nature immunology

Letter https://doi.org/10.1038/s41590-023-01724-6

Long COVID manifests with T cell

dysregulation, inflammation and an
uncoordinated adaptive immune
response to SARS-CoV-2

Received: 9 February 2023 Kailin Yin1,2,9, Michael J. Peluso 3,9, Xiaoyu Luo1,2, Reuben Thomas1,
Min-Gyoung Shin1, Jason Neidleman1,2, Alicer Andrew 1,2, Kyrlia C. Young1,2,
Accepted: 29 November 2023
Tongcui Ma1,2, Rebecca Hoh3, Khamal Anglin3, Beatrice Huang 3,
Published online: 11 January 2024 Urania Argueta3, Monica Lopez3, Daisy Valdivieso 3, Kofi Asare3,
Tyler-Marie Deveau4, Sadie E. Munter4, Rania Ibrahim3, Ludger Ständker5,
Check for updates
Scott Lu6, Sarah A. Goldberg 6, Sulggi A. Lee 7, Kara L. Lynch8,
J. Daniel Kelly 6, Jeffrey N. Martin6, Jan Münch 5, Steven G. Deeks3,
Timothy J. Henrich 4 & Nadia R. Roan 1,2

Long COVID (LC) occurs after at least 10% of severe acute respiratory
syndrome coronavirus 2 (SARS-CoV-2) infections, yet its etiology
remains poorly understood. We used ‘omic” assays and serology to
deeply characterize the global and SARS-CoV-2-specific immunity in
the blood of individuals with clear LC and non-LC clinical trajectories,
8 months postinfection. We found that LC individuals exhibited systemic
inflammation and immune dysregulation. This was evidenced by global
differences in T cell subset distribution implying ongoing immune
responses, as well as by sex-specific perturbations in cytolytic subsets.
LC individuals displayed increased frequencies of CD4+ T cells poised to
migrate to inflamed tissues and exhausted SARS-CoV-2-specific CD8+ T cells,
higher levels of SARS-CoV-2 antibodies and a mis-coordination between
their SARS-CoV-2-specific T and B cell responses. Our analysis suggested an
improper crosstalk between the cellular and humoral adaptive immunity
in LC, which can lead to immune dysregulation, inflammation and clinical
symptoms associated with this debilitating condition.

Intense efforts are underway to determine the pathophysiology of of T cells in a well-matched set of LC and fully recovered (R) individuals
long COVID (LC), a set of conditions characterized by immune per- to identify unique immune features associated with LC that inform on
turbations1. T cells have important roles in severe acute respiratory the mechanistic underpinnings of this condition.
syndrome coronavirus 2 (SARS-CoV-2) immunity and pathogenesis2–6, We leveraged a well-characterized cohort (Long-term Impact of
yet relatively little is known about their role in LC. Here we used CyTOF, Infection with Novel Coronavirus (LIINC)7; Supplementary Tables 1–3)
serology, RNA sequencing (RNA-seq), single‐cell RNA-seq (scRNA-seq) to analyze the blood from 27 LC and 16 R individuals, obtained
and plasma proteomics to obtain a deep phenotypic characterization 8 months postinfection (Fig. 1a) before any SARS-CoV-2 vaccination

A full list of affiliations appears at the end of the paper. e-mail: [email protected]; [email protected]

Nature Immunology | Volume 25 | February 2024 | 218–225 218

Letter https://doi.org/10.1038/s41590-023-01724-6

or reinfection. LC individuals, who consistently exhibited LC symptoms compared to R males, but not in females (Extended Data Fig. 7b).
such as fatigue, ‘brain fog’ and sleep disturbance over 8 months, were Cluster A1 was composed of CD45ROloCD45RAhiCD4+ TN cells and
63% female and included 26% previously hospitalized for COVID-19 expressed low levels of activation markers (HLA-DR and Ox40) and
(Extended Data Fig. 1a–c and Supplementary Tables 1–3). Comorbidi- inflammatory tissue-homing receptors (CD29 and CXCR4), as well
ties such as hypertension were more common in LC individuals (6/27 for as high levels of lymph node homing receptors (CD62L and CCR7;
LC and 1/16 for R), who also had higher body mass index (BMI; Extended Extended Data Fig. 7c). Cluster A4 was composed of terminally dif-
Data Fig. 1d,e). A CyTOF panel designed to interrogate the differentia- ferentiated CD45ROhiCD27loCD57hi CD4+ TEM cells and expressed high
tion and/or activation states, effector functions and homing properties levels of receptors associated with homing to inflamed tissues (CD29,
of T cells (Extended Data Fig. 1f and Supplementary Table 4) was applied CXCR4 and CCR5) but not to lymph nodes (CD62L and CCR7). They
to cryopreserved blood at baseline (post-thaw) or following stimula- also had high expression of cytolytic markers perforin and granzyme B
tion with SARS-CoV-2 spike and T-scan peptides (Methods) to identify (Extended Data Fig. 7d). Among CD8+ T cells, cluster B1 was significantly
SARS-CoV-2-specific T cells through intracellular cytokine staining. underrepresented in LC females, while cluster B2 was significantly
Both baseline and poststimulation datasets were gated on CD3+ overrepresented in LC females, compared to their R female counter-
events to identify T cells (Extended Data Fig. 1g,h), which were assessed parts, with no differences observed in males (Extended Data Fig. 7f).
for the expression of a panel of effector molecules, consisting of the Cluster B1 comprised CD8+ T cells expressing markers of cluster A1
cytokines interferon-γ (IFN-γ), tumor necrosis factor (TNF), interleukin (CD45ROloCD45RAhiHLA-DRloOx40loCD29loCXCR4loCD62LhiCCR7hi),
(IL)-2, IL-4, IL-6, IL-17 and CCL4, and the cytolytic markers granzyme B whereas cluster B2 comprised CD8+ T cells expressing markers of
and perforin (Extended Data Fig. 2a,b). Based on criteria comparing cluster A4 (CD27loCD57hiCD29hiCXCR4hiCCR5hiCD62LloCCR7lo). These
stimulated versus baseline samples (Methods), IFN-γ, TNF and/or observations suggested that females with LC had relatively low fre-
IL-2 positivity identified SARS-CoV-2-specific CD4+ T cells, whereas quencies of resting CD4+ and CD8+ TN cells, which expressed low levels
IFN-γ, TNF and/or CCL4 positivity identified SARS-CoV-2-specific of inflammatory tissue-homing receptors, and high frequencies of
CD8+ T cells (Fig. 1b,c and Extended Data Fig. 2a,b). Using Boolean terminally differentiated CD4+ and CD8+ TEM cells, which expressed
gating, we did not find significant differences between the frequen- inflammatory tissue-homing receptors and cytolytic markers.
cies of total SARS-CoV-2-specific CD4+ or CD8+ T cells (Fig. 1d), or The t-distributed stochastic neighbor embedding (t-SNE) visu-
those producing individual effector cytokines IFN-γ, TNF, IL-2 or CCL4 alization of SARS-CoV-2-specific CD4+ T cells indicated that those
(Extended Data Fig. 2c,d) between LC and R individuals. Furthermore, from LC and R individuals tended to concentrate in different areas
the distribution of polyfunctional (producing at least two cytokines) (Fig. 2a). The tissue-homing receptors CXCR4, CXCR5 and CCR6
SARS-CoV-2-specific CD4+ and CD8+ T cells was similar between LC and were expressed higher on SARS-CoV-2-specific CD4+ T cells from
R individuals (Fig. 1e, f). However, SARS-CoV-2-specific IFN-γ+TNF+IL- LC as compared to R individuals (Fig. 2b). Manual gating showed
2+CD4+ T cells and SARS-CoV-2-specific IFN-γ+TNF+CCL4+CD8+ T cells that the percentages of SARS-CoV-2-specific CXCR4+CXCR5+CD4+
were more abundant, without reaching statistical significance, in R T cells and CXCR5+CCR6+CD4+ T cells were significantly increased,
individuals (Fig. 1e,f). IL-6 expression in CD4+ T cells was induced exclu- and CXCR4+CCR6+CD4+ T cells showed a trend toward higher per-
sively in those with LC, albeit only in a small subset (14%; Extended centages, in LC compared to R individuals (Fig. 2c). Higher percent-
Data Fig. 2e,f). ages of total CXCR4+CXCR5+CD4+ T cells and CXCR5+CCR6+CD4+
CD45RA+CD45RO−CCR7+CD95− naïve T (TN) cells, CD45RA+CD45 T cells were found in LC compared to R as well (Fig. 2d). Flow
RO−CCR7+CD95+ stem cell memory T cells (TSCM) cells, CD45RA−CD45 cytometric analysis of the same LC and R specimens found statis-
RO+CCR7+CD27+ central memory T cells (TCM) cells, CD45RA−CD45RO tically significant elevated frequencies of CXCR4+CXCR5+CD4+,
+
CCR7−CD27− effector memory T (TEM) cells, CD45RA−CD45RO+CCR7 CXCR5+CCR6+CD4+ and CXCR4+CCR6+CD4+ T cells in LC compared
−
CD27+ transitional memory T (TTM) cells and CD45RA+CD45RO−CCR7− to R (Extended Data Fig. 8a–c). Expression of CXCR5 is common
effector memory RA T (TEMRA) cells were identified in both CD4+ and among the CXCR4+CXCR5+CD4+ T cell, CXCR5+CCR6+CD4+ T cell and
CD8+ T cell compartments through manual gating (Extended Data pTFH cell subsets, and we observed significant positive associations
Fig. 1i,j). In addition, CD45RA−CD45RO+CD127−CD25+ T regulatory (Treg) between the percentages of pTFH cells and other CXCR5+CD4+ T cells,
cells and CD45RA−CD45RO+PD1+CXCR5+ peripheral T follicular helper particularly in the LC group (Fig. 2e,f).
(pTFH) cells were identified in the CD4+ T cell compartment, and we SARS-CoV-2-specific CD8+ T cells were also globally different
additionally established a more stringent CD45RA-CD45RO+PD1hiCXCR between LC and R (Fig. 3a), and those from the individuals with LC
5hi TFH cell gate (Extended Data Fig. 1i). Total CD4+ TCM, pTFH, TFH and Treg preferentially expressed the checkpoint markers PD1 and CTLA4, but
cell subsets were more frequent in LC compared to R individuals with not TIGIT (Fig. 3b). Consistently, SARS-CoV-2-specific PD1+CTLA4+CD8+
no difference between LC and R in the other total CD4+ T cell subsets T cells were significantly elevated in LC compared to R individuals, while
analyzed (Fig. 1g), while none of these subsets were significantly dif- SARS-CoV-2-specific TIGIT+CTLA4+CD8+ or PD1+TIGIT+CD8+ T cells were
ferent between LC and R when examining SARS-CoV-2-specific CD4+ not (Fig. 3c). However, the frequencies of total PD1+CTLA4+CD8+ T cells
T cells (Fig. 1g,h). All analyzed subsets of total or SARS-CoV-2-specific were similar in the LC and R groups (Fig. 3d and Extended Data Fig. 8d).
CD8+ T cells were statistically similar between LC and R individuals Serological analysis indicated significantly higher (2.3×) total
(Extended Data Fig. 3). receptor binding domain (RBD)-specific antibody titers in LC as com-
Analysis of expression levels of all CyTOF markers in total or pared to R individuals (Fig. 4a). LC individuals with the highest fre-
SARS-CoV-2-specific CD4+ or CD8+ T cells found that no markers were quencies of SARS-CoV-2-specific PD1+CTLA4+CD8+ T cells had near
significantly differentially expressed between LC and R individuals undetectable antibody levels (Fig. 4b). LC individuals with the highest
(Extended Data Figs. 4 and 5). We found no significant differences in frequencies of SARS-CoV-2-specific PD1+CTLA4+CD8+ T cells had the
the percentages of CD4+ or CD8+ T cells expressing the acute activation lowest frequencies of SARS-CoV-2-specific CD4+ Treg cells, and the fre-
markers CD38, HLA-DR and/or Ki67 in LC compared to R individuals quencies of these two subsets of cells negatively correlated in LC, but
(Extended Data Fig. 6). Clustering analyses (Methods) revealed CD4+ not R individuals (Fig. 4b). A significant positive correlation between
T cells fell into six clusters (A1–A6) and CD8+ T cells into five clusters RBD-specific titers and total SARS-CoV-2-specific total CD4+ and CD8+
(B1–B5) clusters that did not differ significantly between LC and R T cell frequencies was detected in R but not LC individuals (Fig. 4c).
individuals (Extended Data Fig. 7a,e). However, cluster A1 was sig- The frequencies of SARS-CoV-2-specific pTFH cells also correlated
nificantly underrepresented in LC compared to R females, but not positively with RBD-specific antibody titers in R but not LC individuals
in males, while cluster A4 was significantly underrepresented in LC (Fig. 4c), suggesting a mis-coordinated humoral and cell-mediated

Nature Immunology | Volume 25 | February 2024 | 218–225 219

Letter https://doi.org/10.1038/s41590-023-01724-6

a T2
PBMCs
SARS-CoV-2 LC Sera
symptom onset diagnosis (WHO) T1 Plasma Vaccination? Reinfection?

LC, n = 27

Symptom load
R, n = 16

Time
Month 3 Month 4 Month 8

b c d 1.5 R
0 0 0 0 0 0 NS LC

Total CD4+ T
1.0

cells (%)
No No
0.5
peptides peptides

10
4
0.05 0.50 0.06 10
4
0.082 0.045 0.024 1.5
+SARS- 3
+SARS- NS

Total CD8+ T
3
10 10

CoV-2 CoV-2 1.0

cells (%)
2 2
10 10

peptides 1
peptides 1
10 10

CCL4
0.5
IFN-γ

IFN-γ

TNF
TNF

IL-2

0 0

0
1 2 3 4
0
1
10 10
2
10
3 4
10 0 10 10 10 10

CD4 CD4 CD4 CD8 CD8 CD8

e g TN cells TSCM cells TCM cells TEM cells R

IFN-γ+
80
Total CD4+ T cells (%)
0.8 40 60

Total CD4 T cells (%)

Total CD4+ T cells (%)

LC
Total CD4+ T cells (%)

IL-2+ NS
NS NS **
TNF+ 60 0.6 30 40

40 0.4 20

+
20
20 0.2 10
0
LC R 0 0 0

TTM cells TEMRA cells pTFH cells TFH cells Treg cells
IFN-γ+IL-2+TNF+CD4+ T cells
20 50 2.0 0.05 10
*
Total CD4+ T cells (%)

Total CD4 T cells (%)

Total CD4 T cells (%)

Total CD4+ T cells (%)

0.066
Total CD4+ T cells (%)

IFN-γ+IL-2+TNF–CD4+ T cells NS NS
40 * 0.04 8
IFN-γ+IL-2–TNF+CD4+ T cells 15 1.5
30 0.03 6
IFN-γ+IL-2–TNF–CD4+ T cells
10 1.0
20
+

+
– + + +
IFN-γ IL-2 TNF CD4 T cells 0.02 4
5 10 0.5
IFN-γ–IL-2+TNF–CD4+ T cells 0.01 2
0
IFN-γ–IL-2–TNF+CD4+ T cells 0 0 0 0

f h TN cells TSCM cells TCM cells TEM cells R

IFN-γ+ 60 3 80 25 NS LC
SARS-CoV-2-specific

SARS-CoV-2-specific

NS
SARS-CoV-2-specific

SARS-CoV-2-specific
CD4+ T cells (%)

CD4+ T cells (%)

CCL4+
CD4+ T cells (%)

CD4+ T cells (%)

20
40 2 60
TNF+ NS NS
15
40
20 1 10
20
0 5
0
0 0
LC R
+ + + +
TTM cells TEMRA cells pTFH cells TFH cells Treg cells
IFN-γ CCL4 TNF CD8 T cells
20
SARS-CoV-2-specific

SARS-CoV-2-specific

SARS-CoV-2-specific

60 30 30 30
SARS-CoV-2-specific

NS NS NS
SARS-CoV-2-specific

+ + – +
IFN-γ CCL4 TNF CD8 T cells NS NS
CD4+ T cells (%)

CD4+ T cells (%)

CD4+ T cells (%)

CD4+ T cells (%)

CD4+ T cells (%)

IFN-γ+CCL4–TNF+CD8+ T cells 25 15
40 20 20
+ – – +
IFN-γ CCL4 TNF CD8 T cells 20 10
IFN-γ–CCL4+TNF+CD8+ T cells 10 10 0.4
20
– + – + 0.2 5
IFN-γ CCL4 TNF CD8 T cells
0 0 0
IFN-γ–CCL4–TNF+CD8+ T cells 0 0

Fig. 1 | CD4+ T cell phenotypes are perturbed in individuals with LC. cells co-express at least two of the cytokines IFN-γ, IL-2 and TNF (e) or IFN-γ, IL-2
a, Strategy of biospecimen selection in individuals who resolved symptoms and CCL4 (f). g, Frequencies of CD45RA+CD45RO−CCR7+CD95− TN cells, CD45R
(R, n = 16) or who continuously experienced symptoms at month 4 (T1) and A+CD45RO−CCR7+CD95+ TSCM cells, CD45RA−CD45RO+CCR7+CD27+ TCM cells, CD
month 8 (T2) postinitial SARS-CoV-2 infection (LC, n = 27). The WHO definition 45RA−CD45RO+CCR7−CD27− TEM cells, CD45RA−CD45RO+CCR7−CD27+ TTM cells,
for LC is persistent symptoms for 3 months or more after infection14. All analyzed CD45RA+CD45RO−CCR7− TEMRA cells, CD45RA−CD45RO+PD1+CXCR5+ peripheral
PBMCs, sera and plasma were from 8 months postinfection, a timepoint when pTFH cells, CD45RA−CD45RO+PD1highCXCR5high TFH cells and CD45RA−CD45RO+CD
none of the participants had been vaccinated nor re-infected. b,c, Expression 127−CD25+ Treg cells among total CD4+ T cells from LC and R individuals. **P < 0.01,
of IFN-γ, TNF or IL-2 in CD4+ T cells (b) or IFN-γ, TNF or CCL4 in CD8+ T cells (c) *P < 0.05 (two-sided Student’s t test). h, Frequencies of TN cells, TSCM cells, TCM
stimulated (bottom) or not (top) with SARS-CoV-2 spike and T-scan peptides cells, TEM cells, TTM cells, TEMRA cells, pTFH cells, TFH cells and Treg cells among
(Methods). d, Frequency of SARS-CoV-2-specific CD4+ or SARS-CoV-2-specific SARS-CoV-2-specific CD4+ T cells from LC and R individuals. Horizontal bars
CD8+ T cells in LC and R individuals (two-sided Student’s t tests). e,f, Frequency indicate mean, error bars indicate s.d., and dots represent individuals, with n = 27
of monofunctional or polyfunctional SARS-CoV-2-specific CD4+ (e) or LC and n = 16 R (d, g and h). NS, not significant; WHO, World Health Organization.
SARS-CoV-2-specific CD8+ (f) T cells in LC versus R individuals. Polyfunctional

Nature Immunology | Volume 25 | February 2024 | 218–225 220

Letter https://doi.org/10.1038/s41590-023-01724-6

a b
R LC High
40
CXCR4 CXCR5 CCR6
0.5 0.4 R
20 0.6
0.4 0.3
LC

% max

% max
% max
0 0.3 0.4
0.2
0.2
−20 0.1 0.2
0.1
t-SNE 2

0 0 0
−40
Low 2 3 4 5 6 0 1 2 3 2 3 4 5
−50 −25 0 25 50
MSI MSI MSI
t-SNE 1

c d
CXCR4+CXCR5+ cells CXCR5+CCR6+ cells CXCR4+CCR6+ cells CXCR4+CXCR5+ cells CXCR5+CCR6+ cells CXCR4+CCR6+ cells
10 10 100 2.0 1.5 80 R

Total CD4+ T cells (%)

NS

Total CD4+ T cells (%)

Total CD4 T cells (%)

NS
SARS-CoV-2-specific

SARS-CoV-2-specific

R
SARS-CoV-2-specific
* * * LC
CD4+ T cells (%)

CD4+ T cells (%)

CD4+ T cells (%)

8 8 80 LC 1.5
1.0 * 60
6 6 60
1.0 40
4 4

+
40
0.5
2 2 0.5 20
20
0 0 0 0 0
0

e f

(% of SARS-CoV-2-specific
R R

(% of SARS-CoV-2-specific
(% of total CD4+ T cells)

(% of total CD4+ T cells)

CXCR4 CXCR5 T cells

CXCR4+CXCR5+ T cells

2.0 2.0 10 10
CXCR5+CCR6+ T cells

LC r = 0.7585 LC

CXCR5+CCR6+ T cells
r = 0.8649 8 P < 0.0001 8 r = 0.7865

CD4+ T cells)
1.5 1.5

CD4+ T cells)
P < 0.0001 + P < 0.0001
r = 0.6501 6 6
1.0 1.0 P = 0.0002
4 4
0.5 0.5
+

r = 0.5787 2 r = –0.3051 2 r = –0.2164

r = 0.4331
P = 0.0188 P = 0.2505 P = 0.4207
0 0 P = 0.0938 0 0
0 0.5 1.0 1.5 2.0 0 0.5 1.0 1.5 2.0 0 10 20 30 0 10 20 30
pTFH cells (% of total CD4+ T cells) pTFH cells (% of total CD4+ T cells) pTFH cells (% of SARS-CoV-2- pTFH cells (% of SARS-CoV-2-
specific CD4+ T cells) specific CD4+ T cells)

Fig. 2 | SARS-CoV-2-specific CD4+ T cells from individuals with LC CXCR5+CCR6+CD4+ and CXCR4+CCR6+CD4+ SARS-CoV-2-specific (c) and
preferentially express homing receptors associated with migration to total (d) CD4+ T cells in LC and R individuals. *P < 0.05 (two-sided Student’s
inflamed tissues. a, t-SNE contour depiction of SARS-CoV-2-specific CD4+ T t test). e,f, Associations of percentages of total (e) or SARS-CoV-2-specific
cells from LC and R individuals. b, Expression of CXCR4, CXCR5 and CCR6 in (f) CXCR4+CXCR5+CD4+ and CXCR5+CCR6+CD4+ T cells with percentages of
SARS-CoV-2-specific CD4+ T cells from LC and R individuals. MSI corresponds to pTFH cells in LC and R individuals. Data were analyzed by Pearson correlation
the mean signal intensity of the indicated markers’ expression level, reported coefficient and two-tailed unpaired t tests. Horizontal bars indicate mean, error
as arcsinh-transformed CyTOF data. c,d, Percentages of CXCR4+CXCR5+CD4+, bars indicate s.d. and dots represent individuals, with n = 27 LC and n = 16 R (c,d).

response, previously implicated in severe COVID-19 (ref. 8), may also abundant (P = 0.006) and the platelet cluster more abundant (P = 0.01)
be a hallmark of LC. in LC compared to R individuals, while the other clusters (CD4+ T cells,
Bulk RNA-seq identified only two genes, OR7D2 and ALAS2, that CD8+ T cells, CTLs, B cells, monocytes and NKT/NK/MAIT/γδ T cells) did
were significantly differentially expressed between LC and R. OR7D2 not differ between the groups (Fig. 5d). Visualization based on LC versus R
encodes a G-protein-coupled receptor that is activated by odorant mol- status, or based on OR7D2hi LC versus ALAS2hi LC, did not reveal profound
ecules, whereas ALAS2 encodes an enzyme that catalyzes the first step differences (Extended Data Fig. 9a,b). Among all cells, OR7D2 expression
in heme synthesis to generate δ-aminolevulinic acid from succinyl-CoA was highest in cells of the OR7D2hi LC group and ALAS2 was highest in cells
and glycine. Both OR7D2 and ALAS2 were overexpressed in LC individu- of the ALAS2hi LC group, and all clusters except granulocytes and platelets
als although not necessarily together, as the four individuals with the expressed OR7D2 and ALAS2 (Extended Data Fig. 9c–e).
highest OR7D2 expression in peripheral blood mononuclear cells Interrogation of cluster-specific gene expression identified three
(PBMCs) did not have the highest ALAS2 expression (Fig. 5a). Super- additional genes (THEMIS, NUDT2 and PPIE) that were differentially
vised clustering found upregulation of a module of genes that regu- expressed (P < 0.05) in LC individuals, two within CD8+ T cell cluster 1
late heme synthesis and carbon dioxide transport (ALAS2, HBB, CA1, and one within monocyte cluster 3 (Fig. 5e). Using a less stringent cut-
HBA1, SLC4A1, HBD and HBA2) and the downregulation of a module off (P < 0.1), we found 16 differentially expressed genes (DEGs) within
consisting of immunoglobin kappa, lambda and heavy chain genes in CD8+ T cell cluster 1 (for example, THEMIS, HMGB2 and TNFRSF18),
LC compared to R individuals (Fig. 5b,c), suggesting the involvement monocyte cluster 3 (PPIE) and CD4+ T cell cluster 7 (for example, CAST
of heme biosynthesis and immune dysregulation in LC. and APBA2; Fig. 5f and Supplementary Table 5). Gene Ontology (GO)
To gain a more granular view of the transcriptome, we selected a pathway analysis found significant (P < 0.05) differences between LC
subset of the specimens analyzed by bulk RNA-seq for repeat analysis by and R individuals within monocyte cluster 3, in pathways associated
scRNA-seq. We limited these studies to females because individuals with with transcriptional regulation and splicing, protein regulation and
high levels of OR7D2 or ALAS2 were mostly female (the top five OR7D2 neutrophil degranulation (Supplementary Table 6). Trends (P < 0.1)
expressors were female, as were five of the top six ALAS2 expressors). were observed for pathways associated with apoptosis and metabolism
For comparison, we included four randomly selected females from the and/or oxidative stress in CD8+ T cell cluster 1 (Supplementary Table 7).
R specimens. Integration of data from all 12 samples identified 11 clusters CXCR4, CXCR5 and CCR6 were upregulated in CD4+ T cell clusters 0
of cells and revealed that the granulocyte cluster was significantly less and 7 from LC compared to their counterpart clusters in R (Extended

Nature Immunology | Volume 25 | February 2024 | 218–225 221

Letter https://doi.org/10.1038/s41590-023-01724-6

a High
b
R LC
40
PD1 CTLA4 TIGIT
0.8 R
20
0.6 0.9 LC
0.6

% max

% max

% max
0 0.4 0.6
0.4

−20 0.2 0.2 0.3

t-SNE 2

0 0 0
−40
0 1 2 3 1 2 3 4 0 1 2 3
−50 −25 0 25 50 Low
MSI MSI MSI
t-SNE 1

c d
PD1+CTLA4+ cells TIGIT+CTLA4+ cells PD1+TIGIT+ cells PD1+CTLA4+ cells TIGIT+CTLA4+ cells PD1+TIGIT+ cells
40 60 40 R 15 NS 40 NS 8 NS R
*

Total CD8+ T cells (%)

SARS-CoV-2-specific

SARS-CoV-2-specific

SARS-CoV-2-specific

Total CD8+ T cells (%)

Total CD8+ T cells (%)

NS NS LC LC
CD8+ T cells (%)

CD8+ T cells (%)

CD8+ T cells (%) 30

30 30 6
40 10
20 20 20 4
20 10 5
10 10 2

0 0
0 0 0 0

Fig. 3 | SARS-CoV-2-specific CD8+ T cells from individuals with LC preferentially c,d, Percentages of PD1+CTLA4+CD8+, TIGIT+CTLA4+CD8+ and PD1+TIGIT+CD8+
express the exhaustion markers PD1 and CTLA4. a, t-SNE contour depiction of SARS-CoV-2-specific (c) and total (d) CD8+ T cells in LC and R individuals. *P < 0.05
SARS-CoV-2-specific CD8+ T cells from LC and R individuals. b, Expression of PD1, (two-sided Student’s t test). Horizontal bars indicate mean, error bars indicate s.d.
CTLA4 and TIGIT on SARS-CoV-2-specific CD8+ T cells from LC and R individuals. and dots represent individuals, with n = 27 LC and n = 16 R (c,d).

a b
PD1 CTLA4 T cells (% of SARS-

PD1+CTLA4+ T cells (% of SARS-

R 40 40
CoV-2-specific CD8+ T cells)

CoV-2-specific CD8+ T cells)

5
* R
RBD antibody (RFU ×103)

LC
4 LC
30 30
r = –0.5395
3
P = 0.0037
20 20
2
+

1 10 10 r = 0.1605
P = 0.5528
0
+

0 0
0 1 2 3 4 5 0 5 10 15 20

RBD antibody (RFU ×103) SARS-CoV-2-specific CD4+ Treg cells (%)

c 1.5 1.5
0.025 R
r = 0.7458
SARS-CoV-2-specific

SARS-CoV-2-specific

LC
SARS-CoV-2-specific
CD4+ pTFH cells (%)
CD4+ T cells (%)

P = 0.0014
CD8+ T cells (%)

0.02
1.0 r = 0.5308 1.0 r = 0.7764
P = 0.0418 P = 0.0007
0.015

r = –0.0164 0.01 r = 0.0557

0.5 0.5
P = 0.7868 r = –0.1493
P = 0.9367
0.005
P = 0.4667

0 0 0
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
3 3 3
RBD antibody (RFU ×10 ) RBD antibody (RFU ×10 ) RBD antibody (RFU ×10 )

Fig. 4 | Humoral and cellular immunity are discoordinated in individuals with humoral response are circled in green, and those with the highest percentages of
LC. a, Total SARS-CoV-2 RBD-specific antibody levels in LC and R individuals. PD1+CTLA4+ SARS-CoV-2-specific CD8+ T cells are circled in purple (left). c, Plot
*P < 0.05 (two-sided Student’s t test). Horizontal bars indicate mean, error depicting the association between RBD antibody levels and the percentages of
bars indicate s.d. and dots represent individuals. LC (n = 26), R (n = 15). b, Plot SARS-CoV-2-specific CD4+ T cells, SARS-CoV-2-specific CD4+ pTFH cells (middle)
depicting the percentage of PD1+CTLA4+ cells among SARS-CoV-2-specific CD8+ T and SARS-CoV-2-specific CD8+ T cells (right) in LC and R individuals. Data were
cells and RBD antibody levels in LC and R individuals. Individuals with the highest analyzed by Pearson correlation coefficient and two-tailed unpaired t tests.

Data Fig. 9f). Comparison of OR7D2hi LC versus R revealed 35 DEGs Table 9). GO pathways associated with the OR7D2hi LC DEGs included
in the OR7D2hi LC group (Extended Data Fig. 9g and Supplementary lipid transport and stress responses in CD4+ T cell cluster 7, RNA splicing
Table 8) including upregulation of the histone family genes HIST1H2AM, in CD8+ T cell cluster 5 and immunoglobulin (Ig) production in B cell
HIST2H2AC and HIST1H1E, while comparison of ALAS2hi LC versus R cluster 8 (Supplementary Table 10), while those associated with the
revealed 14 DEGs including upregulation of THEMIS and downregula- ALAS2hi LC DEGs included apoptosis and oxidative stress responses in
tion of BACH2 (P < 0.05; Extended Data Fig. 9h and Supplementary CD8+ T cell cluster 1 (Supplementary Table 11).

Nature Immunology | Volume 25 | February 2024 | 218–225 222

Letter https://doi.org/10.1038/s41590-023-01724-6

a b R LC

OR7D2-normalized counts
80
* 20
* R Status
* * LC

ALAS2-normalized
OR7D2 4
60 15 SCN3A

counts (×102)
PPP1R9A 2
L1TD1
10 AC011247.2
40
* 0
*
ADRA2A
LINC02280
20
* 5 * ARC
–2
KNDC1
–4
** TRIM10
TRIM58
–6
0 0 GREB1L
SNTG2
GPAT2P1
c TMEM158
AL390195.1
CHD5
DEPDC1B CDC20 RELB LINC01841
MTND2P28
IGLV10-54
IGHG2
IGHG4
BIRC5 UBA52 GP9 CDC20
HBA2 CADM1
BIRC5
IGHG1
IGKV2-24
RPS27A GATA1 ALAS2
UBC HBB HBB
CA1
HBE1 HBA1
SLC4A1
HP SERPINE1 HBD
GLI1
HBA2
AL450405.1
HBA1 HBQ1 HBG2 SERPINE1
TRBV5-6
IQCD
log2(fold change) HBD JAZF1-AS1
ELL2P1
TOP1MT
AHSP RELB
ALAS2 DEPDC1B
–4.0 –2.0 0 2.0 4.0 GP9
MT-TV
SLC4A1 HDAC11
HBG1 GALNTL6
CLECL1
CA1 AC020659.1
GLI1

d e Cluster 1 Cluster 1 Cluster 3

300 40 25 R
* * * LC
20
30
THEMIS (cpm)

NUDT2 (cpm)
+
CD4 T cells 200

PPIE (cpm)
+
CD8 T cells/CTLs 15
B cells 20
6
10
Monocytes 100
NKT/NK/MAIT/gd 10
2 8 + 5
CD8 T cells
B cells 0 0 0
5 +
CD4 T cells
B cells f Cluster 1 Cluster 3 Cluster 7
UMAP 2

9 Granulocytes NUDT2 6 APBA2 NIBAN1

4 Platelets TNFRSF18 THEMIS
7 PHTF1 HMGB2 PPIE CAST
3 STAG3 TSPAN3 NSMAF 4 RNF157-AS1
1
10
4
MARCH3 DPM3 ABCD2 4
–log10(P)

–log10(P)

–log10(P)
3

2
0 2 2
1

0 0 0

UMAP 1 –2 –1 0 1 2 –2 –1 0 1 2 –2 –1 0 1 2
log2(fold change) log2(fold change) log2(fold change)

g R LC h
Status 4
IL-4
3
SPRY2
2 IL12B
HSD11B1
IL-5 IL-4
1
VEGFD
0 CCL22
IL-33 TNF
JUN –1 CXCL8 LAIR1
KRT19 –2
ITGB6 PLAUR
–3 IL-5
NTF3 IL-6 CD83
ERBB3 CD40
IDS IL-10
HSD11B1
LGALS9
IL-1B
CD4
SIRPB1
HGF
AGRN PTPRC
CCL21
SLAMF1
ITGAM
CCL22
TNF CD86
CXCL10
CD200 KRT19
CD48 CD8A
CSF1 LGALS9
HGF Difference
CD22 CD22 CD79B
CRHBP
IL1RN –1.0 –0.5 0 0.5 1.0

Olink proteomics indicated elevated expression of proteins of IL-4 and decreased expression of IL-5 compared to R individuals
associated with inflammation (LGALS9, CCL21, CCL22, TNF, CXCL10 (Fig. 5g,h), although both cytokines are associated with T helper 2
and CD48) and immune regulation (IL1RN and CD22) in LC compared (TH2) cell responses. CCL22, a ligand for the TH2 cell marker CCR4, was
to R individuals (Fig. 5g). LC individuals had elevated expression expressed at elevated levels in LC compared to R individuals (Fig. 5h).

Nature Immunology | Volume 25 | February 2024 | 218–225 223

Letter https://doi.org/10.1038/s41590-023-01724-6

Fig. 5 | Global changes in gene and gene product expression in the blood of cluster 1 and PPIE in monocyte cluster 3 in LC versus R individuals as determined
individuals with LC. a, Relative expression of OR7D2 and ALAS2 as determined by scRNA-seq analysis. *P < 0.05 (two-sided empirical Bayes quasi-likelihood F
by bulk RNA-seq analysis of whole blood from LC versus R individuals. *P < 0.05 tests, with Benjamini–Hochberg correction). Horizontal bars indicate mean, error
(two-sided Wald test, Benjamini–Hochberg correction). Purple asterisks identify bars indicate s.d. and dots represent individuals. LC (n = 8) and R (n = 4). f, Volcano
the female donors selected for scRNA-seq analyses. Horizontal bars indicate mean, plots depicting DEGs in LC versus R individuals in scRNA-seq-defined clusters.
error bars indicate s.d. and dots represent individuals. LC (n = 23) and R (n = 13). DEGs with P < 0.1 (two-sided empirical Bayes quasi-likelihood F tests, Benjamini–
b, Heatmap of the top 50 DEGs in LC versus R individuals based on clustering Hochberg correction) are labeled. The x axes represent the log2(fold change) of
analysis of bulk RNA-seq data. Genes are grouped into k clusters based on similarity. the mean expression of each gene between the comparison groups, and the y axes
c, Network mapping of DEGs from bulk RNA-seq analysis. Each node corresponds represent the raw −log10(P values). Dashed horizontal lines delineate thresholds
to a gene; colors of nodes indicate the extent of change; red indicates upregulation corresponding to Benjamini–Hochberg adjusted P values of <0.1. g, Clustered
and blue indicates downregulation in LC compared to R. Edges depict the functional heatmap of the top 25 differentially expressed proteins from Olink analysis
relevance between pairs of genes, where thickness corresponds to confidence performed on plasma of LC and R individuals with markers grouped into k-means
of evidence. d, UMAP of clusters of all LC and R PBMCs analyzed by scRNA-seq. clusters based on similarity. LC (n = 25) and R (n = 15). h, Network mapping of related
LC (n = 8) and R (n = 4). e, Relative expression of THEMIS and NUDT2 in CD8+ T cell differentially expressed proteins as detected by Olink. Graph representations as in c.

IL-4, but not IL-5 or CCL22, significantly positively associated with the and competing interests; and statements of data and code availability
percentages of total CXCR4+CXCR5+CD4+ and CXCR5+CCR6+CD4+ are available at https://doi.org/10.1038/s41590-023-01724-6.
T cells in LC individuals (Extended Data Fig. 8e), suggesting an elevated,
yet mis-coordinated, TH2 cell response during LC. References
In summary, using multiple ‘omics’ analytical approaches, we 1. Davis, H. E., McCorkell, L., Vogel, J. M. & Topol, E. J. Long COVID:
found that LC individuals exhibited phenotypic perturbations in both major findings, mechanisms and recommendations. Nat. Rev.
total and SARS-CoV-2-specific CD4+ and CD8+ T cells and changes in Microbiol. 21, 133–146 (2023).
gene expression among CD4+ T cells, CD8+ T cells, monocytes and 2. Ma, T. et al. Protracted yet coordinated differentiation
B cells. We found higher proportions of CD4+ TCM cells, TFH cells and of long-lived SARS-CoV-2-specific CD8+ T cells during
Treg cells in LC compared to R individuals. SARS-CoV-2-specific CD8+ convalescence. J. Immunol. 207, 1344–1356 (2021).
T cells, but not total CD8+ T cells, more frequently expressed the 3. Neidleman, J. et al. SARS-CoV-2-specific T cells exhibit phenotypic
exhaustion markers PD1 and CTLA4, consistent with ongoing stimu- features of helper function, lack of terminal differentiation, and
lation by viral antigens. Further supporting a potential persistent high proliferation potential. Cell Rep. Med. 1, 100081 (2020).
reservoir was our observation of higher SARS-CoV-2 antibody levels 4. Neidleman, J. et al. Distinctive features of SARS-CoV-2-specific
in LC individuals, consistent with reports of higher spike-specific T cells predict recovery from severe COVID-19. Cell Rep. 36,
IgG in LC compared to R individuals9. CyTOF, flow cytometry and 109414 (2021).
scRNA-seq indicated that CD4+ T cells from LC individuals prefer- 5. Neidleman, J. et al. mRNA vaccine-induced T cells respond
entially expressed CXCR4, CXCR5 and CCR6. CXCR4 expression identically to SARS-CoV-2 variants of concern but differ in
is elevated on bystander CD4+ and CD8+ T cells in fatal COVID-19 longevity and homing properties depending on prior infection
(ref. 4) and on pulmonary CD4+ T cells, B cells, macrophages and gran- status. eLife 10, e72619 (2021).
ulocytes in the context of LC following SARS-CoV-2 infection of mice10. 6. Suryawanshi, R. K. et al. Limited cross-variant immunity from
Although fully recovered individuals exhibited coordinated humoral SARS-CoV-2 Omicron without vaccination. Nature 607, 351–355
and cellular immune responses to SARS-CoV-2, this coordination was (2022).
lost in LC individuals, consistent with observations that about half of 7. Peluso, M. J. et al. Persistence, magnitude, and patterns of
individuals with LC with no detectable SARS-CoV-2 antibodies have postacute symptoms and quality of life following onset of
detectable SARS-CoV-2-specific T cell responses11. How the humoral SARS-CoV-2 infection: cohort description and approaches for
response becomes divorced from the cellular response is unclear, measurement. Open Forum Infect. Dis. 9, ofab640 (2022).
but could involve a misalignment between IL-4 and IL-5 production 8. Rydyznski Moderbacher, C. et al. Antigen-specific adaptive
by TH2 cells, as indicated by our Olink analysis. immunity to SARS-CoV-2 in acute COVID-19 and associations with
Our study has limitations. First, the cohort analyzed included only age and disease severity. Cell 183, 996–1012 (2020).
43 participants; however, the rigor with which participants were charac- 9. Files, J. K. et al. Duration of post-COVID-19 symptoms is
terized mitigates the limitations of the small sample size. Some findings associated with sustained SARS-CoV-2-specific immune
were driven by small subsets of LC individuals, which is consistent with responses. JCI Insight 6, e151544 (2021).
the notion of LC being a heterogeneous disease, and will require vali- 10. Ma, T. et al. Post-acute immunological and behavioral sequelae
dation in larger cohorts. Second, due to limited channels available for in mice after Omicron infection. Preprint at bioRxiv https://doi.
CyTOF, we did not examine additional markers that would have been of org/10.1101/2023.06.05.543758 (2023).
interest such as the exhaustion marker thymocyte selection-associated 11. Krishna, B. A. et al. Evidence of previous SARS-CoV-2 infection in
high mobility group protein (TOX)12, the activation marker CD40L and seronegative patients with long COVID. EBioMedicine 81, 104129
the proliferation marker 5-Iodo-2'-deoxyuridine (IdU)13. Third, the (2022).
changes we saw in the blood subsets could reflect migration to tissues. 12. Khan, O. et al. TOX transcriptionally and epigenetically programs
Finally, our study was for the most part descriptive. However, for new CD8+ T cell exhaustion. Nature 571, 211–218 (2019).
and poorly understood diseases, in-depth ‘omics’-based characteriza- 13. Devine, R. D. & Behbehani, G. K. Use of the pyrimidine analog,
tion of a well-annotated cohort is the critical first step for better under- 5-iodo-2′-deoxyuridine (IdU) with cell cycle markers to establish
standing the condition’s etiology and mechanistic underpinnings. cell cycle phases in a mass cytometry platform. J. Vis. Exp.,
https://doi.org/10.3791/60556 (2021).
Online content 14. World Health Organization. A clinical case definition of
Any methods, additional references, Nature Portfolio reporting summa- post COVID-19 condition by a Delphi consensus. www.who.
ries, source data, extended data, supplementary information, acknowl- int/publications/i/item/WHO-2019-nCoV-Post_COVID-19_
edgements, peer review information; details of author contributions condition-Clinical_case_definition-2021.1 (2021).

Nature Immunology | Volume 25 | February 2024 | 218–225 224

Letter https://doi.org/10.1038/s41590-023-01724-6

Publisher’s note Springer Nature remains neutral with regard to material in this article are included in the article’s Creative Commons
jurisdictional claims in published maps and institutional affiliations. license, unless indicated otherwise in a credit line to the material. If
material is not included in the article’s Creative Commons license
Open Access This article is licensed under a Creative Commons and your intended use is not permitted by statutory regulation
Attribution 4.0 International License, which permits use, sharing, or exceeds the permitted use, you will need to obtain permission
adaptation, distribution and reproduction in any medium or format, directly from the copyright holder. To view a copy of this license, visit
as long as you give appropriate credit to the original author(s) and http://creativecommons.org/licenses/by/4.0/.
the source, provide a link to the Creative Commons license, and
indicate if changes were made. The images or other third party © The Author(s) 2024

Gladstone Institutes, University of California, San Francisco, San Francisco, CA, USA. 2Department of Urology, University of California, San Francisco, San
1

Francisco, CA, USA. 3Division of HIV, Infectious Diseases, and Global Medicine, University of California, San Francisco, San Francisco, CA, USA. 4Division
of Experimental Medicine, University of California, San Francisco, San Francisco, CA, USA. 5Core Facility Functional Peptidomics, Ulm University Medical
Center, Ulm, Germany. 6Department of Epidemiology and Biostatistics, University of California, San Francisco, San Francisco, CA, USA. 7Zuckerberg
San Francisco General Hospital and the University of California, San Francisco, San Francisco, CA, USA. 8Division of Laboratory Medicine, University of
California, San Francisco, San Francisco, CA, USA. 9These authors contributed equally: Kailin Yin, Michael J. Peluso. e-mail: [email protected];
[email protected]

Nature Immunology | Volume 25 | February 2024 | 218–225 225

Letter https://doi.org/10.1038/s41590-023-01724-6

Methods Beads (Standard BioTools) and run on a Helios CyTOF instrument (UCSF
Study participants Parnassus Flow Core).
Participants were enrolled in LIINC (www.liincstudy.org;
NCT04362150)7, a prospective observational study enrolling individu- CyTOF data analyses
als with prior nucleic acid-confirmed SARS-CoV-2 infection, regardless Data preprocessing. EQ bead-normalized CyTOF datasets were con-
of the presence or absence of postacute symptoms. At each study visit, catenated, de-barcoded and normalized using Standard BioTools
participants underwent an interviewer-administered assessment of Software version 6.7. Following arcsinh transformation of the data18,
32 physical symptoms that were newly developed or had worsened cells were analyzed by FlowJo (version 10.8.1, BD Biosciences). Intact
since the COVID-19 diagnosis. Detailed data regarding medical history, (Ir191+Ir193+), live (Pt195−), singlet events were identified, followed by
COVID-19 history, SARS-CoV-2 vaccination and SARS-CoV-2 reinfec- gating on CD3+ T cells, and sub-gating on CD4+ T cells and CD8+ T cells
tion were collected. Two participants had biospecimens collected via (Extended Data Fig. 1g,h).
the COVID-19 Host Immune Response Pathogenesis (CHIRP) study5.
For the present study, we selected participants who consistently met CyTOF antibody validation. CyTOF antibodies in our panel (Supple-
a case definition for LC based on the presence or absence of at least mentary Table 4) were validated using methods previously described,
one symptom attributable to COVID-19 for the 8 months following including the use of human lymphoid aggregate cultures generated
SARS-CoV-2 infection (Fig. 1a). The LC group (n = 27) had a median age from tonsils2–5,18,19. The observed expression patterns among tonsillar
of 46 years, and was comprised of 63% females and 26% of whom were T and B cells (Extended Data Fig. 10a) were similar to those previously
previously hospitalized for COVID-19. The R group (n = 16) had a median observed18. To validate the detection of cytokines and other effectors,
age of 45.5 years, and was comprised of 44% females and 12.5% of whom we stimulated PBMCs with 16 nM phorbol 12-myristate 13-acetate
were previously hospitalized for COVID-19 (Supplementary Table 1). (PMA) (Sigma-Aldrich) and 1 μM ionomycin (Sigma-Aldrich), or 1 μg ml−1
Participants were deliberately not matched by age and sex, but we lipopolysaccharides (LPS; eBioscience), for 4 h in the presence of
ensured that there was overlap in the groups. Blood samples were col- 3 μg ml−1 BFA solution (eBioscience), combined the cells and prepared
lected between September 16, 2020 and April 6, 2021. All participants them for CyTOF as described above. We observed the expected induc-
provided a post-COVID blood sample before a SARS-CoV-2 vaccination tion of cytokines or cytolytic markers (Extended Data Fig. 10b)2–5 and
to exclude the potential effects of SARS-CoV-2 vaccination on our study. preferential expression of Treg lineage marker Foxp3 among CD3+CD4
+
Specimens were collected 8 months postinfection from individuals. CD45RO+CD45RA−CD127−CD25+ Treg cells (Extended Data Fig. 10c). We
All assays were performed from the same parent set of n = 27 LC and also observed preferential expression of CD30 and Ki67 in CD4+ TM as
n = 16 specimens. All participants provided written informed consent. compared to CD4+ TN cells (Extended Data Fig. 10d). Examples of pTFH
and TFH gates are depicted in Extended Data Fig. 10e.
Biospecimen collection
Whole blood was collected in EDTA tubes followed by isolation of Identification of SARS-CoV-2-specific T cells. For identification of
PBMCs and plasma as described in ref. 15. Serum was obtained con- SARS-CoV-2-specific T cells, we compared unstimulated specimens to
comitantly from serum-separator tubes. their peptide-stimulated counterparts. Effector cytokines (IFN-γ, TNF,
IL-2, IL-4, IL-6, IL-17 and CCL4) and cytolytic effectors (granzyme B and
Serology perforin) were assessed for the ability to identify antigen-specific T cells
Antibody responses against SARS-CoV-2 spike RBD were measured on at the single-cell level. The following criteria were established to iden-
sera using the Pylon COVID-19 total antibody assay (ET Health) and tify effector molecules appropriate for identifying SARS-CoV-2-specific
reported as relative fluorescence units (RFUs). T cells: (1) counts of positive cells in unstimulated sample (not receiving
peptide) was less than 5 events, or the frequency of positive cells was
SARS-CoV-2 peptides lower than 0.1%; (2) counts of positive cells in the peptide-stimulated
Peptides used for T cell stimulation comprised a mix of overlapping sample was not less than 5, or the frequency was higher than 0.1%; (3)
15-mers spanning the entire SARS-CoV-2 spike protein (PM-WCPV-S-1, differences in frequencies of positive cells between unstimulated
purchased from JPT), and peptides corresponding to CD8+ T cell and peptide-stimulated samples cells was not less than 0.01%; (4)
epitopes identified by T-scan16 synthesized in-house (Supplementary fold change in frequencies of positive cells between unstimulated
Table 12). Final peptide concentrations were 300 nM for the 15-mers and peptide-stimulated samples cells was greater than 10 and (5) the
and 450 nM for the T-scan peptides. aforementioned four criteria could identify SARS-CoV-2-specific T cells
among >50% of participants. Effectors that fulfilled all five criteria
CyTOF were IFN-γ, TNF and IL-2 for CD4+ T cells and IFN-γ, TNF and CCL4 for
Sample preparation was performed similar to methods described2–5. CD8+ T cells. For a sub-analysis to identify responding cells that may
Upon revival of cryopreserved PBMCs, cells were rested overnight to only exist in a small subset of individuals, we removed criterion 5 and
allow for antigen recovery17 and then divided equally into two aliquots. reduced the positive cell counts to number 3 within criteria 1 and 2.
To the first aliquot, we added 3 µg ml−1 brefeldin A (BFA; to enable This approach allowed us to determine that SARS-CoV-2-specific CD4+
intracellular cytokine detection), the costimulation agonists anti-CD28 T cells producing IL-6 were exclusively detected from LC (Extended
(2 µg ml−1; BD Biosciences) and anti-CD49d (1 µg ml−1; BD Biosciences), Data Fig. 2f). SARS-CoV-2-specific T cells were detected at a median of
and the SARS-CoV-2 peptide pool prepared as described above. To the 163 cells (134 for CD4+ T cells and 29 for CD8+ T cells) and a mean of 221.7
second aliquot, we added 1% DMSO (Sigma-Aldrich) and 3 µg ml−1 BFA. cells (185.2 for CD4+ T cells and 36.4 for CD8+ T cells), per participant.
Cells from both treatments were incubated at 37°C for 6 h. Cells were SARS-CoV-2-specific T cells, once identified, were analyzed by Boolean
treated with cisplatin (Sigma-Aldrich) as a live/dead distinguisher gating20 and exported for further analyses.
and fixed in paraformaldehyde (Electron Microscopy Sciences) as
described2–5. CyTOF antibody conjugation was performed using the SPICE. SPICE analyses were performed using version 6.1 software21.
Maxpar X8 Antibody Labeling Kit (Standard BioTools) according to CD4+ and CD8+ T cells were subjected to manual gating based on the
the manufacturer’s instructions. CyTOF staining was performed as expression of cytokines used to define SARS-CoV-2-specific T cells
described2–5, but using the CyTOF panel created for this study (Supple- (IFN-γ, TNF, IL-2 and CCL4, see above) using operations of Boolean logic.
mentary Table 4). Stained samples were washed with CAS buffer (Stand- The parameters for running the dataset were as follows: iterations for
ard BioTools), spiked with 10% (vol/vol) EQ Four Element Calibration permutation test = 10,000 and highlight values = 0.05. The parameters

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

for the query structure were set as follows: values = frequency of single fragmentation and random priming, cDNA syntheses were completed.
cytokine positive cells in total CD4+/CD8+ T cells; category = IFN-γ, TNF, End repair, 5′ phosphorylation and dA-tailing were performed, followed
IL-2 and CCL4; overlay = patient type (LC versus non-LC); group = all by adaptor ligation, PCR enrichment and sequencing on an Illumina
other variables in the data matrix. HiSeq platform using PE150 (paired-end sequencing, 150 bp for reads
1 and 2). Raw reads (480 Gb in total) were trimmed using Trimmomatic
T cell subsetting. Manual gating was performed using R (ver- (version 0.36) to remove adapter sequences and poor-quality reads.
sion 4.1.3). Arcsinh-transformed data corresponding to total or Trimmed reads were mapped to Homo sapiens GRCh37 using star
SARS-CoV-2-specific CD4+ or CD8+ T cells were plotted as 2D plots aligner (version 2.5.2b)28. log2 fold changes were calculated between
using the CytoExploreR package. Visualization of datasets by t-SNE was LC versus R individuals. Two-sided P values corresponding to a null
performed using methods similar to those described2–5. CytoExploreR hypothesis of fold change of 1 were calculated using DESeq2’s (ref.29)
and tidyr packages were used to load the data, and t-SNE was performed Wald test and were adjusted for multiple testing using false discovery
using Rtsne and RColorBrewer packages on arcsinh-transformed mark- rates. Genes with an adjusted P value < 0.05 and absolute log2(fold
ers. Total CD4+/CD8+ T cells were downsampled to n = 8,000 (maximal change) > 1 were considered significant DEGs. Clustered heatmaps of
cell number for individual samples) before t-SNE analysis. The param- DEGs were constructed with groups of genes (rows) defined using the
eters for t-SNE were set as iteration = 1,000, perplexity = 30 and θ = 0.5. k-means algorithm to cluster genes into k clusters based on their simi-
larity. K = 4 was determined using the Hierarchical Ordered Partition-
T cell clustering analysis. Flow cytometry standard (FCS) files corre- ing and Collapsing Hybrid (HOPACH) algorithm30, which recursively
sponding to total and SARS-CoV-2-specific CD4+ and CD8+ T cells were partitions a hierarchical tree while ordering and collapsing clusters at
imported in R for data transformation. Packages of flowcore, expss, each level to identify the level of the tree with maximally homogene-
class and openxlsx were loaded in R. Arcsinh-transformed data were ous clusters.
then exported as CSV files for clustering analyses. Biological (LC status,
biological sex and hospitalization status) and technical (batch/run of scRNA-seq
processing) variables were visualized using the DimPlot function of scRNA-seq was performed on PBMCs from 8 LC and 4 R individuals from
Seurat22. Batch correction was performed by RunHarmony23. Optimal our cohort, obtained from aliquots of specimens analyzed by CyTOF.
clustering resolution parameters were determined using Random Library preparation was performed using the Chromium Next GEM
Forests24 and a silhouette score-based assessment of clustering validity Single-Cell 5′ Reagent Kits v2 (10x Genomics) and sequenced on the
and subject-wise cross-validation, as detailed in ref. 25. A generalized Illumina NovaSeq 6000 S4 300 platform. Samples were sequenced at a
linear mixed model (GLMM, implemented in the lme4 (ref. 26) package mean of >50k reads per cell (minimum 51k, maximum 120k and median
in R with family argument set to the binomial probability distribution) 83k). A median of 7,888 cells was analyzed per donor (minimum 4,189
was used to estimate the association between cluster membership and and maximum 9,511). Demultiplexed fastq files were aligned to human
LC status and the sex of the participant, with the participant modeled reference genome GRCh38 using the 10x Genomics Cell Ranger v7.1.0
as a random effect. For each individual, cluster membership of cells count pipeline31. The include-introns flag for the count pipeline was set
was encoded as a pair of numbers representing the number of cells in to true to count reads mapping to intronic regions. The filtered count
the cluster and the number of cells not in the cluster. Clusters having matrices generated by the Cell Ranger count pipeline were processed
fewer than three cells were discarded. The sex-specific log odds ratio using Seurat22. Each sample was preprocessed as a Seurat object, and
of cluster membership association with LC status was estimated using the top 1% of cells per sample with the highest numbers of unique genes,
the emmeans27 R package using the GLMM model fit. The estimated log cells with ≤200 unique genes and cells ≥10% mitochondrial genes were
odds ratio represented the change (due to LC status) in the average over filtered out for each sample. The samples were then merged into a sin-
all participants of a given sex in the log odds of cluster membership. gle Seurat object, and normalization and variance stabilization were
The two-sided P values corresponding to the null hypothesis of an performed using sctransform86 with the ‘glmGamPoi’ method32 for
odds ratio value of 1 were computed based on a z statistic in the GLMM initial parameter estimation.
model fit. These P values were adjusted for multiple testing using the Graph-based clustering was performed using the Seurat22 func-
Benjamini–Hochberg method. tions FindNeighbors and FindClusters. First, the cells were embedded
in a k-nearest neighbor graph (with k = 20) based on the Euclidean
Flow cytometry distance in the principal component analysis (PCA) space. The edge
Flow cytometry was performed on PBMCs from 25 LC and 15 R individu- weights between the two cells were further modified using Jaccard simi-
als from our cohort, obtained from aliquots of specimens analyzed larity. Next, clustering was performed using the Louvain algorithm33
by CyTOF. Cells were stained with the panel shown in Supplementary implementation in the FindClusters Seurat function. Clustering with
Table 13, using Zombie UV or Zombie NIR (BioLegend) as viability 15 principal components (PCs, determined based on the location of
indicators. All cells were analyzed on a Fortessa X-20 (BD Biosciences). the elbow in the plot of variance explained by each of the top 25 PCs)
FCS files were exported into FlowJo (BD, version 10.9.0) for further and 0.1 resolution (determined using the resolution optimization
analysis. Flow cytometric data were arcsinh-scaled before analyses. In method described above for CyTOF data clustering) resulted in 11 dis-
flow cytometric experiments, SARS-CoV-2-specific CD8+ T cells were tinct biologically relevant clusters (clusters 0–11), which were used for
defined as those specifically inducing IFN-γ and/or TNF in response further analyses. Marker genes for each cluster were identified using
to SARS-CoV-2 peptide stimulation, as the CCL4 antibody exhibited the FindAllMarkers Seurat function. Marker genes were filtered to keep
background staining in flow cytometry and could not be used to define only expressed genes detected in at least 25% of the cells, with at least
SARS-CoV-2-specific T cells. 0.5 log2 fold change. Cluster annotation was performed according to
subset definitions previously established34–36. Classification markers
RNA-seq included CD19, MS4A1 and CD79A for B cells; CD3D, CD3E, CD5 and IL7R
RNA-seq was performed on PBMCs from 23 LC and 13 R individuals for CD4+ T cells; CD3D, CD3E, CD8A, CD8B and GZMK (CTL subset) for
from our cohort, obtained from aliquots of specimens analyzed by CD8+ T cells; CD14, CD68, CYBB, S100A8, S100A9, S100A12 and LYZ for
CyTOF. Samples were prepared using the AllPrep kit (Qiagen) per the monocytes; CSF2RA, LYZ, CXCL8 and CD63 for granulocytes and PF4,
manufacturer’s instructions. RNA libraries, next-generation Illumina CAVIN2, PPBP, GNG11 and CLU for platelets.
sequencing, quality control analysis, trimming and alignment were The counts-per-million reads for ALAS2 and OR7D2 were assessed
performed by Genewiz (Azenta). Briefly, following oligo dT enrichment, using edgeR37, and associations with group status were made using

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

the two-sample Welch t test, followed by multiple correction testing membership differences assumed a binomial probability distribution,
using the Holm38 procedure. For establishing associations between and those involving RNA expression differences assumed a negative
clusters and group status, GLMM implemented in the lme4 R package binomial probability distribution, but these were not formally tested.
was used. The model was performed with the family argument set to All other tests were based on the normality assumption but this was
the binomial probability distribution and with the ‘nAGQ’ parameter not formally tested.
set to 10 corresponding to the number of points per axis for evaluat-
ing the adaptive Gauss–Hermite approximation for the log-likelihood Statistics and reproducibility
estimation. Cluster membership was modeled as a response variable No statistical method was used to predetermine the sample size. Sam-
by a two-dimensional vector representing the number of cells from ples were chosen based on the availability of specimens meeting our LC
a given sample belonging or not to the cluster under consideration. criteria. No samples were excluded from the analyses. Randomization
The corresponding sample from which the cell was derived was the was not implemented as the study compared LC to R individuals. Data
random effect variable, and the group (R, LC, OR7D2high LC, or ALAS2high collection and analysis were not performed blind to the conditions of
LC) was considered the fixed variable. The log odds ratio for all pair- the experiments.
wise comparisons was estimated using the model fits provided to the
emmeans function in the emmeans R package27. The resulting P values Reporting summary
for the estimated log odds ratio and clusters were adjusted for multiple Further information on research design is available in the Nature Port-
testing using the Benjamini–Hochberg method39. For associations folio Reporting Summary linked to this article.
of gene expression with group status, raw gene counts per cell were
loaded as a SingleCellExperiment object. Cells from clusters 9 and 10 Data availability
were not included in this analysis as the median number of cells across The raw CyTOF datasets for this study corresponding to total and
samples was less than 20 per cluster. The aggregateData function in the SARS-CoV-2-specific CD4+ and CD8+ T cells are publicly accessible
muscat bioconductor package40 was used to pseudo-bulk the gene read through the following link: https://datadryad.org/stash/dataset/
counts across cells for each cluster group. Genes with raw counts less doi:10.7272/Q6WD3XTB. The raw Olink data are also downloadable
than ten in more than eight samples were removed from the analyses. through this link. The raw bulk RNA-seq and scRNA-seq data from
The pbDS function implementing the statistical methods in the edgeR this study are deposited in the Gene Expression Omnibus database—
package37 was used to assess associations of gene expression with GSE224615 (for bulk RNA-seq) and GSE235050 (for scRNA-seq).
group identity. Results from the cluster-specific pseudo-bulked gene
expression association analyses were visualized as volcano plots using References
EnhancedVolcano41,42. Select genes of interest or genes that passed a 15. Peluso, M. J. et al. Long-term SARS-CoV-2-specific immune and
multiple testing-adjusted P value threshold of 0.05 or 0.1 as indicated inflammatory responses in individuals recovering from COVID-19
were indicated in the volcano plots. For gene set enrichment analyses, with and without post-acute symptoms. Cell Rep. 36, 109518
the raw P values for each gene derived from hypothesis tests for asso- (2021).
ciations of interest were combined with a list of genes annotated with 16. Ferretti, A. P. et al. Unbiased screens show CD8+ T cells of
each of the gene sets in the biological processes domain of GO43 and COVID-19 patients recognize shared epitopes in SARS-CoV-2 that
analyzed via the simultaneous enrichment analysis method44 using largely reside outside the spike protein. Immunity 53, 1095–1107
the rSEA R package45. The family-wise error rate-adjusted P values for (2020).
cluster-specific associations of interest with each of the annotated 17. Costantini, A. et al. Effects of cryopreservation on lymphocyte
gene sets were used to identify significant associations. immunophenotype and function. J. Immunol. Methods 278,
145–155 (2003).
Olink 18. Cavrois, M. et al. Mass cytometric analysis of HIV entry,
The Olink EXPLORE 384 inflammation protein extension assay was replication, and remodeling in tissue CD4+ T cells. Cell Rep. 20,
performed per manufacturer’s protocol as published in ref. 46. 984–998 (2017).
19. Neidleman, J. et al. Phenotypic analysis of the unstimulated
Data visualization in vivo HIV CD4 T cell reservoir. eLife 9, e60933 (2020).
HOPACH30 was used to find the best cluster number. Gene expression 20. Steiner, S. et al. SARS-CoV-2 T cell response in severe and fatal
values were log-transformed and centered using the average expres- COVID-19 in primary antibody deficiency patients without specific
sion value. Clustering was performed by running the k-means algorithm humoral immunity. Front. Immunol. 13, 840126 (2022).
using the best cluster number k found, and the results were plotted 21. Roederer, M., Nozzi, J. L. & Nason, M. C. SPICE: exploration and
using the pheatmap package47. For gene network analyses, the STRING analysis of post-cytometric complex multivariate datasets.
interaction database was used to reconstruct gene networks using Cytometry A 79, 167–174 (2011).
stringApp48 for Cytoscape49. For the network, the top 50 genes or 25 22. Hao, Y. et al. Integrated analysis of multimodal single-cell data.
proteins with the lowest P values were selected from the RNA-seq data Cell 184, 3573–3587 (2021).
and Olink data, respectively. They were then subjected to stringApp 23. Korsunsky, I. et al. Fast, sensitive and accurate integration of
with an interaction score cutoff = 0.5 and the number of maximum single-cell data with Harmony. Nat. Methods 16, 1289–1296 (2019).
additional indirect interactors cutoff = 10. 24. Breiman, L. Random forests. Mach. Learn. 45, 5–32 (2001).
25. George, A. F. et al. Deep phenotypic analysis of blood
Statistical tests and lymphoid T and NK cells from HIV+ controllers and
Unless otherwise indicated, permutation tests, two-tailed unpaired ART-suppressed individuals. Front. Immunol. 13, 803417 (2022).
Student’s t tests and Welch’s t test were used for statistical analyses. 26. Bates, D., Mächler, M., Bolker, B. & Walker, S. Fitting linear
*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 and NS. Error bars mixed-effects models using lme4. J. Stat. Softw. https://doi.
corresponded to s.d. Graphs were plotted by GraphPad Prism (version org/10.48550/arXiv.1406.5823 (2015).
9.4.1). All measurements were taken from distinct samples, no sam- 27. Lenth, R., Singmann, H., Love, J., Buerkner, P. & Herve, M.
ples were measured repeatedly to generate data. Where appropriate, Emmeans: estimated marginal means, aka least-squares means.
P values were corrected for multiple testing (across three pairwise R. package version 1 https://github.com/rvlenth/emmeans
comparisons) using the Holm procedure38. Tests involving cluster (2018).

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

28. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Acknowledgements

Bioinformatics 29, 15–21 (2013). This work was supported by the Van Auken Private Foundation,
29. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold D. Henke, P. Taft and E. Taft; philanthropic funds donated to Gladstone
change and dispersion for RNA-seq data with DESeq2. Genome Institutes by the Roddenberry Foundation and individual donors
Biol. 15, 1–21 (2014). devoted to COVID-19 research; the Program for Breakthrough
30. Van der Laan, M. & Pollard, K. A new algorithm for hybrid Biomedical Research, which is partly funded by the Sandler
clustering of gene expression data with visualization and the Foundation and awards 2164 and 2208 from Fast Grants, a part
bootstrap. J. Stat. Plan Inference 117, 275–303 (2003). of Emergent Ventures at the Mercatus Center, George Mason
31. Zheng, G. X. et al. Massively parallel digital transcriptional University (to N.R.R.). We acknowledge the National Institutes
profiling of single cells. Nat. Commun. 8, 14049 (2017). of Health (NIH) DRC Center Grant P30 DK063720 and the S10
32. Ahlmann-Eltze, C. & Huber, W. glmGamPoi: fitting 1S10OD018040-01 for use of the CyTOF instrument and the NIH S10
Gamma-Poisson generalized linear models on single cell count RR028962 and the James B. Pendleton Charitable Trust for use of
data. Bioinformatics 36, 5701–5702 (2020). the Fortessa X-20. This study was also funded by the Ministerium
33. Blondel, V. D., Guillaume, J.-L., Lambiotte, R. & Lefebvre, E. Fast für Wissenschaft, Forschung und Kunst, Baden Württemberg,
unfolding of communities in large networks. J. Stat. Mech. Theory Germany (KNKC.031) and the Deutsche Forschungsgemeinschaft
Exp. 2008, P10008 (2008). (DFG; German Research Foundation)—Projektnummer 316249678—
34. McGinnis, C. S. et al. No detectable alloreactive transcriptional SFB 1279 (to J.M.). Funding from the PolyBio Research Foundation
responses under standard sample preparation conditions during supported both the experiments reported herein and the parent
donor-multiplexed single-cell RNA sequencing of peripheral cohort (LIINC); specimen and clinical data collection were also
blood mononuclear cells. BMC Biol. 19, 10 (2021). supported by NIH 3R01AI141003-03S1 and NIH R01AI158013 (to
35. Xu, C. et al. Comprehensive multi-omics single-cell data M.J.P. and T.J.H.). The funders had no role in study design, data
integration reveals greater heterogeneity in the human immune collection and analysis, decision to publish or preparation of the
system. iScience 25, 105123 (2022). manuscript.
36. Ianevski, A., Giri, A. K. & Aittokallio, T. Fully-automated and We thank S. Tamaki, V. Nguyen, P. Sanchez and C. Bispo for
ultra-fast cell-type identification using specific marker CyTOF assistance at the Parnassus Flow Core, J. Srivastava and
combinations from single-cell transcriptomic data. Nat. Commun. V. Saware for technical assistance in flow cytometry, M. Karacan
13, 1246 (2022). and N. Preising for technical assistance in peptide synthesis, E.
37. Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Ghosn for guidance on annotation of cell clusters identified by
Bioconductor package for differential expression analysis of scRNA-seq, J. Carroll for assistance on graphics, F. Chanut for
digital gene expression data. Bioinformatics 26, 139–140 (2010). editorial assistance and R. Givens for administrative assistance. We
38. Holm, S. A. A simple sequentially rejective multiple test are grateful to the study participants and their medical providers.
procedure. Scand. J. Stat. 6, 65–70 (1979). We acknowledge current and former LIINC clinical study team
39. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: members T. Abualhsan, A. Alvarez, M. Arreguin, M. Buitrago,
a practical and powerful approach to multiple testing. J. R. Stat. M. Deswal, N. DelCastillo, E. Fehrman, H. Grebe, H. Hartig, Y.
Soc. Ser. B Methodol. 57, 289–300 (1995). Hernandez, M. Kerbleski, R. Kirtikar, J. Lombardo, M. Luna, L. Ngo,
40. Crowell, H., Germain, P., Soneson, C., Sonrel, A. & Robinson, M. E. Ortiz, A. Rodriguez, J. Romero, D. Ryder, R. Sanchez, M. So,
muscat: multi-sample multi-group scRNA-seq data analysis tools. C. Song, V. Tai, A. Tang, C. Thanh, F. Ticas, L. Torres, B. Tran, D.
R package version 1.14.10. https://github.com/HelenaLC/muscat Varma and M. Williams. We also acknowledge LIINC laboratory
(2023). team members A. Buck, J. Donatelli, J. Hakim, N. Iyer, O. Janson, B.
41. Blighe, K., Rana, S. & Lewis, M. Publication-ready volcano plots LaFranchi, C. Nixon, I. Thomas and K. Turcios. We thank J. Chen, A.
with enhanced colouring and labeling. R package version 1.18.10. Donovan and C. Forman for assistance with data entry and review.
https://github.com/kevinblighe/EnhancedVolcano (2020). We thank the UCSF AIDS Specimen Bank for processing specimens
42. Huber, W. et al. Orchestrating high-throughput genomic analysis and maintaining the LIINC biospecimen repository. We are grateful
with Bioconductor. Nat. Methods 12, 115–121 (2015). to E. Eilkhani and M. Deswal for regulatory support. We are also
43. The Gene Ontology Consortium The gene ontology resource: 20 grateful for the contributions of additional current and former LIINC
years and still going strong. Nucleic Acids Res 47, D330–D338 leadership team members—M. Durstenfeld, P. Hsue, B. Greenhouse,
(2019). I. Rodriguez-Barraquer and R. Rutishauser.
44. Ebrahimpoor, M., Spitali, P., Hettne, K., Tsonaka, R. & Goeman,
J. Simultaneous enrichment analysis of all possible gene-sets: Author contributions
unifying self-contained and competitive methods. Brief. K.Y. designed the experiments, performed CyTOF, flow cytometry
Bioinform. 21, 1302–1312 (2020). and scRNA-seq experiments, conducted analyses, and prepared
45. Ebrahimpoor, M. rSEA: simultaneous enrichment analysis. R figures and tables. M.J.P. designed the LIINC cohort, oversaw LIINC
package version 2.1.1. CRAN.R-project.org/package=rSEA cohort procedures and interpreted data. X.L. developed pipelines for
(2020). data analyses and performed scRNA-seq. R.T. performed clustering
46. Assarsson, E. et al. Homogenous 96-plex PEA immunoassay and scRNA-seq analyses. M.S. performed RNA-seq and Olink
exhibiting high sensitivity, specificity, and excellent scalability. analyses. J.N. prepared peptides and helped with experiments. A.A.
PLoS ONE 9, e95192 (2014). and K.C.Y. prepared and analyzed CyTOF specimens. T.M. designed
47. Kolde, R. pheatmap: Pretty Heatmaps. R package version 1.0.12. protocols for CyTOF analyses. R.H., K.A. and B.H. managed the LIINC
https://github.com/raivokolde/pheatmap (2018). cohort, recruited participants, collected clinical data and collected
48. Doncheva, N. T., Morris, J. H., Gorodkin, J. & Jensen, L. J. biospecimens. U.A., M.L., D.V., K.A., T.D. and S.E.M. recruited LIINC
Cytoscape StringApp: network analysis and visualization of participants, collected clinical data and collected biospecimens. T.D.
proteomics data. J. Proteome Res. 18, 623–632 (2018). and S.E.M. processed specimens. L.S. and J.M. synthesized peptides.
49. Shannon, P. et al. Cytoscape: a software environment for R.I. entered, cleaned and performed quality control on LIINC data.
integrated models of biomolecular interaction networks. Genome S.L. and S.A.G. managed LIINC data and selected biospecimens.
Res. 13, 2498–2504 (2003). K.L.L. performed antibody assays. S.A.L. designed the CHIRP cohort

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

and oversaw CHIRP cohort procedures. J.D.K. and J.N.M. designed Additional information
the LIINC cohort and interpreted LIINC clinical data. S.G.D. designed Extended data is available for this paper at
the LIINC cohort, oversaw cohort procedures and interpreted LIINC https://doi.org/10.1038/s41590-023-01724-6.
clinical data. T.J.H. designed the LIINC cohort, oversaw cohort
procedures, performed the RNA-seq and Olink studies, interpreted Supplementary information The online version contains supplementary
data and prepared figures. N.R.R. conceived the study, performed material available at https://doi.org/10.1038/s41590-023-01724-6.
supervision, conducted data analyses and prepared figures and
tables. K.Y., M.J.P., T.J.H. and N.R.R. wrote the manuscript. All authors Correspondence and requests for materials should be addressed to
have read and approved this manuscript. Timothy J. Henrich or Nadia R. Roan.

Competing interests Peer review information Nature Immunology thanks Christina Zielinski
M.J.P. reports consulting fees from Gilead Sciences and AstraZeneca, and the other, anonymous, reviewer(s) for their contribution to the peer
outside the submitted work. S.G.D. reports grants and/or personal fees review of this work. Primary Handling Editor: Ioana Staicu, in collaboration
from Gilead Sciences, Merck & Co., Viiv, AbbVie, Eli Lilly, ByroLogyx with the Nature Immunology team. Peer reviewer reports are available.
and Enochian Biosciences, outside the submitted work. T.J.H. receives
grant support from Merck and consults for Roche. All other authors Reprints and permissions information is available at
report no conflicts of interest. www.nature.com/reprints.

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

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

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

Extended Data Fig. 1 | Cohort characteristics, study design, and subset individuals, with n = 27 LC and n = 16R. f. Schematic of experimental design and
identification. a–c, Number of sequelae symptoms at 4 (M4) and 8 (M8) data analyses. Blood specimens from 27 LC and 16 R individuals were subjected
months post-infection (n = 27 LC, n = 16 R) (a), and the numbers of individuals to Olink, serology, CyTOF, and RNA-seq and scRNA-seq analysis. The indicated
that were male or female (b) and that were hospitalized at the time of acute tools on the right were then used for analyses of the resulting high-dimensional
COVID-19 infection (c), in LC and R study participants. *p < 0.05 (two-sided paired datasets. g,h, Gating strategy to identify T cell populations. Intact, live, singlet
sample t-test). d, The numbers of indicated co-morbidities in LC vs R study cells from baseline (g) or SARS-CoV-2 peptide-treated (h) samples were gated
participants. e, BMI in LC vs R study participants. *p < 0.05 (two-sided student’s for CD3+ T cells followed by sub-gating on CD4+ and CD8+ T cells as indicated.
t-test). Horizontal bars indicate mean, error bars indicate SD, and dots represent i,j, Gating strategy to define classical CD4+ (i) and CD8+ (j) T cell subsets.

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

a
0 0 0 0 0

No peptides

0.021 0.13 0.01 0 0

4
10

3
10

+SARS-CoV-2 peptides
2
10

1
IFN-γ

10

TNF

IL-2

IL-4

IL-6
0

1 2 3 4
0 10 10 10 10

CD4 CD4 CD4 CD4 CD4

0 0 3.99 1.41

No peptides

10
4
0 0 4.27 1.32
3

Granzyme B
10

+SARS-CoV-2 peptides
10

Perforin
1
CCL4

10
IL-17

1 2 3 4
0 10 10 10 10

CD4 CD4 CD4 CD4

b
0 0 0 0 0

No peptides

10
4

0.055 0.047 0 0 0
3
10

2
10

+SARS-CoV-2 peptides
IFN-γ

1
10
TNF

IL-2

IL-4

IL-6
0

1 2 3 4
0 10 10 10 10

CD8 CD8 CD8 CD8 CD8

0 0 47.3 5.97

No peptides

0 0.025 50.8 6.44

4
10
Granzyme B

3
10

2
10

+SARS-CoV-2 peptides
Perforin

1
CCL4

10
IL-17

1 2 3 4
0 10 10 10 10

CD8 CD8 CD8 CD8

c IFN-γ+ cells IL-2+ cells TNF+ cells e

4
10

0.15 0.15 1.5 R 0 per million 45 per million

Total CD4+ T cells (%)

Total CD4+ T cells (%)

Total CD4+ T cells (%)

NS
10

NS NS LC
2
10

0.10 0.10 1.0

1
10
IL-6

0.05 0.05 0.5

1 2 3 4
0 10 10 10 10

CD4
0.00 0.00 0.0 No peptides +SARS-CoV-2 peptides

d IFN-γ+ cells CCL4+ cells TNF+ cells f IL-6+ cells

Total CD4 T cells (per M)

R 80 R
1.0 1.0 1.0
Total CD8+ T cells (%)

Total CD8+ T cells (%)

Total CD8+ T cells (%)

NS LC * LC
0.8 60
NS
NS 0.6
0.5 0.5 40
0.4
+

20
0.2
0.0 0.0 0
0.0

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

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

Extended Data Fig. 2 | Cytokine and effector molecule expression in SARS- observed in LC individuals. e, CD4+ T cells from representative donor, stimulated
CoV-2-specific T cells. a,b CD4+ (a) or CD8+ (b) T cells from representative donor, (right) or not (left) with SARS-CoV-2 spike and T-scan peptides (Methods). f, The
stimulated (bottom) or not (top) with SARS-CoV-2 spike and T-scan peptides percentages of SARS-CoV-2-specific CD4+ T cells inducing IL-6 in response to
(Methods). Red boxes highlight the cytokines used to define the SARS-CoV-2- SARS-CoV-2 peptide stimulations. *p < 0.05 (two-sided Welch’s t-test). Horizontal
specific T cells. c,d The percentages of SARS-CoV-2-specific CD4+ (c) and CD8+ (d) bars indicate mean, error bars indicate SD, and dots represent individuals, with
T cells as defined by induction of IFN-γ, IL-2, CCL4, or TNF in response to SARS- n = 27 LC and n = 16 R (c, d, f).
CoV-2 peptide stimulations (two-sided student’s t-test). e,f, IL-6+ CD4+ T cells are

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

a
TN cells TSCM cells TCM cells TEM cells TTM cells TEMRA cells
80 0.8 20 40 30 NS 50 R
NS NS p=0.09 NS
Total CD8+ T cells (%)

Total CD8+ T cells (%)

Total CD8+ T cells (%)

Total CD8+ T cells (%)

Total CD8+ T cells (%)

Total CD8+ T cells (%)

NS LC
40
60 0.6 15 30
20
30
40 0.4 10 20
20
10
20 0.2 5 10
10

0 0.0 0 0 0 0

b
TN cells TSCM cells TCM cells TEM cells TTM cells TEMRA cells
60 4 50 40 50 80 R
NS NS NS p=0.05
SARS-CoV-2-specific

SARS-CoV-2-specific

SARS-CoV-2-specific

SARS-CoV-2-specific

SARS-CoV-2-specific

SARS-CoV-2-specific
NS NS LC
CD8+ T cells (%)

CD8+ T cells (%)

CD8+ T cells (%)

CD8+ T cells (%)

CD8+ T cells (%)

CD8+ T cells (%)

3 40 30 40 60
40
2 30 30
20 40
1 20 20
20 10 20
0 10 10
0 0
0 0 0

Extended Data Fig. 3 | Subset distribution of total and SARS-CoV-2-specific R individuals (two-sided student’s t-test). b, Frequencies of TN cells, TSCM cells, TCM
CD8+ T cells among LC and R individuals. a, Frequencies of TN cells, TSCM cells, cells, TEM cells, TTM cells, and TEMRA cells among SARS-CoV-2-specific CD8+ T cells
TCM cells, TEM cells, TTM cells, and TEMRA cells among total CD8+ T cells from LC and from LC and R individuals (two-sided student’s t-test).

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

Extended Data Fig. 4 | MSI of CyTOF phenotyping markers among total CD4+ individuals for any of the antigens (two-sided t-test with multiple correction by
and CD8+ T cells from LC and R individuals. Antigens are shown in the order Sidak adjustment). Box plots represent the median (middle bar), 75% quartile
listed in Supplementary Table 4. Results are gated on live, singlet CD4+ (a) or (upper hinge) and 25% (lower hinge) with whiskers extending 1.5× interquartile
CD8+ (b) T cells. No significant differences were observed between LC and R range, dots represent individuals with n = 27 LC and n = 16 R.

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

Extended Data Fig. 5 | MSI of CyTOF phenotyping markers among SARS-CoV- R individuals for any of the antigens (two-sided t-test with multiple correction
2-specific CD4+ and CD8+ T cells from LC and R individuals. Results are similar by Sidak adjustment). Box plots represent the median (middle bar), 75% quartile
to that shown in Extended Data Fig. 4, but gated on SARS-CoV-2-specific CD4+ (upper hinge) and 25% (lower hinge) with whiskers extending 1.5× interquartile
(a) or CD8+ (b) T cells. No significant differences were observed between LC and range, dots represent individuals with n = 27 LC and n = 16 R.

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

Extended Data Fig. 6 | Activated T cells are not more abundant in individuals expressing acute activation markers CD38, HLA-DR, and/or Ki67 in LC and R
with LC. The percentages of total CD4+ T cells (a), total CD8+ T cells (b), individuals (two-sided student’s t-tests). Horizontal bars indicate mean, error
SARS-CoV-2-specific CD4+ T cells (c), and SARS-CoV-2-specific CD8+ T cells (d) bars indicate SD, and dots represent individuals, with n = 27 LC and n = 16 R.

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

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

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

Extended Data Fig. 7 | Sex-dimorphic T cell cluster distribution in individuals cell clusters B1 and B2 in male and female individuals, with or without LC. Two-
with LC. a, Cluster distribution among total CD4+ T cells as depicted by UMAP. sided p-values were derived from a GLMM fit (see Methods). Individual points
b, The distributions of CD4+ T cell clusters A1 and A4 in male and female represent individuals, with n = 10 LC and n = 9 R in the male group and n = 17 LC
individuals, with or without LC. Two-sided p-values were derived from a GLMM and n = 7 R in the female group, and where the value corresponds to % of cells
fit (see Methods). Individual points represent individuals, with n = 10 LC and n = 9 belonging to clusters B1 or B2. g, Expression levels of differentiation markers
R in the male group and n = 17 LC and n = 7 R in the female group, and where the (CD45RA, CD45RO, CD27), activation markers (HLA-DR, OX40), tissue homing
value corresponds to % of cells belonging to clusters A1 or A4. c, Expression levels receptors (CD29, CXCR4), and lymph node homing receptors (CD62L, CCR7) on
of differentiation markers (CD45RA, CD45RO, CD27), activation markers (HLA- CD8+ T cell cluster B1 compared to total baseline CD8+ T cells. h, Expression levels
DR, OX40), tissue homing receptors (CD29, CXCR4), and lymph node homing of differentiation markers (CD45RA, CD45RO, CD27, CD57), cytolytic effectors
receptors (CD62L, CCR7) on CD4+ T cell cluster A1 compared to total baseline (perforin, granzyme B), tissue homing receptors (CD29, CXCR4, CCR5), and
CD4+ T cells. d, Expression levels of differentiation markers (CD45RA, CD45RO, lymph node homing receptors (CD62L, CCR7) on CD8+ T cell cluster B2 compared
CD27, CD57), cytolytic effectors (perforin, granzyme B), tissue homing receptors to total baseline CD8+ T cells. ****p < 0.0001 (two-sided paired t-test, c,d,g,h).
(CD29, CXCR4, CCR5), and lymph node homing receptors (CD62L, CCR7) on CD4+ Horizontal bars indicate mean, error bars indicate SD, and dots represent
T cell cluster A4 compared to total baseline CD4+ T cells. e, Cluster distribution individuals, with n = 27 LC and n = 16 R (b–d,f–h).
among total CD8+ T cells as depicted by UMAP. f, The distributions of CD8+ T

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

a
CXCR4 CXCR5 CCR6
80 r=0.4638 80 r=0.4109 150 r=0.3183
p=0.0026 p=0.0084 p=0.0453
60 60
MFI (x102)

100

MFI

MFI
40 40
50
20 20

0 0 0
2 3 4 5 0.0 0.1 0.2 0.3 0.4 0.5 0 1 2 3

CXCR4-APC
CXCR4+CXCR5+
4
10.8
MSI MSI MSI
10

cells
3

b
10

Cells Singlets Live cells CD4+ T cells

250K 250K 250K

Memory
5 5
10 10 3 4
0 10 10

96.9 CXCR5-BV421

CD3-FITC

CD3-FITC
200K 200K 200K

15.5 T cells
SSC-A

FSC-H

FSC-A
4 4
10 10

CXCR5+CCR6+
150K 150K

81.4
150K

CXCR5-BV421
54.8 10
3

3.63 cells
3 3
100K 100K 100K 10 10

50K 50K 50K

0 0
40.5
3 3
-10 -10
0 0 0 2
10
0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 3 3 4 5 4 5 6 3 3 4 5
-10 0 10 10 10 0 10 10 10 -10 0 10 10 10

FSC-A FSC-A Zombie UV CD4-BV750 CD45RO-APC/Cy7

0

2
-10

2 2 3
-10 0 10 10

CCR6-BV605
c CXCR4+CXCR5+ cells CXCR5+CCR6+ cells CXCR4+CCR6+ cells
15 4 15 * R 9.2
Total CD4+ T cells (%)

Total CD4+ T cells (%)

Total CD4+ T cells (%)

CXCR4+CCR6+

CXCR4-APC
4
10

* * LC cells
3 10
3

10 10
2 0

5 5
3 4
0 10 10

1 CCR6-BV605

0 0 0

d PD1+CTLA4+ cells IFNγ+TNF+ cells

e
R 2.0 1.5 R
1.0
SARS-CoV-2-specific

Total CD8+ T cells (%)

NS
(%o f total CD4+ T cells)

(%o f total CD4+ T cells)

CXCR4+CXCR5+ T cells

* LC CXCR5+CCR6+ T cells LC
CD8+ T cells (%)

100 0.8 1.5

R=0.4662 1.0
0.6 R=-0.6594 R=0.4395
1.0 R=-0.6108 p=0.0188
p=0.0279
p=0.0075
50 0.4 p=0.0156 0.5
0.5
0.2
0.0 0.0
0 0.0
7 8 9 10 11 12 13 7 8 9 10 11 12 13
IL-4 (Normalized Counts) IL-4 (Normalized Counts)

Extended Data Fig. 8 | Flow cytometric validation and association analyses. a, percentages of cells dually expressing PD1 and CTLA4 among SARS-CoV-2-
Association of flow cytometric (mean fluorescence intensity, MFI) vs CyTOF (MSI) specific CD8+ (left) or cells dually expressing IFN-γ and TNF among total CD8+
expression levels of CXCR4, CXCR5, and CCR6. Data were analyzed by Pearson T cells (right), as determined by flow cytometry. *p < 0.05 (two-sided student’s
correlation coefficient and two-tailed unpaired t-tests. b, Flow cytometric gating t-test). Horizontal bars indicate mean, error bars indicate SD, and dots represent
strategy to identify memory CD4+ T cells expressing various combinations individuals, with n = 25 LC and n = 15 R (c,d). e, Associations of percentages of
of CXCR4, CXCR5, and CCR6. c, The percentages of CXCR4+CXCR5+CD4+, CXC4+CXCR5+CD4+ T cells or CXCR5+CCR6+CD4+ T cells with IL-4 levels in LC vs R
CXCR5+CCR6+CD4+, and CXCR4+CCR6+CD4+ T cells in LC vs R individuals as individuals. Data were analyzed by Pearson correlation coefficient and two-tailed
determined by flow cytometry. *p < 0.05 (two-sided student’s t-test). d, The unpaired t-tests.

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

Extended Data Fig. 9 | See next page for caption.

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

Extended Data Fig. 9 | scRNAseq analysis reveals OR7D2 and ALAS2 with LC, depicted as mean % of cells that were positive for OR7D2 or ALAS2 reads.
expression in multiple subsets, validates tissue-homing chemokine receptor f, Volcano plots showing LC vs R individuals for scRNA-seq-identified CD4+ T cell
expression among LC CD4+ T cells, and identifies DEGs among subsets in clusters 0 and 7, depicting CXCR4, CXCR5, and CCR6. g,h, Volcano plots depicting
LC individuals. a,b, UMAP of cells analyzed by scRNA-seq among LC (n = 8) vs scRNA-seq-defined clusters 0, 1, 5, 7, and 8 for OR7D2high vs. R (g), or clusters 1, 5,
R (n = 4) individuals (a), and among the LC individuals classified as OR7D2high 6, 7, and 8 for ALAS2high vs. R (h) individuals. DEGs with p < 0.05 (as determined
(n = 4) vs. ALAS2high (n = 4) (b). c, OR7D2 and ALAS2 expression in the OR7D2high LC, empirical Bayes quasi-likelihood F-tests, with Benjamini-Hochberg correction)
ALAS2high LC, and R individuals. **p < 0.01 (two-sided Welch two-sample t-test). are labeled. Genes preferentially expressed in LC individuals are depicted on the
Box plots represent the median (middle bar), 75% quartile (upper hinge) and 25% right, and those preferentially expressed in R individuals on the left. The x-axes
(lower hinge) with whiskers extending 1.5× interquartile range, dots represent represent the log2(fold-change) of the mean expression of each gene between the
individuals with n = 8 LC and n = 4R. d, UMAP depictions of cells expressing (blue) comparison groups, and the y-axes represent the raw –log10(p-values). Dashed
or not expressing (grey) OR7D2 or ALAS2 in individuals with LC. e, OR7D2 and horizontal lines delineate the thresholds corresponding to Benjamini-Hochberg
ALAS2 expression in scRNA-seq-identified clusters labeled in Fig. 5d in individuals adjusted two-tailed p-values of <0.05 (Methods).

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

Extended Data Fig. 10 | See next page for caption.

Nature Immunology
Letter https://doi.org/10.1038/s41590-023-01724-6

Extended Data Fig. 10 | Validation of CyTOF antibodies. a, CyTOF analysis and CD4+ TN cells, as assessed by CyTOF. d, Expression of CD30 and Ki67 among
of human lymphoid aggregate cultures generated from tonsils depicting CD3+ CD3+ CD45RO+CD45RA−CD4+ T memory (TM) cells and CD4+ TN cells, as assessed
T cells on the top and CD3− B cells on the bottom as indicated, analogous to by CyTOF. ****p < 0.0001 (two-sided paired t-test). e, Illustration of pTFH gate
methods previously described18. b, CyTOF analysis of PMA/ionomycin- or implemented on PBMC samples, and TFH gate implemented on tonsil samples.
LPS-stimulated PBMCs, depicting CD3+ T cells on the top and CD3− cells on the Cells were pre-gated on CD4+ TM cells.
bottom, similar to prior studies2–5. c, Expression of Foxp3 among CD4+ Treg cells

Nature Immunology