to express the Vγ​9/Jγ​1.2 TCR, which may
suggest that these cells are depleted after early infection14,16. Our
data therefore uncover a previously unrecognized cellular mecha-
nism by which early immune activation following HIV-1 infec-
tion can occur through TLR4 signalling, leading to induction of
pro-inflammatory cytokines and polyclonal activation of CD8+ T
cells, as well as subsequent intestinal inflammation and damage.
To this end, the co-upregulation of gut-homing receptors CCR9/
CD103 on Δ​42PD1+Vδ​2 cells suggests their potential function in
modulating gut mucosal integrity. Previous studies show that the
frequency of Vδ​2 cells is significantly reduced both in peripheral
blood and in the gut despite HAART during HIV-1 infection when
compared with healthy people14,29,31. In our study, up to 20% of Vδ​2
cells expressed Δ​42PD1 in acute HIV-1-infected individuals, of
which a majority were positive for gut-homing receptors CCR9 and
CD103 (Fig. 2a,d). This is in line with previous studies that showed
high α​4β​7 and CCR5 expression on Vδ​2 cells following in vitro
HIV-1 infection of human PBMCs16. The translocation ability of
Δ​42PD1+Vδ​2 cells was further demonstrated in a humanized mouse
1398 Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NaTuRe MICRObIOLOgy
model (Fig. 3e,f). Once reaching the gut, they may engage in TLR4
signalling, which is abundantly expressed by DCs and other muco-
sal cell types, including macrophages, myofibroblasts, endothelial
cells and circulating polymorphonuclear neutrophils (PMNs) to
induce pro-inflammatory cytokines33. This may be the underlying
mechanism behind the increased IL-6 expression in the small intes-
tines in humanized mice and/or gut pathology (Fig.  6d,f). While
TLR4 signalling is an important regulator in intestinal homeosta-
sis49, the onset of CCR9+CD103+Δ​42PD1+Vδ​2 cells in natural acute
HIV-1 infection can stimulate an inflammatory response through
TLR4 that may contribute to gut CD4+ and Th17 T cell deple-
tion once migrated11,50–52. Furthermore, positive correlation of the
frequency of Δ​42PD1+Vδ​2 cells with FABP2 but not haptoglobin in
acute HIV-1 patients suggest that intestinal epithelial cell apoptosis
occurred rather than tight junction defects. The resulting muco-
sal integrity damage caused by Δ​42PD1–TLR4 and possibly sub-
sequent microbial translocation may account for systemic chronic
immune activation—a hallmark of HIV-1 pathogenesis. However,
other factors such as oxidative stress may contribute to acute HIV-
associated inflammatory responses (Supplementary Fig. 20), as has
been suggested for chronic HIV-153,54.
HIV/AIDS is mainly a sexually transmitted disease, for which
early innate immunity at mucosal surfaces remains incompletely
characterized. As mounting evidence points to a role of the acute
innate immune response in the control or development of HIV-1
disease, our findings may provide new insights into the understand-
ing of a new pathway that underlies acute HIV-1 infection. Vδ​2 cells
are characterized for other functions such as antigen presentation
and cytotoxicity1,55,56, whereas TLR4 has been implicated to affect
T cell stimulation and function36,57,58, and future work to define the
role of Δ​42PD1–TLR4 in the functional aspects of these cell types
may identify new immune mechanisms. Because Vδ​2 cells are
important players in other mucosal infections and disease mod-
els59,60, studies should examine the role of the Δ​42PD1–TLR4 path-
way in settings besides HIV/AIDS.
Methods
Study subjects. All study subjects gave informed consent, and procedures
were reviewed and approved by the Beijing You-An Hospital Research Ethics
Committee, The First Hospital of China Medical University Ethics Committee
and the Ethics Review Committee of Shenzhen Third People’s Hospital. A cohort
of members of the high-risk MSM (men who have sex with men) group were
enrolled in Beijing You-An Hospital, Beijing, China, and was screened every two
months for HIV-1 infection. A total of 45 acute HIV-1-infected individuals were
enrolled in this study at Beijing You-An Hospital, Capital Medical University,
Beijing, China. Another cohort of 12 acute HIV-1-infected individuals (MSM)
were enrolled from The First Hospital of China Medical University, Shenyang,
China. They were subjected to voluntary HIV-1 testing and counselling, given
necessary healthcare, and monitored regularly for HIV PCR testing at 6 week
to 3 month intervals. Apart from signed informed consent forms for the
collection of samples, all subjects received subsequent analyses and completed an
administered epidemiological questionnaire. Once identified as infected, patients
were followed up and blood samples were obtained in 1 or 2 week intervals.
All new infections were classified based on the detection of HIV-1-specific
RNA, antigen and antibody in plasma, according to the Fiebig system. Acute
HIV positivity was achieved when real-time PCR showed positive detection
of HIV RNA, with or without western blot detection of all three Gag, Env and
Pol antigens, no antibody against HIV was detected by ELISA, but with HIV
seroconversion occurring within 6 months of follow-up. A further 17 chronic
patients infected for more than 2 years were also recruited from Beijing You-An
Hospital and 33 from Shenzhen Third People’s Hospital. None of the acute or
chronic patients received antiretroviral treatment. The patient characteristics are
shown in Supplementary Table 1.
PBMC, MoDC and transfection. Fresh PBMCs were collected following gradient
centrifugation (Ficoll-Paque, GE Healthcare) of healthy human blood donor buffy
coats. Use of buffy coats received ethics approval from the Institutional Review
Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster
#UW13-476. MoDCs were generated by 5 or 7 day culture of freshly purified
CD14+ cells from healthy human PBMCs using microbeads (>​95% purity, Miltenyi
Biotec) in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS,
GIBCO), 2 mM l-glutamine (GIBCO), 1% streptomycin/penicillin (GIBCO),
NaTuRe MICRObIOLOgy
10 mM HEPES (GIBCO), 2 mM β​-mercaptoethanol (GIBCO), 20 ng ml–1 each of
recombinant human GM-CSF and IL-4 (Sigma-Aldrich). A 50% medium change
was carried out every 3 days. MoDCs were transfected with 1 μ​g of plasmids
encoding shRNA (InvivoGen) against TLR4 (psiRNA-TLR4) or CD14 (psiRNA-
CD14), or a control plasmid (psiRNA-Ctrl), using Lipofectamine 2000 (Invitrogen)
in Opti-MEM medium (Invitrogen) for 48 h with Zeocin selection before treatment
with stimulating agents. Success of downregulation was observed by western
blotting. The HEKBlue-TLR4 cell line was used according to the manufacturer’s
instructions (InvivoGen) and expression of human TLR4, CD14 and MD2 was
monitored by flow cytometry.
Antibodies and flow cytometry. The following antibodies were used for western
blotting: mouse anti-His (R&D Systems #MAB050), anti-mouse IgG HRP (GE
Healthcare #NA931V) and anti-rabbit IgG HRP (GE Healthcare #NA934V). For
flow cytometry, the following antibodies against human surface proteins were
used: anti-CD3 Pacific Blue (PB; clone UCHT1, BD Biosciences #558117),
anti-γ​δ​-TCR APC (clone B1, BD Biosciences #555718), anti-Vδ​2 PE (clone
B6, BD Biosciences #555739), anti-CD4 PerCP-Cy5.5 (clone OKT4, Biolegend
#317428), anti-CD8a PE/Cy7 (clone SK1, eBioscience #25-0087-42), anti-CD8
Alexa Fluor 700 (clone RPA-T8, BD Biosciences #561453) anti-CD11c APC (clone
3.9, eBioscience #17-0116-42), anti-CD19 PE (clone 1D3, eBioscience #12-0193-
82), anti-CD14 PE/Cy7 (eBioscience), anti-CD56 APC (clone HCD56, Biolegend
#318310), anti-CD69 APC/Cy7 (clone FN50, Biolegend #310914), anti-CD103
Alexa Fluor 488 (clone Ber-ACT8, Biolegend), anti-CCR9 PerCP/Cy5.5
(clone L053E8, Biolegend) and anti-HLA-DR PerCP/Cy5.5 (clone LN3,
eBioscience #25-9956-42). Relevant matching isotype controls were used.
Two mouse monoclonal antibodies (clones CH34 and CH101) against Δ​42PD1
were generated as described in ref. 25 (by using a Δ​42PD1 expression plasmid DNA
to prime BALB/c mice followed by a boost with soluble Δ​42PD1 protein). Briefly,
serum samples were collected at week 10 post-vaccination and analysed for anti-sΔ​
42PD1 antibodies by ELISA. Splenocytes from mice showing high antibody titre
were used to fuse with SP2/0-Ag14 myeloma cells. The hybridoma cells, which
produced Δ​42PD1-specific monoclonal antibody (no cross reaction with wild-
type PD-1), were purified twice using a limited-dilution subcloning technique.
The positive clones (CH34 and CH101) were confirmed by flow cytometry using
Δ​42PD1-expressing 293T cells. These antibodies are capable of distinguishing
between wild-type PD-1 and Δ​42PD1 by flow cytometry25 and the CH101
clone has a blocking effect against Δ​42PD1-induced signalling (Supplementary
Fig. 14). Anti-Δ​42PD1 antibodies were labelled using an Antibody Labeling
Kit (Invitrogen). Clone CH34 was labelled with Alexa Fluor (AF) 488 and clone
CH101 with Alexa Fluor 647.
Viruses. HIV-1NL4-3 (X4-tropic) and HIV-1Henan (R5-tropic) virus stocks and
pseudoviruses ADA and VSV-G were prepared by human PBMC expansion and
used for infection at 50 ng p24 (determined using ZeptoMatrix p24 ELISA kit) for
5 ×​  106 cells per ml RPMI 1640 +​ 10% FBS for 2 h, as previously described61. Human
cytomegalovirus strain AD169 was prepared as previously described62. Freshly
isolated PBMCs without PHA or IL-2 stimulation were used for infection.
RT–PCR and TCR-γ analysis. PBMCs were infected with HIV-1NL4-3 for 3 days
before sorted for Vδ​2+ cells using anti-phycoerythrin (PE) microbeads kit after
staining with anti-Vδ​2 PE antibody, or further stained with anti-Δ​42PD1 antibody
and FACSorted for Δ​42PD1+ and Δ​42PD1− cells using a BD AriaIII instrument.
RNA was then extracted with RNAiso Plus (Takara) and cDNA generated by
PrimeScript RT kit (Takara). PCR was performed using the following primers63,64
with PrimeStar HS DNA Polymerase (Takara): Vγ​I, 5′​-GAAGATCTAGACAGGC
CGACTGGGTCATCTGC-3′​; Vγ​II, 5′​-GAAGATCTAGACAGCCCGCCTG
GAATGTGTGG-3′​; Vγ​III, 5′​-AGCATGGGTAAGACAAGCAAA-3′​; Jγ​1.1/2.1,
5′​-GAAGATCTAGACTTACCAGGTGAAGTTACTATAAGC-3′​; Jγ​1.2, AAGAA
AACTTACCTGTAATGATAAGC-3′​; Jγ​1.3/2.3, 5′​-CTAGTCTAGACCGTA
TATGCACAAAGCCAGATC-3′​; hu-β​-actin forward, 5′​-GGATGATGATATC
GCCGCG-3′​, reverse, 5′​-GGATAGCAACGTACATGGCTGGG-3′​. PCR products
were electrophoresed on 1% (wt/vol) agarose gel, extracted using a QIAquick
gel extraction kit (Qiagen) or MEGA-spin(TM) Agarose Gel DNA Extraction
Kit (Intron) and sent for Sanger sequencing. The original gel photos are provided
in Supplementary Fig. 22. For TCR-Vγ​II/Jγ​1.2 sequence analysis, sequences
were aligned to the best-matched germline V and J gene (TRGV9*01:
accession no. X07205; and TRGJP*01: accession no. M12950) on the
IMGT/V-QUEST database65 (www.imgt.org).
Humanized mice experiments. NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ mice (NSG,
Jackson Laboratories) were bred and housed in the minimal disease area in
HKU Laboratory Animal Unit. All animal experiments were approved by
the Committee on the Use of Live Animals in Teaching and Research at the
Laboratory Animal Unit of the University of Hong Kong. Mice were randomly
chosen with no blinding, and gender was chosen between groups randomly. Four-
to six-week-old mice were reconstituted with 1 ×​  107 fresh healthy human PBMCs
by intraperitoneal (i.p.) injection in 0.5 ml RPMI. At day 10, blood was drawn
to assess the existence of human lymphocytes by red blood cell lysis (BD Pharm
Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles
Lyse) and flow cytometry. Three weeks after reconstitution, mice were injected i.v.
with 1 ×​  105 CD3+Vδ​2+ cells FACSorted from the same donor PBMCs following
3 days of in vitro HIV-1 infection. For tracking experiments, FACSorted CD3+Vδ​
2+ cells were labelled with CFSE using the CellTrace CFSE proliferation kit or
Qdot 655 nanocrystals according to the manufacturer’s instructions (Thermo
Fisher Scientific), before i.v. injection. For the in vivo blocking experiment,
anti-Δ​42PD1 (CH101 clone) antibody or TLR4 inhibitor CLI-095 (InvivoGen)
were administered i.p. at 2 mg kg–1 and 3 mg kg–1, respectively, 2 h before and
2 days after CD3+Vδ​2+ cell transfer. Three days after transfer, mice were killed
for blood and the following tissues collected for analysis: spleen, lymph nodes,
small intestine and large intestine. LP cells and IELs were isolated according to
a published protocol66. After formalin fixation and paraffin embedding, tissue
sections and H&E staining were performed at the Department of Pathology,
HKU LKS Faculty of Medicine, Queen Mary Hospital, Pokfulam, Hong Kong.
Gut pathology was graded in at least ten fields of vision per tissue sample slide
under light microscopy according to refs. 67,68, and the median was used for
analysis. Briefly, as shown in Fig. 3c, the small intestinal pathology was graded as
follows: Gruenhagen’s space and vacuolization, grade 1 (black arrows); epithelium
detachment, grade 2 (asterisk); mucosal ulceration, grade 3; disintegration of the
LP, grade 4 (red arrow). Intact epithelial villi are graded 0 (as seen in mock).
Fluorescent IHC. Tissue sections of small intestines from humanized mice
were immunostained as previously described32, with the primary antibodies
anti-human Vδ​2-TCR (clone 15D, Thermo Fisher Scientific), anti-Δ​42PD1 (clone
CH34 or CH101), anti-TLR4 (Abcam) and anti-IL-6 (clone M-19, Santa Cruz),
followed by the secondary antibodies anti-mouse IgG1-Alexa Fluor 488, mouse
IgG1-Alexa Fluor 568, anti-mouse IgG2a-Alexa Fluor 488 and anti-mouse IgG2b-
Alexa Fluor 568 (Invitrogen), counterstained using Hoechst 33258, and mounted
using fluorescence mounting medium (DAKO). Signals were captured using
a Carl Zeiss LSM700 confocal microscope (×​20 and ×​40 objective lenses at 1.3
aperture) and acquired using ZEN software at room temperature. ImageJ software
(http://imagej.nih.gov/ij/) was used for analysis and to merge images.
Purification of recombinant proteins. HEK-293F cells (Freestyle 293-F cells, Life
Technologies, mycoplasma negative) were transfected with pVAX-sPD1-Fc, pVAX-sΔ​
42PD1-Fc or pVAX-sΔ​42PD1-Flag plasmids using polyethylenimine (PEI; Sigma).
After 24 h, cells were 1:1 diluted with pre-warmed medium supplemented with
valproic acid (adjusted to a final concentration of 2.2 mM in the cell culture). Half
of the cell suspension was collected on day 5 post-transfection and an equivalent
volume of pre-warmed medium was supplemented to the culture. Four days after the
first collection, cells in the suspension were collected. The harvested cell suspension
was centrifuged at 3,220g for 10 min at 4 °C. Supernatant was then aspirated and
0.2 μ​m-filtered. Supernatant containing sPD1fc or sΔ​42PD1fc proteins was purified
by passing through a column with protein G agarose beads (Sigma), while that
containing sΔ​42PD1flag was passed through a column with anti-FLAG M2 affinity
gel (Sigma). The columns were appropriately washed and the purified protein was
then eluted, neutralized and concentrated using Amicon Ultra-4 Centrifugal Filter
Units (50K for sΔ​42PD1fc and 10K for sΔ​42PD1flag, Merck Millipore). An endotoxin
removal step with EndoTrap red (Hyglos) was then performed for the proteins
according to the manufacturer’s instructions. Protein was eluted and concentrated
using the above centrifugal filter units, its concentration determined by BCA protein
assay kit (Thermo Fisher Scientific) and then stored at –80 °C in multiple aliquots.
Analysis of cytokine expression. Endotoxin contamination was not detected
in all protein preparations tested using an E-TOXATE kit (<​0.03 endotoxin
units ml–1, Sigma-Aldrich) or Pierce LAL Chromogenic Endotoxin Quantitation
Kit (Thermo Fisher Scientific). LPS (100 ng ml–1) were also used for treatment.
MoDCs were treated with LPS or recombinant human sΔ​42PD1fc proteins as
described previously24. For inhibition experiments, MoDCs were pre-treated with
polyclonal blocking antibodies against TLR2, TLR4 and TLR6 (InvivoGen) for
30 min, or the TLR4-specific inhibitor CLI-095 (InvivoGen) for 1 h. After 6 h,
qRT–PCR was performed for the human genes tnfa, il6, il1b and ifna using specific
primers. HIV-1 infection of freshly unstimulated human PBMCs was performed
for 24 h before supernatants were collected for cytokine analysis. Cytokine release
was measured using the Human Th1/Th2/Th9/Th17/Th22 13plex, Human
Inflammation 20plex Ready-to-Use FlowCytomix Multiplex kit and analysed using
FlowCytomixPro software (version 3) or Human IL-6 Quantikine ELISA Kit (R&D
Systems) according to the manufacturer’s instructions (eBioscience).
Co-IP and western blotting. Recombinant proteins sΔ​42PD1fc and sPD1fc were
generated as described above. Purified sΔ​42PD1fc (0.5 μ​g) was mixed with 0.5 μ​g
of commercial carrier-free purified TLR4-His or TLR4/MD2-His proteins
(R&D Systems), or with equal amounts of LBP and HMGB1 (R&D Systems)
in 500 μ​l IP buffer (25 mM Tris, 150 mM NaCl, pH 7.2) with gentle rocking
overnight at 4 °C. Following addition of anti-rabbit IgG agarose beads for 2 h,
protein mixture was centrifuged at 5,000 r.p.m. for 10 min before being washed
in IP buffer and centrifuged again. Bead-bound proteins were eluted using IgG
elution buffer and neutralized with 1 M Tris (pH 9.0) buffer (10 μ​l per 100 μ​l
protein eluate). Proteins with loading dye were then boiled for 5 min and loaded
1399
Articles
onto 4–10% SDS–PAGE gel for electrophoresis and transferred to polyvinylidene
difluoride (PVDF) membrane (Millipore) using a Trans-Blot SD Semi-Dry
Transfer Cell (Bio-Rad). Membrane was blocked using PBS-T with 0.5% bovine
serum albumin (BSA) and 5% blotting-grade blocker (Bio-Rad) for 1 h, before
incubation with primary mouse anti-His (R&D Systems) or anti-Δ​42PD1 (clone
101) antibody, followed by anti-mouse HRP-conjugated secondary antibody (GE
Healthcare). Membrane blot was developed using Superbright Femto enhanced
chemiluminescence (ECL) substrate (Thermo Fisher Scientific) and signals were
captured by an Amersham Hyperfilm enhanced chemiluminescence (ECL) (GE
Healthcare) and Developer/Fixative solutions (Kodak) or scanned with a LICOR
C-DiGit scanner. Negative data for sPD1fc are not shown. The original photos are
provided in Supplementary Fig. 21.
DC co-culture experiments. MoDCs were generated as above (see section ‘PBMC,
MoDC and transfection’) and used at 5 days after culture with GM-CSF/IL-4 and
co-cultured with CD3+Vδ​2+ cells obtained from FACSorting of HIV-1-infected
PBMC cultures using the BDAriaIII instrument. After sorting, cells were used
immediately in RPMI +​10% FBS medium. Cells were pre-treated with blocking
antibodies against TLR4 (pAb, InvivoGen) or Δ​42PD1 (CH101) or isotype IgG1
(DAKO) for 30 min before co-culture at a 1:1 ratio in a U-bottomed 96-well culture
plate (Greiner Bio-One), or separated in a Transwell-96 culture plate (Corning).
Cytokine production was assessed by qRT–PCR and IL-6 in the supernatant by
ELISA (Biolegend) after 24 h of co-culture.
Generation of MD2 mutants. MD2-F126A or -C95Y mutation was introduced
into pEFBOS/humanMD2-FlagHis expression plasmid (provided by K. Miyake,
University of Toyko) by site-directed mutagenesis using the QuikChange II Site-
Directed Mutagenesis Kit (Agilent Technologies) according to the manufacturer’s
protocol. The primer pair for mutagenesis of human MD2-F126A mutant
comprised 5­′​-­CA­AC­AC­AT­TT­GT­AT­TT­TC­CC­TT­AG­GC­AA­TT­TT­AT­TC­CC­TT­
GAAGGAGAATGATATTGTT-3′​and 5′​-­AACAATATCATTCTCCTTCAAGG
GAATAAAATTGCCTAAGGGAAAATACAAATGTGTTG-3′​, both of which
were designed via the web-based QuikChange Primer Design Program available
online at http://www.genomics.agilent.com/primerDesignProgram.jsp.
The primer pair for mutagenesis of human MD2-C95Y mutant comprised
5′​-GCAAAGAAGTTATTTACCGAGGATCTGATGACGATTAC-3′​ and
5′​-GTAATCGTCATCAGATCCTCGGTAAATAACTTCTTTGC-3′​, as described
previously39. Introduction of the mutation was confirmed by DNA sequencing
using the forward primer of MD2 (5′​-GCATTTGTAAAGCTTTGGAGA-3′​) at
Beijing Genomics Institute (Hong Kong) (Supplementary Fig. 15a). Detection
of the expression of wild-type or mutant MD2 by flow cytometry in transfected
cells was performed using the anti-Flag antibody (clone M2, Sigma).
Reporter cell line experiments. HEKBlue-hTLR4 cells (InvivoGen) were obtained
commercially and tested by LPS as a positive control to authenticate the cell line.
HEKBlue-hTLR4 cells were first transfected with psiRNA-hCD14 (InvivoGen) sub-
cultured in Zeocin-selection medium, then transiently transfected with MD2 siRNAs
(ID s24262 and s24264, Silencer Select pre-designed siRNA; Thermo Fisher Scientific)
or Silencer Select negative control #1 siRNA (Thermo Fisher Scientific). After 48 h,
the effect of downregulation was determined by flow cytometry using anti-Flag (for
MD2) and anti-CD14 antibodies. The 293A cells were co-transfected with NF-kB
Luciferase reporter expression vector (Clontech, provided by K. H. Kok, University
of Hong Kong), pUNO1-hTLR4-HA3x (InvivoGen), pEFBOS/humanMD2 and its
F126A mutant. The 293A cells are mycoplasma negative. Expression of TLR4 and
MD2 was determined on 293A cells by flow cytometry. For co-culture experiments,
~2.5 ×​ 104 transfected HEKBlue-hTLR4 cells or ~5 ×​ 104 transiently transfected 293A
cells were co-cultured 1:1 with 293A cells (negative or stably transfected with PD1
and Δ​42PD1) or γ​δ​T cells (untreated or treated with IL-12 (4 ng ml–1) and IL-15
(20 ng ml–1) (Peprotech) for 5 days, then pre-treated with anti-Δ​42PD1 antibody
(clone CH101) or IgG1 isotype (DAKO)), or treated with sΔ​42PD1fc or sΔ​42PD1flag
purified protein (2 μ​g ml–1), LPS (100 ng ml–1) or TNF-α​ (10 ng ml–1). Results were
determined using the VICTOR3 1420 Multilabel Counter (PerkinElmer) after 12 h
incubation for SEAP reporter response (absorbance at 620 nm) from HEKBlue-
hTLR4 cells, or after 24 h for 293A cells for luciferase activity. For luciferase, cells
were washed with PBS and then lysed. The lysates were used in a Luciferase Assay
System (Promega) and CellTiter-Glo Luminescent Cell Viability Assay (Promega) to
normalize for cell number differences.
Plasma endotoxin, oxidized phospholipid, and biomarker ELISA. Healthy human
plasma or HIV-1 patient plasma was assessed for endotoxin level using a Pierce LAL
Chromogenic Endotoxin Quantification kit (Thermo Fisher Scientific). Plasma
was centrifuged at 16,000g for 5 min, then 5 μ​l was diluted 1:10 with endotoxin-
free water and incubated at 75 °C for 15 min, then assayed according to the
manufacturer’s instructions. All recombinant proteins used contained <​0.1 EU ml–1.
Acute and chronic HIV-1 patient plasma was assessed for the biomarkers soluble
CD14, FABP2 and haptoglobin using Quantikine ELISA kits (R&D Systems), as well
as oxidized phospholipid using the Human Oxidized LDL ELISA kit (oxLDL, Cell
Biolabs) according to the manufacturer’s instructions. Recombinant proteins showed
negative results for oxLDL (below the detection limit).
1400 Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NaTuRe MICRObIOLOgy
FRET. 293T cells were co-transfected with plasmids encoding sΔ​42PD1-CFP
(acceptor) and TLR4-YFP (donor) or TLR9-YFP (CFP and YFP encoding plasmids
were provided by K. H. Kok, University of Hong Kong). The 293T cells were
mycosplasma negative. Settings were assisted by single transfections. Photobleaching
was carried out using an LSM510 confocal microscope and detection of FRET in
the YFP channel of the acceptor with the following settings: bleach for 5 s, repeat 5 s,
20 iterations. FRET analysis was performed as previously described69, and regions
of interest were identified in multiple areas without cells and areas inside the cells
without photobleaching and from areas with photobleaching to calculate the FRET
efficiency. ImageJ and the PixFRET plug-in were used to analyse regions of FRET
and increase in pixel intensity70.
SPR. Recombinant purified sΔ​42PD1fc or sPD1fc protein was immobilized
on a CM5 chip by amine coupling for SPR in a Biacore X100 instrument (GE
Healthcare). Commercial TLR4/MD2-His complex or PD-L1 purified proteins
(R&D Systems) was used to flow over the chip at different concentrations. LPS
(Sigma) was also examined against coated sΔ​42PD1fc. Analysis was performed on
Biacore X100 evaluation software (GE Healthcare).
Fluorescence binding assay. Recombinant proteins of soluble domains of ∆​42PD1
were expressed in Escherichia coli and purified as described previously71. Proteins
were allowed to refold overnight and purified by size-exclusion chromatography.
Purified recombinant TLR4/MD2 and PD-L1 were purchased from R&D Systems
and conjugated with Alexa Fluor 594 succinimidyl ester (Thermo Fisher Scientific)
according to the manufacturer’s manual. The labelled protein was then purified
from unlabelled dyes using a spin column. The binding experiment was performed
by titrating increasing concentrations of soluble ∆​42PD1 with 10 nM labelled
TLR4/MD2 or PD-L1 protein in PBS buffer. The change in fluorescence was
measured and the binding isotherm was fitted with the Hill equation using
IgorPro (WaveMetrics).
Molecular docking of PD1 to TLR4/MD2. Protein–protein docking was
performed using the ClusPro web server with default parameters72. Briefly, the
structure of PD1 (5GGS) was docked with the TLR4/MD2 complex (3FXI) and
the docking model with the best cluster scores calculated by balanced coefficients
was selected. The model was visualized and analysed with Discovery Studio
visualizer (Accelrys).
Statistics. All statistical analyses were performed using a paired two-tailed
Student’s t-test to calculate P values for equal numbers between groups. P <​  0.05
was considered statistically significant. Data are presented as mean ±​ s.e.m. of at
least three independent experiments unless indicated in the figure legends. Linear
regression was performed for correlation analysis with data from acute HIV
patients with R2 and P values depicted in the graphs. Data comparisons between
acute HIV and chronic HIV patient samples, and healthy controls, or groups of
mice were analysed using the Mann–Whitney U-test.
Data availability. The data that support the findings of this study are available
from the corresponding author upon reasonable request.
Ethical approvals. Informed consent was obtained for all participants and
reviewed and approved by the Beijing You-An Hospital Research Ethics
Committee and Ethics Review Committee, Shenzhen Third People’s Hospital. All
animal experiments were approved by the Committee on the Use of Live Animals
in Teaching and Research at the Laboratory Animal Unit of the University of
Hong Kong. Use of healthy human PBMCs was approved by the Institutional
Review Board of the University of Hong Kong/Healthy Authority Hong Kong
West Cluster.
Received: 24 October 2016; Accepted: 5 July 2017;
Published online: 14 August 2017
References
1. Vantourout, P. & Hayday, A. Six-of-the-best: unique contributions
of γ​δ​T cells to immunology. Nat. Rev. Immunol. 13, 88–100 (2013).
2. McCarthy, N. E. et al. Proinflammatory Vδ​2+​T cells populate the human
intestinal mucosa and enhance IFN-γ​production by colonic α​β​T cells.
J. Immunol. 191, 2752–2763 (2013).
3. McCarthy, N. E. et al. Azathioprine therapy selectively ablates human Vδ​2(+​)
T cells in Crohn’s disease. J. Clin. Invest. 125, 3215–3225 (2015).
4. Brandes, M., Willimann, K. & Moser, B. Professional antigen-presentation
function by human γ​δ​T cells. Science 309, 264–268 (2005).
5. Bjarnason, I. et al. Intestinal inflammation, ileal structure and function in
HIV. AIDS 10, 1385–1391 (1996).
6. Batman, P. A. et al. Jejunal enteropathy associated with human
immunodeficiency virus infection: quantitative histology. J. Clin. Pathol. 42,
275–281 (1989).
NaTuRe MICRObIOLOgy
7. Heise, C. et al. Primary acute simian immunodeficiency virus infection of
intestinal lymphoid tissue is associated with gastrointestinal dysfunction.
J. Infect. Dis. 169, 1116–1120 (1994).
8. Kotler, D. P., Reka, S. & Clayton, F. Intestinal mucosal inflammation
associated with human immunodeficiency virus infection. Dig. Dis. Sci. 38,
1119–1127 (1993).
9. McGowan, I. et al. Increased HIV-1 mucosal replication is associated with
generalized mucosal cytokine activation. J. Acquir. Immune Defic. Syndr. 37,
1228–1236 (2004).
10. Olsson, J. et al. Human immunodeficiency virus type 1 infection is associated
with significant mucosal inflammation characterized by increased expression
of CCR5, CXCR4, and β​-chemokines. J. Infect. Dis. 182, 1625–1635 (2000).
11. Harouse, J. M. et al. Distinct pathogenic sequela in rhesus macaques infected
with CCR5 or CXCR4 utilizing SHIVs. Science 284, 816–819 (1999).
12. Autran, B. et al. T cell receptor γ​/δ​+ lymphocyte subsets during HIV
infection. Clin. Exp. Immunol. 75, 206–210 (1989).
13. Gan, Y. H., Pauza, C. D. & Malkovsky, M. γ​δ​T cells in rhesus monkeys
and their response to simian immunodeficiency virus (SIV) infection.
Clin. Exp. Immunol. 102, 251–255 (1995).
14. Poles, M. A. et al. Human immunodeficiency virus type 1 induces persistent
changes in mucosal and blood T cells despite suppressive therapy. J. Virol. 77,
10456–10467 (2003).
15. Glatzel, A. et al. Patterns of chemokine receptor expression on peripheral
blood γ​δ​T lymphocytes: strong expression of CCR5 is a selective feature of
Vδ​2/Vγ​9 γ​δ​T cells. J. Immunol. 168, 4920–4929 (2002).
16. Li, H. & Pauza, C. D. HIV envelope-mediated, CCR5/α​4β​7-dependent killing
of CD4-negative γ​δ​T cells which are lost during progression to AIDS. Blood
118, 5824–5831 (2011).
17. Hermier, F. et al. Decreased blood TcR γ​δ​+ lymphocytes in AIDS and
p24-antigenemic HIV-1-infected patients. Clin. Immunol. Immunopathol. 69,
248–250 (1993).
18. Kosub, D. A. et al. γ​δ​T-cell functional responses differ after pathogenic
human immunodeficiency virus and nonpathogenic simian
immunodeficiency virus infections. J. Virol. 82, 1155–1165 (2008).
19. Li, Z. et al. γ​δ​T cells are involved in acute HIV infection and associated with
AIDS progression. PLoS ONE 9, e106064 (2014).
20. Poccia, F. et al. Peripheral Vγ​9/Vδ​2 T cell deletion and anergy to nonpeptidic
mycobacterial antigens in asymptomatic HIV-1-infected persons. J. Immunol.
157, 449–461 (1996).
21. Wallace, M. et al. Functional γ​δ​T-lymphocyte defect associated with human
immunodeficiency virus infections. Mol. Med. 3, 60–71 (1997).
22. Cardone, M. et al. HIV-1-induced impairment of dendritic cell cross-talk
with γ​δ​T lymphocytes. J. Virol. 89, 4798–4808 (2015).
23. Cimini, E. et al. Primary and chronic HIV infection differently modulates
mucosal Vδ​1 and Vδ​2 T-cells differentiation profile and effector functions.
PLoS ONE 10, e0129771 (2015).
24. Zhou, J. et al. Potentiating functional antigen-specific CD8+ T cell immunity
by a novel PD1 isoform-based fusion DNA vaccine. Mol. Ther. 21,
1445–1455 (2013).
25. Cheng, L. et al. Monoclonal antibodies specific to human Δ​42PD1: a novel
immunoregulator potentially involved in HIV-1 and tumor pathogenesis.
MAbs 7, 620–629 (2015).
26. Stacey, A. R. et al. Induction of a striking systemic cytokine cascade prior to
peak viremia in acute human immunodeficiency virus type 1 infection, in
contrast to more modest and delayed responses in acute hepatitis B and C
virus infections. J. Virol. 83, 3719–3733 (2009).
27. Brenchley, J. M. et al. Microbial translocation is a cause of systemic
immune activation in chronic HIV infection. Nat. Med. 12,
1365–1371 (2006).
28. García, V. E. et al. IL-15 enhances the response of human γ​δ​T cells to
nonpetide microbial antigens. J. Immunol. 160, 4322–4329 (1998).
29. Enders, P. J. et al. HIV-mediated γ​δ​T cell depletion is specific for Vγ​2+ cells
expressing the Jγ​1.2 segment. AIDS Res. Hum. Retroviruses 19, 21–29 (2003).
30. Evans, P. S. et al. In vitro stimulation with a non-peptidic alkylphosphate
expands cells expressing Vγ​2-Jγ​1.2/Vδ​2 T-cell receptors. Immunology 104,
19–27 (2001).
31. Hebbeler, A. M. et al. Failure to restore the Vγ​2-Jγ​1.2 repertoire in
HIV-infected men receiving highly active antiretroviral therapy (HAART).
Clin. Immunol. 128, 349–357 (2008).
32. Wu, X. et al. Brain invasion by CD4+ T cells infected with a transmitted/
founder HIV-1 during acute stage in humanized mice. J. Neuroimmune
Pharmacol. 11, 572–583 (2016).
33. Abreu, M. T., Fukata, M. & Arditi, M. TLR signaling in the gut in health and
disease. J. Immunol. 174, 4453–4460 (2005).
34. Balazs, A. B. et al. Vectored immunoprophylaxis protects humanized mice
from mucosal HIV transmission. Nat. Med. 20, 296–300 (2014).
35. Im, E. et al. Elevated lipopolysaccharide in the colon evokes intestinal
inflammation, aggravated in immune modulator-impaired mice. Am. J.
Physiol. Gastrointest. Liver Physiol. 303, G490–G497 (2012).
Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles
36. Erridge, C. Endogenous ligands of TLR2 and TLR4: agonists or assistants?
J. Leukoc. Biol. 87, 989–999 (2010).
37. Teghanemt, A. et al. Novel roles in human MD-2 of phenylalanines 121
and 126 and tyrosine 131 in activation of Toll-like receptor 4 by endotoxin.
J. Biol. Chem. 283, 1257–1266 (2008).
38. Yu, L. et al. NMR studies of hexaacylated endotoxin bound to wild-type and
F126A mutant MD-2 and MD-2.TLR4 ectodomain complexes. J. Biol. Chem.
287, 16346–16355 (2012).
39. Schromm, A. B. et al. Molecular genetic analysis of an endotoxin
nonresponder mutant cell line: a point mutation in a conserved region
of MD-2 abolishes endotoxin-induced signaling. J. Exp. Med. 194,
79–88 (2001).
40. Gruber, A. et al. Structural model of MD-2 and functional role of its basic
amino acid clusters involved in cellular lipopolysaccharide recognition.
J. Biol. Chem. 279, 28475–28482 (2004).
41. Apetoh, L. et al. Toll-like receptor 4-dependent contribution of the immune
system to anticancer chemotherapy and radiotherapy. Nat. Med. 13,
1050–1059 (2007).
42. Andersson, U. & Tracey, K. J. HMGB1 is a therapeutic target for sterile
inflammation and infection. Annu. Rev. Immunol. 29, 139–162 (2011).
43. Pal, D. et al. Fetuin-A acts as an endogenous ligand of TLR4 to promote
lipid-induced insulin resistance. Nat. Med. 18, 1279–1285 (2012).
44. Schmidt, M. et al. Crucial role for human Toll-like receptor 4 in the
development of contact allergy to nickel. Nat. Immunol. 11, 814–819 (2010).
45. Conti, L. et al. Reciprocal activating interaction between dendritic cells and
pamidronate-stimulated γ​δ​T cells: role of CD86 and inflammatory cytokines.
J. Immunol. 174, 252–260 (2005).
46. Scotet, E. et al. Bridging innate and adaptive immunity through γ​δ​
T-dendritic cell crosstalk. Front. Biosci. 13, 6872–6885 (2008).
47. Mancek-Keber, M. & Jerala, R. Postulates for validating TLR4 agonists.
Eur. J. Immunol. 45, 356–370 (2015).
48. Park, B. S. et al. The structural basis of lipopolysaccharide recognition by the
TLR4-MD-2 complex. Nature 458, 1191–1195 (2009).
49. Rakoff-Nahoum, S. et al. Recognition of commensal microflora by Toll-like
receptors is required for intestinal homeostasis. Cell 118, 229–241 (2004).
50. Chen, Z. et al. CD4+ lymphocytopenia in acute infection of Asian macaques
by a vaginally transmissible subtype-C, CCR5-tropic simian/human
immunodeficiency virus (SHIV). J. Acquir. Immune Defic. Syndr. 30,
133–145 (2002).
51. Mattapallil, J. J. et al. Massive infection and loss of memory CD4+ T cells in
multiple tissues during acute SIV infection. Nature 434, 1093–1097 (2005).
52. Raffatellu, M. et al. Simian immunodeficiency virus-induced mucosal
interleukin-17 deficiency promotes Salmonella dissemination from the gut.
Nat. Med. 14, 421–428 (2008).
53. Mancek-Keber, M. et al. Toll-like receptor 4 senses oxidative stress mediated
by the oxidation of phospholipids in extracellular vesicles. Sci. Signal. 8,
ra60 (2015).
54. Zidar, D. A. et al. Oxidized LDL levels are increased in HIV infection and
may drive monocyte activation. J. Acquir. Immune Defic. Syndr. 69,
154–160 (2015).
55. Brandes, M. et al. Cross-presenting human γ​δ​T cells induce robust CD8+ α​β​
T cell responses. Proc. Natl Acad. Sci. USA 106, 2307–2312 (2009).
56. Poonia, B. et al. γ​δ​T cells are ADCC effectors in elite HIV controllers.
Retrovirology 7(Suppl 1), O7 (2010).
57. González-Navajas, J. M. et al. TLR4 signaling in effector CD4+ T cells
regulates TCR activation and experimental colitis in mice. J. Clin. Invest. 120,
570–581 (2010).
58. Jin, B. et al. The effects of TLR activation on T-cell development and
differentiation. Clin. Dev. Immunol. 2012, 836485 (2012).
59. Tu, W. et al. The aminobisphosphonate pamidronate controls influenza
pathogenesis by expanding a γ​δ​T cell population in humanized mice.
J. Exp. Med. 208, 1511–1522 (2011).
60. Xiang, Z. et al. Targeted activation of human Vγ​9Vδ​2-T cells controls
Epstein–Barr virus-induced B cell lymphoproliferative disease. Cancer Cell 26,
565–576 (2014).
61. Kang, Y. et al. CCR5 antagonist TD-0680 uses a novel mechanism for
enhanced potency against HIV-1 entry, cell-mediated infection, and a
resistant variant. J. Biol. Chem. 287, 16499–16509 (2012).
62. Cheung, A. K. L. et al. The role of the human cytomegalovirus UL111A gene in
down-regulating CD4+ T-cell recognition of latently infected cells: implications
for virus elimination during latency. Blood 114, 4128–4137 (2009).
63. Ghosh, J. K., Romanow, W. J. & Murre, C. Induction of a diverse T cell
receptor γ​/δ​repertoire by the helix-loop-helix proteins E2A and HEB in
nonlymphoid cells. J. Exp. Med. 193, 769–776 (2001).
64. van Dongen, J. J. et al. Design and standardization of PCR primers and
protocols for detection of clonal immunoglobulin and T-cell receptor
gene recombinations in suspect lymphoproliferations: report of the
BIOMED-2 Concerted Action BMH4-CT98-3936. Leukemia 17,
2257–2317 (2003).
1401
Articles
65. Lefranc, M. P. et al. IMGT®, the international ImMunoGeneTics information
system® 25 years on. Nucleic Acids Res. 43, D413–D422 (2015).
66. Weigmann, B. et al. Isolation and subsequent analysis of murine lamina propria
mononuclear cells from colonic tissue. Nat. Protoc. 2, 2307–2311 (2007).
67. Erben, U. et al. A guide to histomorphological evaluation of intestinal
inflammation in mouse models. Int. J. Clin. Exp. Pathol. 7,
4557–4576 (2014).
68. Feinman, R. et al. HIF-1 mediates pathogenic inflammatory responses to
intestinal ischemia–reperfusion injury. Am. J. Physiol. Gastrointest. Liver
Physiol. 299, G833–G843 (2010).
69. Karpova, T. & McNally, J. G. Detecting protein–protein interactions with
CFP-YFP FRET by acceptor photobleaching. Curr. Protoc. Cytom. 12,
12.7.1–12.7.11 (2007).
70. Feige, J. N. et al. PixFRET, an ImageJ plug-in for FRET calculation that can
accommodate variations in spectral bleed-throughs. Microsc. Res. Tech. 68,
51–58 (2005).
71. Nicoludis, J. M. et al. Structure and sequence analyses of clustered
protocadherins reveal antiparallel interactions that mediate homophilic
specificity. Structure 23, 2087–2098 (2015).
72. Kozakov, D. et al. The ClusPro web server for protein–protein docking. Nat.
Protoc. 12, 255–278 (2017).
Acknowledgements
The authors thank K. Miyake for providing the MD2 plasmid, K.H. Kok for CFP,
YFP and NF-κ​B-luciferase plasmids and C. Cheng-Mayer for critical discussions.
The authors thank L. Liu for technical advice with immunohistochemical staining.
The authors acknowledge the Faculty Core Facility of the LKS Faculty of Medicine,
HKU, for technical assistance with confocal microscopy. This work was supported
1402 Nature Microbiology | VOL 2 | OCTOBER 2017 | 1389–1402 | www.nature.com/naturemicrobiology
© 2017 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
NaTuRe MICRObIOLOgy
by research grants from the Hong Kong Research Grant Council (RGC: HKU5/
CRF/13G, RGC17103514, RGC17122915 and A-HKU709/14 to Z.C.); the Health and
Medical Research Fund (HMRF: 14130582 to Z.C., 15140372 to A.K.L.C.); the San-
Ming Project of Medicine in Shenzhen (to Z.C. and H.Wa.); the National Science and
Technology Major Project (2012ZX10001-009-001-001 to Z.C., 2012ZX1000-1006-
001-009 to H.S.) Beijing Key Laboratory of HIV/AIDS Research (BZ0089 to H.Wu)
and Beijing Municipal of Science and Technology Major Project (D161100000416003
to H.Wu) and the University Development Fund of the University of Hong Kong and
Li Ka Shing Faculty of Medicine Matching Fund to AIDS Institute.
Author contributions
A.K.L.C. and Z.C. designed experiments, analysed data and wrote the manuscript.
A.K.L.C., Y.H., H.-y.K., M.C., Y.M., X.W., K.-s.L., H.-k.K, T.C.K.L., J.Z. and B.K.L.
performed experiments. J.L. and L.C. generated the Δ​42PD1-specific antibodies. Q.P.,
X.L., M.A., H.Wa., H.S., B.Z. and H.Wu provided HIV patient samples. A.X. and K.-Y.Y.
provided critical comments and materials.
Competing interests
The authors declare no competing financial interests.
Additional information
Supplementary information is available for this paper at doi:10.1038/s41564-017-0006-5.
Reprints and permissions information is available at www.nature.com/reprints .
Correspondence and requests for materials should be addressed to Z.C.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
 nature research | life sciences reporting summary
Corresponding Author: Zhiwei Chen

Date: Jun 27, 2017

Life Sciences Reporting Summary

Nature Research wishes to improve the reproducibility of the work we publish. This form is published with all life science papers and is intended to
promote consistency and transparency in reporting. All life sciences submissions use this form; while some list items might not apply to an individual
manuscript, all fields must be completed for clarity.
For further information on the points included in this form, see Reporting Life Sciences Research. For further information on Nature Research policies,
including our data availability policy, see Authors & Referees and the Editorial Policy Checklist.

` Experimental design
1. Sample size
Describe how sample size was determined. At least 3 replicate experiments using individual donor cells were
performed to achieve Student t-test statistics. Acute patient cases are rare
and we were able to obtained 57 individual samples. With at least 16
samples comparing between acute and chronic patients, a minimum
statistical power of 80% can be achieved. Mann-Whitney test was
performed for the patient data. Linear regression was performed for
patient plasma data. As indicated in Statistics of the Materials and
Methods.
2. Data exclusions
Describe any data exclusions. The criteria for HIV patients categorized as acute patients is reported in
the Materials and Methods section, under Study Subjects, page 29. There
were no inclusion/exclusion criteria for animals.
3. Replication
Describe whether the experimental findings were reliably reproduced. At least 3 independent replicate experiments were performed. Statistical
analyses were done to illustrate significance.
4. Randomization
Describe how samples/organisms/participants were allocated into Animals chosen for different groups, sequence of adoptive transfer and
experimental groups. sacrificed were randomized.
5. Blinding
Describe whether the investigators were blinded to group allocation No blinding was performed.
during data collection and/or analysis.
Note: all studies involving animals and/or human research participants must disclose whether blinding and randomization were used.

May 2017

1
6. Statistical parameters

nature research | life sciences reporting summary

For all figures and tables that use statistical methods, confirm that the following items are present in relevant figure legends (or the Methods
section if additional space is needed).

n/a Confirmed

The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement (animals, litters, cultures, etc.)
A description of how samples were collected, noting whether measurements were taken from distinct samples or whether the same sample
was measured repeatedly.

A statement indicating how many times each experiment was replicated

The statistical test(s) used and whether they are one- or two-sided (note: only common tests should be described solely by name; more
complex techniques should be described in the Methods section)
A description of any assumptions or corrections, such as an adjustment for multiple comparisons

The test results (e.g. p values) given as exact values whenever possible and with confidence intervals noted

A summary of the descriptive statistics, including central tendency (e.g. median, mean) and variation (e.g. standard deviation, interquartile range)

Clearly defined error bars

See the web collection on statistics for biologists for further resources and guidance.

` Software
Policy information about availability of computer code
7. Software
Describe the software used to analyze the data in this study. Flowjo, ImageJ with relevant plug-ins, Graphpad Prism, Excel, Biacore X100
software, FlowCytomixPro, ZEN.
For all studies, we encourage code deposition in a community repository (e.g. GitHub). Authors must make computer code available to editors and reviewers upon
request. The Nature Methods guidance for providing algorithms and software for publication may be useful for any submission.

` Materials and reagents

Policy information about availability of materials
8. Materials availability
Indicate whether there are restrictions on availability of unique Materials commercially available are indicated, in-house antibodies are
materials or if these materials are only available for distribution by a used where appropriate.
for-profit company.
9. Antibodies
Describe the antibodies used and how they were validated for use in Materials and Methods section, page 31.
the system under study (i.e. assay and species).
10. Eukaryotic cell lines
a. State the source of each eukaryotic cell line used. Cell line sources are found in Materials and Methods. Materials and
methods, under Cells and transfection, page 30, for HEKBlue-hTLR4; page
31 and 38, for FreeStyle 293F cells page 38. HEK-293A page 39

b. Describe the method of cell line authentication used. Cell lines were authenticated using the relevant stimulation and is
mentioned in the Methods section.

c. Report whether the cell lines were tested for mycoplasma The cell lines were ordered directly from the company and used. Some lab
contamination. cells were mycoplasma previously.

d. If any of the cell lines used in the paper are listed in the database Two cell lines was used in this study: HEKBlue-hTLR4 from Invivogen
of commonly misidentified cell lines maintained by ICLAC, (Materials and methods, page 31 and 38), not listed in the misidentified
provide a scientific rationale for their use. cell line database; FreeStyle 293-F cells (Thermo Fischer, previously
Invitrogen) were used for protein production (page 35); HEK293T (from
ATCC) were used for FRET assay (page 40); 293A cells were used for
transfection.
May 2017

2
` Animals and human research participants

nature research | life sciences reporting summary

Policy information about studies involving animals; when reporting animal research, follow the ARRIVE guidelines
11. Description of research animals
Provide details on animals and/or animal-derived materials used in Materials and methods, section Humanized mice experiments, page 34 -
the study. Strain and age are reported but not the gender. Equal number of male/
female were randomized into each group where possible.

Policy information about studies involving human research participants

12. Description of human research participants
Describe the covariate-relevant population characteristics of the Materials and methods, Study subjects, page 29-30; Cells and transfection,
human research participants. page 30. Consent was obtained for each patient as stated on page 29-30.

May 2017

3
 nature research | flow cytometry reporting summary
Corresponding Author: Zhiwei Chen

Date: Jun 27, 2017

Flow Cytometry Reporting Summary

Form fields will expand as needed. Please do not leave fields blank.

` Data presentation
For all flow cytometry data, confirm that:
1. The axis labels state the marker and fluorochrome used (e.g. CD4-FITC).
2. 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).
3. All plots are contour plots with outliers or pseudocolor plots.
4. A numerical value for number of cells or percentage (with statistics) is provided.

` Methodological details
5. Describe the sample preparation. In vitro experiment PBMCs or subsequent purified using
microbeads, were obtained from healthy individuals,
immunostained with relevant antibodies and fixed with 2% PFA.
Patient PBMCs were resuscitated, immunostained, and fixed with
2% PFA.

6. Identify the instrument used for data collection. BD Aria III and BD FACSCalibur

7. Describe the software used to collect and analyze the flow Acquiring data using BD Diva, analysis performed with Flowjo.
cytometry data.

8. Describe the abundance of the relevant cell populations Purity was determined by relevant staining using flow cytometry,
within post-sort fractions. these are stated in the manuscript where appropriate.

9. Describe the gating strategy used. All samples are FSC-A and SSC-A gated, followed by FSC-A/FSC-H
gating to select singlet cells. Subsequent relevant gating was
conducted, shown in the main figures or supplementary.

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

June 2017