CXCL13 is a plasma biomarker of germinalcenter activityColin Havenar-Daughtona,b, Madelene Lindqvistb,c, Antje Heitb,d, Jennifer E. Wua, Samantha M. Reissa, Kayla Kendrica,Simon Bélangera, Sudhir Pai Kasturie,f, Elise Landaisg, Rama S. Akondyf, Helen M. McGuireh, Marcella Bothwelli,j,Parsia A. Vagefik, Eileen Scullyc, IAVI Protocol C Principal Investigatorsl,1, Georgia D. Tomarasm, Mark M. Davish,n,Pascal Poignardg,o, Rafi Ahmedb,f, Bruce D. Walkerb,c,p, Bali Pulendranb,e,f, M. Juliana McElrathb,d,Daniel E. Kaufmannb,c,q,r, and Shane Crottya,b,2

aDivision of Vaccine Discovery, La Jolla Institute for Allergy and Immunology, La Jolla, CA 92037; bCenter for HIV/AIDS Vaccine Immunology and ImmunogenDiscovery, La Jolla, CA 92037; cRagon Institute of Massachusetts General Hospital, Massachusetts Institute of Technology and Harvard University, Boston, MA02114; dVaccine and Infectious Disease Division, Fred Hutchinson Cancer Research Center, Seattle, WA 98109; eYerkes National Primate Research Center, EmoryUniversity, Atlanta, GA 30322; fEmory Vaccine Center, Emory University School of Medicine, Atlanta, GA 30322; gNeutralizing Antibody Center, InternationalAIDS Vaccine Initiative, La Jolla, CA 92037; hDepartment of Microbiology and Immunology, Stanford University School of Medicine, Stanford, CA 94304;iDepartment of Surgery, University of California, San Diego, CA 92123; jPediatric Otolaryngology, Rady Children’s Hospital–San Diego, San Diego, CA 92123;kDepartment of Surgery, Massachusetts General Hospital, Boston, MA 02114; lInternational AIDS Vaccine Initiative, New York, NY 10038; mDuke Human VaccineInstitute, Duke University School of Medicine, Durham, NC 27710; nHoward Hughes Medical Institute, Stanford University School of Medicine, Stanford, CA 94304;oDepartment of Immunology and Microbial Science, The Scripps Research Institute, La Jolla, CA 92037; pHoward Hughes Medical Institute, Chevy Chase, MD 20815;qDepartment of Medicine, Université de Montréal, Montreal, QC H2X 0A9, Canada; and rCentre de Recherche du Centre Hospitalier de l’Université de Montréal(CRCHUM), Montreal, QC H2X 0A9, Canada

Edited by Rino Rappuoli, GSK Vaccines, Siena, Italy, and approved January 21, 2016 (received for review October 9, 2015)

Significantly higher levels of plasma CXCL13 [chemokine (C-X-C motif)ligand 13] were associated with the generation of broadly neutral-izing antibodies (bnAbs) against HIV in a large longitudinal cohortof HIV-infected individuals. Germinal centers (GCs) perform theremarkable task of optimizing B-cell Ab responses. GCs arerequired for almost all B-cell receptor affinity maturation andwill be a critical parameter to monitor if HIV bnAbs are to beinduced by vaccination. However, lymphoid tissue is rarely avail-able from immunized humans, making the monitoring of GC activityby direct assessment of GC B cells and germinal center CD4+ T fol-licular helper (GC Tfh) cells problematic. The CXCL13–CXCR5 [chemo-kine (C-X-C motif) receptor 5] chemokine axis plays a central role inorganizing both B-cell follicles and GCs. Because GC Tfh cells canproduce CXCL13, we explored the potential use of CXCL13 as ablood biomarker to indicate GC activity. In a series of studies, wefound that plasma CXCL13 levels correlated with GC activity indraining lymph nodes of immunized mice, immunized macaques,and HIV-infected humans. Furthermore, plasma CXCL13 levels inimmunized humans correlated with the magnitude of Ab responsesand the frequency of ICOS+ (inducible T-cell costimulator) Tfh-likecells in blood. Together, these findings support the potential use ofCXCL13 as a plasma biomarker of GC activity in human vaccine trialsand other clinical settings.

vaccines | CXCL13 | HIV | antibodies | Tfh

The germinal center (GC) reaction is a critical immunologicalprocess that occurs in draining lymph nodes after immuni-

zation (1). The GC response consists of antigen-specific B cellsundergoing affinity maturation through a process of somatichypermutation (SHM) of the B-cell receptor. SHM is necessaryfor producing high-affinity Ab responses after immunizations andinfections. Influenza neutralizing Abs have substantial SHM.Particularly high levels of SHM, 15–30% amino acid mutation(2, 3), are present and necessary for broad Ab neutralization ofdiverse HIV strains (4, 5). Therefore, as candidate influenza andHIV vaccines are evaluated for the ability to induce broadly neu-tralizing antibodies (bnAbs), the quantitation and functional char-acterization of GC responses will be a key parameter for study.Serological analysis of vaccine-specific Ab titers provides importantinformation, but those data are limited. Serological outcomes aremeasured at time points long after initial immunizations. Neutral-izing Ab responses are commonly only measurable after multipleboosts. Those outcomes likely depend on GC activity and affinity

maturation at much earlier time points. Several state of the artHIV vaccine strategies rely on long, multistage immunization pro-tocols (6, 7). With bnAb responses as the goal, means of earlyanalysis of the immune response will be essential to understand andimprove on vaccination schemes that may end in failure or onlypartial success. One critical parameter to assess will be the ability ofeach immunization to generate GC responses.

Central to the GC reaction and SHM is the interaction of GC Bcells with germinal center T follicular helper (GC Tfh) cells (8). GCTfh cells are both required and limiting for the GC reaction (9, 10).GC Tfh cells control the number of GC B-cell divisions and there-fore, the amount of SHM by individual GC B-cell clones (11). Cur-rently, the preferred means of quantifying the GC response is the cellularenumeration and analysis of GC Tfh and GC B cells (8). However, inhuman and nonhuman primate (NHP) vaccination studies, directanalysis of draining lymph node tissues is impractical or undesirable

Significance

A major challenge for vaccine science is that there is no way tomeasure germinal center activity in humans. This challenge isparticularly acute for human clinical trials of candidate vaccines(and most nonhuman primate studies of candidate vaccines), be-cause germinal centers are the engines of Ab affinity maturation,and generation of highly affinity-matured Ab responses is thegoal of all Ab-eliciting vaccines. Here, we report that we haveidentified the chemokine CXCL13 [chemokine (C-X-C motif) ligand13] as a biomarker of germinal center activity. We show explicitrelationships between plasma CXCL13 concentrations and ger-minal center frequencies in lymph nodes in a series of differentconditions, including licensed and experimental vaccines, and inhumans, nonhuman primates, and mice.

Author contributions: C.H.-D. and S.C. designed research; C.H.-D., M.L., A.H., J.W., S.M.R.,K.K., S.B., E.L., R.S.A., and H.M.M. performed research; S.P.K., E.L., R.S.A., H.M.M., M.B.,P.A.V., E.S., I.P.C.P.I., G.D.T., M.M.D., P.P., R.A., B.D.W., B.P., M.J.M., and D.E.K. contributednew reagents/analytic tools; C.H.-D., M.L., A.H., J.W., S.M.R., K.K., S.B., S.P.K., E.L., R.S.A.,H.M.M., G.D.T., M.M.D., P.P., R.A., B.D.W., B.P., M.J.M., D.E.K., and S.C. analyzed data; andC.H.-D. and S.C. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1A complete list of the International AIDS Vaccine Initiative (IAVI) Protocol C PrincipalInvestigators is in SI Text.

2To whom correspondence should be addressed. Email: shane@lji.org.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1520112113/-/DCSupplemental.

2702–2707 | PNAS | March 8, 2016 | vol. 113 | no. 10 www.pnas.org/cgi/doi/10.1073/pnas.1520112113

for fear of removing the primary site of the ongoing immune response.Therefore, identification of a plasma biomarker for GC activity wouldbe of great value in immunization studies as well as have utility in anumber of other biomedically relevant contexts.

The CXCL13 [chemokine (C-X-C motif) ligand 13]–CXCR5[chemokine (C-X-C motif) receptor 5] chemokine axis plays a majorrole in organizing both B-cell follicles and GCs (12, 13). CXCL13 isexpressed by both follicular dendritic cells (14) and GC Tfh cells (15,16) in the B-cell follicles. B- and Tfh-cell expressions of CXCR5, thereceptor for CXCL13, are necessary for migration to the B-cellfollicle. Although CXCL13 acts locally, it can also be detected inhuman plasma in the steady state, and perturbations in plasmaCXCL13 have been associated with immune activity (17–21). BecauseGC Tfh cells regulate the size of the GC response and can be majorproducers of CXCL13, we explored whether plasma CXCL13 changesmay reflect lymphoid tissue GC activity.

ResultsPlasma CXCL13 Is Elevated in bnAb+ HIV-Infected Individuals. Thehigh levels of SHM seen in bnAbs generated against HIV and theassociation of circulating memory Tfh cells with the generation ofbnAbs (22) suggest that individuals able to generate HIV bnAbs mayhave superior GC responses (6). Because directly measuring thesetissue-resident GC responses in humans is not generally feasible, weasked whether CXCL13 was higher in the plasma of individuals ableto generate HIV bnAbs than in the plasma of individuals who werenot. We tested plasma samples from ART (antiretroviral therapy)-naive HIV+ individuals enrolled in International AIDS VaccineInitiative (IAVI) Protocol C, a large longitudinal cohort of HIV+

individuals monitored early after infection and followed every 3–6mo. This cohort has been extensively characterized for the ability ofeach individual to produce neutralizing Abs against HIV (22, 23). Aneutralization score was calculated based on both the breadth andthe potency of the neutralizing Ab present in individual samples foreach time point after infection; 15% of 228 individuals followedbeyond 4 y postinfection were found to have a neutralization score ofgreater than one and were termed top neutralizers (Fig. 1A). Topneutralizers had Ab responses capable of neutralizing an average of73% (27 of 37) of HIV pseudoviruses using a principally tier 2 viruspanel. Most HIV-infected individuals had Ab responses that scoredless than 0.5, neutralizing an average of 27% (10 of 37) of HIVstrains; these individuals were termed low neutralizers. Plasma fromeach individual was tested for neutralizing Abs and CXCL13 con-centration at both the earliest time point available after HIV in-fection (∼4 mo postinfection) and the time of bnAb development(∼40 mo postinfection). Broad neutralization at the later time pointwas associated with elevated concentrations of plasma CXCL13 at boththe ∼4-mo (top neutralizers median: 92.7 pg/mL vs. low neutralizersmedian: 31.3 pg/mL; P = 0.012) and the ∼40-mo (top neutralizers me-dian: 78.9 pg/mL vs. low neutralizers median: 32.2 pg/mL; P = 0.008)time points postinfection (Fig. 1 B and C). A positive correlation be-tween Ab breadth and CXCL13 was also observed in an independentcohort containing a small number (five) of individuals with neutralizingbreadth (24). The generation of HIV neutralizing Abs is also positivelycorrelated with HIV viral load (23, 25), and viral load could affectplasma CXCL13 levels (24, 26). We, therefore, asked whether viral loadand CXCL13 were independent variables. The difference in plasmaCXCL13 between top and low neutralizers remained significant at the∼40-mo time point and continued to show a strong trend at the ∼4-motime point. Multivariate analysis determined that plasma CXCL13and viral load are largely independent factors [early time pointANCOVA (analysis of covariance), P = 0.021; bnAb development timepoint ANCOVA, P = 0.066]. Therefore, elevated plasma CXCL13 intop HIV neutralizers suggested that these individuals may have strongerGC responses.

Plasma CXCL13 Is Correlated with Lymphoid Tissue GCs in Humans.GC Tfh cells are a producer of CXCL13 in secondary lymphoid tissue,such as tonsil (15, 16, 27). We found that GC Tfh cells were uniquelyable to produce CXCL13 when analyzed by intracellular FACS anal-ysis (Fig. 2A). Other cell types have been reported to produce CXCL13

under various inflammatory conditions (26, 28). On examination,CXCL13 protein was not produced in the other cell types present intonsil cell preparations (Fig. 2B) or monocytes in peripheral bloodmononuclear cells (PBMCs) (Fig. S1 A and B). Similar results wereobtained from human spleen and lymph node tissues, showingCXCL13 expression to be restricted to GC Tfh cells (Fig. 2C).

In an additional cohort of HIV+ and HIV− individuals at Mas-sachusetts General Hospital, we obtained lymph node biopsies, allowingus to directly compare plasma CXCL13 to GC activity in human lym-phoid tissue. GC Tfh cells (CXCR5hi PD-1hi) were identified (Fig. 2D)and quantified. In 14 matched plasma and lymph node samples, plasmaCXCL13 concentration positively correlated with GC Tfh cells in thelymph node (r = 0.75; P = 0.003) (Fig. 2E). Furthermore, plasmaCXCL13 also correlated with GC B cells (r = 0.62; P = 0.02) (Fig. S2A).The strong correlation observed between plasma CXCL13 and lymphnode GC Tfh cells within a relatively small human donor set togetherwith the ability of GC Tfh cells to produce CXCL13 suggest thatCXCL13 can be a biomarker of GC activity.

Plasma CXCL13 Is Correlated with Lymph Node GCs in Mice AfterImmunization. We next examined the relationship between plasmaCXCL13 and GC activity after protein immunization in animal vacci-nation models (6, 29–31). We first immunized mice with 4-hydroxy-3-nitrophenyl acetyl haptenated ovalbumin in aluminum hydroxideadjuvant (alum + NP-OVA). One-half of the immunized mice re-ceived OT-II TCR (T-cell receptor) transgenic CD4 T cells to en-hance the antigen-specific and GC Tfh CD4 T-cell response. Sevendays after immunization, plasma CXCL13 was increased comparedwith that in unimmunized mice (Fig. 3A). Plasma CXCL13 was alsoincreased 7 d after acute infection with LCMV (lymphocytic chorio-meningitis virus) Armstrong or vaccinia virus (Fig. 3A). In alum + NP-OVA immunized mice, GC Tfh cells were quantified in the draininglymph node (CXCR5hi PD-1hi) (Fig. 3B), and plasma CXCL13 corre-lated with the frequency of vaccine-induced GC Tfh cells (Fig. 3C) (r =0.82; P = 0.002) and GC B cells (r = 0.74; P = 0.008) (Fig. S2B).

In a larger study of immunized mice that did not receive transgenicCD4 T cells, keyhole limpet hemocyanin plus aluminum hydroxide(KLH + alum) immunized mice had higher plasma concentrations ofCXCL13 after immunization compared with those at the preimmuni-zation time point in the same mice, peaking at 2 wk postimmunization(Fig. 3D). The mice were given a booster immunization with KLH +alum 50 d after the primary immunization. Plasma CXCL13 concen-trations were increased at both 10 and 18 d postboost (Fig. 3E) andagain, correlated with the frequency of GC Tfh cells in the draininglymph node (r = 0.69; P = 0.023) (Fig. 3F). Thus, in small animalimmunization models, plasma CXCL13 concentrations were a positivebiomarker of GC activity and GC Tfh cells.

0

100

200

300

400

LOD

CXC

L13

(pg/

ml)

0.012

CXC

L13

(pg/

ml)

>1.0 <0.5Neut. Score

0.008800

300400

200

100

0LOD

>1.0 <0.5Neut. Score

0-0.5

1-3

0.5-1

IAVI Protocol C

CBANeutralization Score

(n=228)

Fig. 1. Plasma CXCL13 concentration is associated with HIV bnAb develop-ment. (A) HIV neutralizers grouped by neutralization score for IAVI Protocol C.The neutralization score is derived from both breadth and potency data from a6- or 37-virus cross-clade pseudovirus panel. (B) Plasma CXCL13 from top(neutralization score >1.0) and low (neutralization score <0.5) neutralizingindividuals at the earliest time point available after infection (range of 1–9 mo;mean of 4 mo). (C) Plasma CXCL13 from top and low neutralizing individuals atthe time of bnAb generation (range of 24–54 mo; mean of 40 mo). CXCL13ELISA limit of detection (LOD) was 8 pg/mL. Means and interquartile ranges areshown in B and C. Each point represents an individual.

Havenar-Daughton et al. PNAS | March 8, 2016 | vol. 113 | no. 10 | 2703

IMMUNOLO

GYAND

INFLAMMATION

Plasma CXCL13 Is Correlated with GC Activity in Macaques AfterImmunization. NHPs are considered the best animal model for manypreclinical vaccine studies. We considered that CXCL13 expression byGC Tfh cells in NHPs might be more similar to that of humans. We,therefore, examined the relationship between plasma CXCL13 and GCactivity in rhesus macaques after protein immunization. Our previousidentification of an anti-human CXCR5 Ab (clone MU5UBEE) re-active to rhesus macaque CXCR5 (used in the works of refs. 32–35)allowed detection of CXCR5hi PD1hi (programmed cell death 1) GCTfh cells in macaque lymphoid tissue (Fig. 4A). In pilot experiments,we determined that GC Tfh cells in macaques highly express Bcl6,ICOS (inducible T-cell costimulator), and CD200 (Fig. 4A), directlyanalogous to human GC Tfh cells (8). Seven days after a protein andadjuvant immunization, GC Tfh- and GC B-cell responses were foundin the draining lymph node but not in a nondraining lymph node(Fig. 4B). We examined plasma CXCL13 in these animals at thesame day 7 postimmunization time point. We observed a strongpositive correlation between plasma CXCL13 concentration andGC Tfh-cell abundance in the draining lymph node (r = 0.71; P =0.013) (Fig. 4C). A nonsignificant positive trend was observed withGC B-cell frequency (Fig. S2C). The correlation between plasmaCXCL13 and GC Tfh cells in immunized rhesus macaques wasconfirmed in an additional study of 12 animals. In summary, plasmaCXCL13 responses in both mice and rhesus macaques stronglycorrelated with GC Tfh-cell frequencies in draining lymph nodeafter immunization.

Plasma CXCL13 Is Increased in Humans After Immunization. Becauseplasma CXCL13 was elevated after immunization in animal models,correlated with GC activity, and correlated with bnAb developmentin HIV+ individuals, we investigated plasma CXCL13 responses af-ter vaccination in humans. We initially examined plasma CXCL13 ina small cohort of influenza vaccine recipients. We found mixed plasmaCXCL13 responses after influenza immunization that did not exhibit a

statistically significance change (Fig. S3). The lack of a clear increase inplasma CXCL13 could be because of the fact that there was lowoverall antiinfluenza Ab response generated to the immunization causedby preexisting Ab titers (36). We, therefore, moved to study immu-nizations that generated more robust immune responses. Two sep-arate cohorts were studied. The first cohort was a vaccine cohortimmunized with the Food and Drug Administration-approved yellowfever virus vaccine (37). The second group comprised study partic-ipants in an HIV Vaccine Trials Network (HVTN) protocol testing acandidate HIV vaccine regimen (HVTN068) (38).

We tested pre- and postvaccination plasma samples obtained from17 yellow fever vaccine recipients. Seven days after immunization,statistically significant increases in plasma CXCL13 were observed(P = 0.04) (Fig. 5A). Plasma CXCL13 was 30% higher than pre-vaccinations levels in 9 of 17 individuals. We next assessed the ki-netics of plasma CXCL13 in 11 vaccinated individuals in the HVTN068.After adenovirus-5 vector with HIV-1 gene inserts (Ad5/HIV) immu-nization, plasma CXCL13 concentrations were significantly increasedover preboost levels at both days 7 (P = 0.001) and 14 (P = 0.014) (Fig.5B). For each individual donor, peak plasma CXCL13 was detected ateither the 7- or 14-d time point. Seven days after immunization, plasmaCXCL13 was 50% higher than at the prevaccination time point in 9 of11 individuals. In a larger set of samples available at only the 7-d timepoint, plasma CXCL13 positively correlated with the vaccine-specificgp41 Env IgG Ab response (r = 0.41; P = 0.037) (Fig. 5C) and showed astrong positive trend with the gp140 Env IgG Ab response (Fig. S4)measured 28 d after immunization. In a small number of individuals forwhom cryopreserved PBMCs were available, we evaluated the frequencyof activated ICOS+PD1+++CXCR5+ CD4 T cells after vaccination.ICOS+PD1+++CXCR5+ CD4 T cells are found in the blood of in-dividuals with ongoing immune responses and may be a subpopu-lation of Tfh cells trafficking through the blood before becoming GCTfh cells (8, 22, 39, 40). Even with only six individuals available foranalysis, plasma CXCL13 concentration responses correlated with

0.1 1 10 1000

20

40

60

80

100

% GC Tfh (LN)

CXC

L13

(pg/

ml)

0.0030.75r =

p =

D

A

Spleen GC Tfh Lymph node GC Tfh

CXCL13

FCS

CXCR5

PD

-1

CXCR5

PD

-1

Tonsil, other CD4+ Tonsil GC Tfh

CXCL13

FCS

GC Tfh

CD4+

non-G

C TfhB ce

llsoth

er

cells

0

10

20

30

40

% C

XCL1

3+

B

C

GC Tfh

other CD4+

E

Tonsil (CD4+ gated)

Human Inguinal LN(CD4+ gated)

GC Tfh

0.35 21.5

13.8 8.1

Fig. 2. Human GC Tfh cells produce CXCL13 and correlate with plasma CXCL13. (A, Left) Identification of GC Tfh cells as PD-1hi CXCR5hi (gated on CD4+CD3+)T cells in human tonsil. (A, Center and Right) Intracellular cytokine staining for CXCL13 in unstimulated GC Tfh cells and all other non-GC Tfh CD4+ T cells inhuman tonsil. (B) Intracellular cytokine staining for CXCL13 in unstimulated cell subsets from nine human tonsils. (C) Intracellular cytokine staining for CXCL13in unstimulated GC Tfh cells in human spleen (representative of two analyzed) and a human lymph node. (D) Identification of GC Tfh cells as PD-1hi CXCR5hi

(gated on CD4+CD3+) T cells in human inguinal lymph node. (E) Matched plasma CXCL13 and lymph node GC Tfh cells in 14 human donors. Black indicatesHIV+ antiretroviral-treated, blue indicates HIV seronegative, and red indicates HIV+ viremic controller. The Spearman r and P values are shown. The GC Tfh-cellpercentage is plotted on log scale for visualization purposes only; the linear correlation results in an r2 value of 0.48. LN, lymph node.

2704 | www.pnas.org/cgi/doi/10.1073/pnas.1520112113 Havenar-Daughton et al.

the increase in activated ICOS+PD1+++CXCR5+ CD4 T cells foundin the blood 7 d after immunization (r = 0.85; P = 0.03) (Fig. 5D).Thus, plasma CXCL13 correlates with GC-associated Ab and Tfh-cell responses in immunized humans.

DiscussionThe GC response is a critical immune mechanism by which Ab af-finity occurs, memory B cells develop, and long-lived plasma cells areproduced. Here, we show a means to monitor GC activity in lym-phoid tissues using a plasma biomarker. Plasma CXCL13 positively

correlates with the lymph node GC response in mice, macaques, andhumans. Increases in plasma CXCL13 were found in a number ofdifferent immune-activating conditions: aluminum hydroxide or TLR(Toll-like receptor) ligand adjuvants plus recombinant protein im-munizations, acute viral infections, an adenovirus vector candidateHIV vaccine, the licensed yellow fever vaccine, and HIV infection.Based on the strong correlation of GC Tfh cells and plasma CXCL13 andthe significant measurable change in plasma CXCL13 in two humanvaccine cohorts, monitoring plasma CXCL13 could be useful in humanand NHP vaccine trials, where direct analysis of lymphoid tissue is either

CB

D

A

0 7 14 210

50

100

150

200

250

CXC

L13

(pg/

ml)

0.004

0 5 10 180

200

400

600

800

CXC

L13

(pg/

ml)

0.002

0.0007 FE

CXCR5

PD

-1

Draining lymph node(CD4+ gated)

Days post immunization Days post boost

0 1 2 3 4 50

50

100

% Tfh of CD4 T cells

CXC

L13

(pg/

ml)

0.82r =p = 0.002

0.0 0.1 0.2 0.3 0.40

1

2

3

4

5

% GC Tfh of CD4 T cells

CXC

L13

Fold

cha

nge

0.69r =p = 0.023

0

50

100

150

200600800

CXC

L13

(pg/

ml) 0.03

0.01

0.06

naïve Alum + OVA

LCMV Vaccinia

Fig. 3. Plasma CXCL13 correlates with GC Tfh cellsin mice after immunization. (A) Plasma CXCL13 innaïve mice or mice 7 d after alum + NP-OVA footpadimmunization, LCMV infection, or vaccinia virus in-fection. Black circles indicate the presence of trans-ferred CD4+ TCR transgenic T cells: OT-II cells into thealum + NP-OVA group and NIP cells into the LCMVgroup. White circles indicate untransferred mice.(B) Identification of GC Tfh cells as PD-1hi CXCR5hi

cells in the draining popliteal lymph node of an NP-OVA immunized mouse. (C) Correlation of plasmaCXCL13 and GC Tfh cells in naïve and alum + NP-OVAimmunized mice from A. GC Tfh cells in naïve micewere set at the limit of detection (0.1% of the totalCD4+ T cells). (D and E) Plasma CXCL13 in B6 miceimmunized in the footpad with KLH + alum. Datarepresentative of two experiments of 10 mice each.(D) Plasma CXCL13 pre- and postprimary immunization.(E) Plasma CXCL13 in the same mice as those in D aftera boost with KLH + alum at 50 d postprimary immu-nization. (F) Correlation of plasma CXCL13 and GC Tfhcells in the draining popliteal lymph node at 10 and 18 dpostboost from animals shown in E. CXCL13 is plotted asfold change of day 10 or 18 postbooster immunizationover preboost (d50) CXCL13 concentration.

0.588 3.29

PD

-1

CXCR5

Non-draining LN Draining LN

1.87 6.49 Non-draining LN Draining LN

CXCR5 Bcl-6 ICOS CD200

PD

-1

45.6

43.9

7.01A

B

C

GC Tfh (gated on CD3+CD4+ T cells) GC B cells (gated on CD20+ B cells)

GC Tfh

0 1 2 3 4 5 60

20

40

60

80

% GC Tfh of CD4 T cells

CXC

L13

(pg/

ml)

0.710.013

r =p =

Bcl-6

Ki6

7

Fig. 4. Plasma CXCL13 correlates with GC Tfh cells in rhesus macaques after immunization. (A) Identification and characterization of GC Tfh cells (PD1hi CXCR5hi CD4+ Tcells) in lymph node from rhesus macaques. Expression of Bcl-6, ICOS, and CD200 is shown for GC Tfh cells (red), mantle CXCR5+ Tfh cells (blue), and non-Tfh cells (gray).(B) Identification of (Left) GC Tfh cells and (Right) GC B cells (identified as Ki67+Bcl6+ cells; gated on CD20+ cells) in nondraining (axillary) and draining lymph nodes(popliteal) of macaques 7 d after protein and adjuvant immunization in the leg. (C) Correlation of plasma CXCL13 and GC Tfh cells in the draining popliteal lymph nodeofmacaques 7 d after the second or third protein and adjuvant immunization. The correlation holds even with removal of the potential outlier point (highest GC Tfh celland highest CXCL13), with values of r = 0.62 and P = 0.048. Aluminum hydroxide or TLR (Toll-like receptor) ligand-encapsulated PLGA [poly(lactic-co-glycolic acid)]nanoparticles were used as adjuvants for gp140 Env and p55 Gag recombinant SIV (Simian immunodeficiency virus) proteins. All data are representative of two similarimmunization experiments in rhesus macaques totaling 22 animals. LN, lymph node.

Havenar-Daughton et al. PNAS | March 8, 2016 | vol. 113 | no. 10 | 2705

IMMUNOLO

GYAND

INFLAMMATION

not possible or undesirable for fear of disturbing the ongoing im-mune response.

If bnAbs against HIV are to be generated by vaccination, the GCresponse will play a central role. Measuring CXCL13 in vaccine studiescan provide data on postvaccination GC activity, a major driver of Abquality by SHM. Furthermore, in some cases, antigen-specific Ab resultsare not measured until after a final boost 6 mo after the primary im-munization. CXCL13 can be measured after each immunization, pro-viding much earlier data on the progress of the immune response to theimmunization scheme, which could be important for in-trial decision-making. Our studies detecting increases in plasma CXCL13 in themajority but not all of the immunized individuals suggest that GCs werenot generated in certain individuals, a potentially critical observation.We do not suggest that CXCL13 analysis should replace antigen-spe-cific Ab titer data, but rather that CXCL13 monitoring be added as avaluable parameter to gain an understanding of the magnitude of theGC activity that is necessary for the development of improved Abquality. Given that GC B cells do not exist in peripheral blood,CXCL13 may be the best available proxy for those inaccessible cells.

Plasma CXCL13 has been proposed to serve as a biomarker of au-toimmune diseases, such as rheumatoid arthritis, systemic lupus eryth-ematosus, Sjogren’s syndrome, and Myasthenia Gravis (41). Elevatedplasma CXCL13 was detected in patients with systemic lupus eryth-ematosus and further increased in individuals with severe disease pre-senting with nephritis or anti-DNA Ab responses (19). In rheumatoidarthritis, CXCL13 was not only followed as a plasma biomarker ofdisease, but also, CXCL13 blockade has been proposed as a treatment(42). It is important to note that analysis of plasma CXCL13 is not anantigen- or disease-specific readout. Plasma CXCL13 reports total GCactivity, and the basal levels detected in unimmunized humans,macaques, and mice likely reflect ongoing GC activity in tonsillar-and gut-associated lymphoid tissue. For these reasons, we consider amultiparameter approach to be the best approach. Analysis ofplasma CXCL13 together with analysis of other potential bio-markers specific to the immunological and pathological setting un-der study are advisable.

We have shown a strong correlation between CXCL13 and lym-phoid tissue resident GC Tfh cells. With the additional observationthat GC Tfh cells are robust producers of CXCL13, it is suggestive of adirect relationship between GC Tfh cells and plasma CXCL13. We didnot identify other cell types in lymphoid tissue, monocytes in PBMCs,or CXCR5+ Tfh cells in PBMCs as producers of CXCL13 by in-tracellular staining. Although follicular dendritic cells (FDC) and somedendritic cell subsets are likely lost during tonsil tissue processing, ahistological study suggests that much of the CXCL13 observed in thetonsil GC costains with PD-1, and PD-1 is an excellent marker of GC Tfhcells (27). In the same study, CXCL13 was only weakly associated withCD21+ FDC. Monocytes (26) and dendritic cells (28) have beenreported to express CXCL13 and could contribute to plasma CXCL13concentrations in different inflammatory settings. The observation that

monocytes can produce CXCL13 in response to IFN-I may be relevantin HIV infection and immunizations with viral vectors or adjuvantscontaining TLR ligands. However, we did not identify CXCL13-pro-ducing monocytes in PBMCs from HIV+ individuals ex vivo or after ashort in vitro stimulation, and IFN-I responses are normally short induration, peaking at 24–48 h after immunization (43, 44). Therefore,monocyte-generated CXCL13 may not significantly impact plasmaCXCL13 measured in our immunization studies. These later time pointsexamined are contemporaneous with GC activity.

Here, we show that CXCL13 acts as a plasma biomarker for GC ac-tivity in generally inaccessible lymphoid tissue. Plasma CXCL13 corre-lates with GC activity after immunization in animal models and in HIV+

humans. Furthermore, plasma CXCL13 is associated with generation ofHIV bnAbs, and CXCL13 was elevated in humans after immunization.Together, these findings support the use of CXCL13 as a plasmabiomarker of GC activity useful for studying differences in humoralimmunity among patients and vaccinees.

Materials and MethodsIAVI Protocol C and Human Vaccine Cohorts. The IAVI Protocol C cohort has beendescribed elsewhere (22, 23). An HIV neutralization score for each plasmasample was determined as in the work by Simek et al. (45) to account for bothbreadth and potency. All donors were ART-free and had CD4 counts >200 cellsper 1 mL at each time point tested.

CXCL13was analyzed in the serumbefore and 1wkafter a single immunizationof TIV Fluzone (Sanofi Pasteur) in vaccine recipients from the Stanford–LucilePackard Children’s Hospital Vaccine Program (36). Yellow fever vaccine recipientsreceived the Food and Drug Administration-approved 17D yellow fever vaccine(37). Plasma samples from HIV-uninfected candidate vaccine recipients from theHVTN068 trial (38) were analyzed for CXCL13 in plasma after a booster immu-nization (Ad5/HIV) given at the 6-mo time point. Primary immunizations wereeither Ad5/HIV at month 0 or DNA/HIV at months 0 and 1. Some participantsreceived placebo.

Human Lymph Node, Spleen, and Tonsil. Inguinal lymph nodes fromHIV-infectedpersons and HIV− healthy volunteers were obtained by excisional surgical bi-opsy under local anesthesia at Massachusetts General Hospital and processedas in the work by Lindqvist et al. (46) Additional nonidentifiable, discarded,excess spleen, lymph node, and tonsil tissues were obtained at MassachusettsGeneral Hospital or Rady Children’s Hospital–San Diego. Spleen and lymph nodetissues (46) and tonsils (15) were processed as previously described.

Macaque and Mouse Immunization. Macaques were immunized s.c. with gp140(SIVmac239) Env and p55 Gag protein mixed with either aluminum hydroxide(Alhydrogel 2% adjuvant; Invivogen) or MPL (monophosphoryl lipid A) + R848-encapsulated PLGA [poly(lactic-co-glycolic acid)] nanoparticles (47). A full de-scription of the vaccine trial methods and results will be published elsewhere.

C57BL/6 (B6) mice (Jackson Laboratory) were immunized with alum + NP-OVA(Sigma-Aldrich); 10 μg NP-OVA was mixed 1:1 with 10% (wt/vol) aluminum hy-droxide (aluminum potassium sulfate dodecahydrate; Sigma-Aldrich) in PBS. Insomemice, 2 × 105 OVA-specific OT-II TCR transgenic CD4 T cells were transferred

0 10 20 30 40 500

1

2

3

CXCL13 (pg/ml)

chan

ge o

f IC

OS+

Tfh

-like

cells

in b

lood

CBA

0

20

40

60

80

CXC

L13

(pg/

ml)

0.001

0.014

0 7 21140

20

40

60

LOD

75100125150

CXC

L13

(pg/

ml)

0.04D

0 20 40 60 800.0

0.5

1.0

1.5

2.0

2.5

CXCL13 (pg/ml)

gp41

ELI

SA (O

D)

0.0370.41r =

p =

r = 0.85

p = 0.03

Day 0 Day 7Day post boost

Fold

Fig. 5. Plasma CXCL13 is increased after immunization in humans. (A) Plasma CXCL13 measured before immunization (day 0) and 7 d postyellow fever vacci-nation in 17 individuals. (B–D) HVTN068 participants who received an Ad5/vector encoding HIV gag and envelope immunization 6 mo postprime. (B) Kineticanalysis of plasma CXCL13 post-Ad5/HIV boost. Plasma CXCL13 was measured in 11 vaccinated individuals. (C) Correlation of plasma CXCL13 7 d postimmunizationand anti-gp41 Env Ab responses (ELISA OD) 4 wk postimmunization in 26 vaccinated individuals. Anti-gp41 Ab ELISA OD is background-subtracted. (D) Correlationof plasma CXCL13 7 d postimmunization and the fold change of ICOS+ blood Tfh-like cells (percentage at day 7 postboost over percentage at preboost time point)in six vaccinated donors. LOD, limit of detection; OD, optical density.

2706 | www.pnas.org/cgi/doi/10.1073/pnas.1520112113 Havenar-Daughton et al.

3.5 d before immunization. For KLH + alum immunized mice, 10 μg KLH wasmixed 1:1 with 10% aluminum hydroxide in PBS. For LCMV immunization exper-iments, 2 × 105 NIP CD4+ TCR transgenic cells (48) were transferred before im-munization with 2 × 106 pfu LCMV Armstrong by i.p. injection. For vaccinia virusimmunization experiments, B6 mice were i.p. injected with 6 × 105 pfu.

CXCL13 ELISA and Flow Cytometry. The Human CXCL13 Quantikine ELISA Kit(R&D Systems) was used for both human and macaque (49) plasma samples asdescribed in the instructions. The mouse CXCL13 DuoSet (R&D Systems) was usedfor quantification of CXCL13 in mouse serum. All samples were tested in duplicate.

GC Tfh cells were characterized by flow cytometry as previously described inmouse lymph node (50), human tonsil (15, 22), or human lymph node (46). Pre-viously cryopreserved human PBMCs from HVTN068 subjects were stained. In-tracellular CXCL13 was detected by intracellular cytokine staining as described (15).Cells were acquired on a BD Fortessa Analyzer.

Statistics. TheMann–Whitney testwasused for evaluatingdifferences amonggroups.TheWilcoxon test was used to evaluate differences between time points for the sameindividuals. The Spearman correlation test was used for all correlative analysis.Covariance of plasma CXCL13 and viral load was evaluated with the ANCOVA

multivariate statistical test (VassarStats). Prism 5.0 (GraphPad) was used for all otherdata statistical analyses.

Study Approvals. Informed, written consent was obtained from all human studyparticipants before enrollment in the human studies listed above and approved bythe local institutional review boards (La Jolla Institute Internal Review Board, TheScripps Research Institute Internal Review Board, Massachusetts General HospitalPartners Human Research Committee, the Institutional Review Board of theResearch Compliance Office at Stanford University, Institutional Review Boards atEmoryUniversity, Centers for Disease Control, and the Institutional ReviewBoard atthe HVTN). Rhesus macaque study procedures were performed in accordance withEmory School of Medicine Institutional Animal Care and Use Committee-approvedprotocols. Mouse study procedures were performed in accordance with approvedanimal protocols at La Jolla Institute for Allergy and Immunology.

ACKNOWLEDGMENTS.We thank all project study volunteers and clinical andstudy team members. This work was supported by grants from the NationalInstitute of Allergy and Infectious Diseases, International AIDS Vaccine Initiative(IAVI), and the Scripps Center for HIV/AIDS Vaccine Immunology and ImmunogenDiscovery (Scripps CHAVI-ID).

1. Victora GD, Nussenzweig MC (2012) Germinal centers. Annu Rev Immunol 30:429–457.2. Haynes BF, et al. (2014) Progress in HIV-1 vaccine development. J Allergy Clin Immunol

134(1):3–10.3. Burton DR, Mascola JR (2015) Antibody responses to envelope glycoproteins in HIV-1 in-

fection. Nat Immunol 16(6):571–576.4. Sok D, et al. (2013) The effects of somatic hypermutation on neutralization and binding in

the PGT121 family of broadly neutralizing HIV antibodies. PLoS Pathog 9(11):e1003754.5. Klein F, et al. (2013) Somatic mutations of the immunoglobulin framework are generally

required for broad and potent HIV-1 neutralization. Cell 153(1):126–138.6. Burton DR, et al. (2012) A blueprint for HIV vaccine discovery. Cell Host Microbe 12(4):

396–407.7. Haynes BF (2015) New approaches to HIV vaccine development. Curr Opin Immunol 35:39–47.8. Crotty S (2014) T follicular helper cell differentiation, function, and roles in disease.

Immunity 41(4):529–542.9. Johnston RJ, et al. (2009) Bcl6 and Blimp-1 are reciprocal and antagonistic regulators

of T follicular helper cell differentiation. Science 325(5943):1006–1010.10. Victora GD, et al. (2010) Germinal center dynamics revealed by multiphoton micros-

copy with a photoactivatable fluorescent reporter. Cell 143(4):592–605.11. Gitlin AD, Shulman Z, Nussenzweig MC (2014) Clonal selection in the germinal centre

by regulated proliferation and hypermutation. Nature 509(7502):637–640.12. Ansel KM, et al. (2000) A chemokine-driven positive feedback loop organizes lym-

phoid follicles. Nature 406(6793):309–314.13. Allen CDC, Okada T, Cyster JG (2007) Germinal-center organization and cellular dy-

namics. Immunity 27(2):190–202.14. Wang X, et al. (2011) Follicular dendritic cells help establish follicle identity and

promote B cell retention in germinal centers. J Exp Med 208(12):2497–2510.15. Kroenke MA, et al. (2012) Bcl6 and Maf cooperate to instruct human follicular helper

CD4 T cell differentiation. J Immunol 188(8):3734–3744.16. Rasheed A-U, Rahn H-P, Sallusto F, Lipp M, Müller G (2006) Follicular B helper T cell activity

is confined to CXCR5(hi)ICOS(hi) CD4 T cells and is independent of CD57 expression. Eur JImmunol 36(7):1892–1903.

17. Widney DP, et al. (2005) Serum levels of the homeostatic B cell chemokine, CXCL13,are elevated during HIV infection. J Interferon Cytokine Res 25(11):702–706.

18. Sansonno D, et al. (2008) Increased serum levels of the chemokine CXCL13 and up-regu-lation of its gene expression are distinctive features of HCV-related cryoglobulinemia andcorrelate with active cutaneous vasculitis. Blood 112(5):1620–1627.

19. Lee HT, et al. (2010) SerumBLC/CXCL13 concentrations and renal expression of CXCL13/CXCR5in patients with systemic lupus erythematosus and lupus nephritis. J Rheumatol 37(1):45–52.

20. Wong CK, et al. (2010) Elevated production of B cell chemokine CXCL13 is correlated withsystemic lupus erythematosus disease activity. J Clin Immunol 30(1):45–52.

21. Panse J, et al. (2008) Chemokine CXCL13 is overexpressed in the tumour tissue and inthe peripheral blood of breast cancer patients. Br J Cancer 99(6):930–938.

22. Locci M, et al. (2013) Human circulating PD-1+CXCR3-CXCR5+memory Tfh cells are highlyfunctional and correlate with broadly neutralizing HIV antibody responses. Immunity39(4):758–769.

23. Landais E, et al. (2016) Broadly neutralizing antibody responses in a large longitudinalsub-Saharan HIV primary infection cohort. PLoS Pathog 12(1):e1005369.

24. Cohen K, Altfeld M, Alter G, Stamatatos L (2014) Early preservation of CXCR5+ PD-1+helper T cells and B cell activation predict the breadth of neutralizing antibody re-sponses in chronic HIV-1 infection. J Virol 88(22):13310–13321.

25. Hessell AJ, Haigwood NL (2012) Neutralizing antibodies and control of HIV: Movesand countermoves. Curr HIV/AIDS Rep 9(1):64–72.

26. Cohen KW, Dugast A-S, Alter G, McElrath MJ, Stamatatos L (2015) HIV-1 single-stranded RNA induces CXCL13 secretion in human monocytes via TLR7 activation andplasmacytoid dendritic cell-derived type I IFN. J Immunol 194(6):2769–2775.

27. Yu H, Shahsafaei A, Dorfman DM (2009) Germinal-center T-helper-cell markers PD-1 andCXCL13 are both expressed by neoplastic cells in angioimmunoblastic T-cell lymphoma.Am JClin Pathol 131(1):33–41.

28. Ishikawa S, et al. (2001) Aberrant high expression of B lymphocyte chemokine (BLC/CXCL13) by C11b+CD11c+ dendritic cells in murine lupus and preferential chemotaxisof B1 cells towards BLC. J Exp Med 193(12):1393–1402.

29. Sanders RW, et al. (2015) HIV-1 VACCINES. HIV-1 neutralizing antibodies induced bynative-like envelope trimers. Science 349(6244):aac4223.

30. Jardine JG, et al. (2015) HIV-1 VACCINES. Priming a broadly neutralizing antibody responseto HIV-1 using a germline-targeting immunogen. Science 349(6244):156–161.

31. Dosenovic P, et al. (2015) Immunization for HIV-1 broadly neutralizing antibodies inhuman Ig knockin mice. Cell 161(7):1505–1515.

32. Hong JJ, et al. (2014) Early lymphoid responses and germinal center formation cor-relate with lower viral load set points and better prognosis of simian immunodefi-ciency virus infection. J Immunol 193(2):797–806.

33. Xu H, Wang X, Lackner AA, Veazey RS (2014) PD-1(HIGH) follicular CD4 T helper cell subsetsresiding in lymph node germinal centers correlate with B cell maturation and IgG productionin rhesus macaques. Front Immunol 5:85.

34. Mylvaganam GH, et al. (2014) Diminished viral control during simian immunodefi-ciency virus infection is associated with aberrant PD-1hi CD4 T cell enrichment in thelymphoid follicles of the rectal mucosa. J Immunol 193(9):4527–4536.

35. Chowdhury A, et al. (2015) Decreased T follicular regulatory cell/T follicular helper cell(TFH) in simian immunodeficiency virus-infected rhesus macaques may contribute toaccumulation of TFH in chronic infection. J Immunol 195(7):3237–3247.

36. Furman D, et al. (2013) Apoptosis and other immune biomarkers predict influenzavaccine responsiveness. Mol Syst Biol 9(1):659.

37. Edupuganti S, et al.; YF-Ig Study Team (2013) A randomized, double-blind, controlled trial ofthe 17D yellow fever virus vaccine given in combination with immune globulin or placebo:Comparative viremia and immunogenicity. Am J Trop Med Hyg 88(1):172–177.

38. De Rosa SC, et al.; National Institute of Allergy and Infectious Diseases HIV VaccineTrials Network (2011) HIV-DNA priming alters T cell responses to HIV-adenovirusvaccine even when responses to DNA are undetectable. J Immunol 187(6):3391–3401.

39. Bentebibel SE, et al. (2013) Induction of ICOS+CXCR3+CXCR5+ TH cells correlates withantibody responses to influenza vaccination. Sci Transl Med 5(176):176ra32.

40. He J, et al. (2013) Circulating precursor CCR7(lo)PD-1(hi) CXCR5⁺ CD4⁺ T cells indicate Tfh cellactivity and promote antibody responses upon antigen reexposure. Immunity 39(4):770–781.

41. Finch DK, Ettinger R, Karnell JL, Herbst R, SleemanMA (2013) Effects of CXCL13 inhibitionon lymphoid follicles in models of autoimmune disease. Eur J Clin Invest 43(5):501–509.

42. Klimatcheva E, et al. (2015) CXCL13 antibody for the treatment of autoimmune dis-orders. BMC Immunol 16(1):6.

43. Thompson LJ, Kolumam GA, Thomas S, Murali-Krishna K (2006) Innate inflammatorysignals induced by various pathogens differentially dictate the IFN-I dependence ofCD8 T cells for clonal expansion and memory formation. J Immunol 177(3):1746–1754.

44. Johnson MJ, et al. (2012) Type I IFN induced by adenovirus serotypes 28 and 35 hasmultiple effects on T cell immunogenicity. J Immunol 188(12):6109–6118.

45. Simek MD, et al. (2009) Human immunodeficiency virus type 1 elite neutralizers: Individualswith broad and potent neutralizing activity identified by using a high-throughput neutral-ization assay together with an analytical selection algorithm. J Virol 83(14):7337–7348.

46. Lindqvist M, et al. (2012) Expansion of HIV-specific T follicular helper cells in chronicHIV infection. J Clin Invest 122(9):3271–3280.

47. Kasturi SP, et al. (2011) Programming the magnitude and persistence of antibodyresponses with innate immunity. Nature 470(7335):543–547.

48. Nance JP, Bélanger S, Johnston RJ, Takemori T, Crotty S (2015) Cutting edge: T follicular helpercell differentiation is defective in the absence of Bcl6 BTB repressor domain function.J Immunol 194(12):5599–5603.

49. Lucero CM, et al. (2013) Macaque paneth cells express lymphoid chemokine CXCL13 andother antimicrobial peptides not previously described as expressed in intestinal crypts. ClinVaccine Immunol 20(8):1320–1328.

50. Choi YS, et al. (2013) Bcl6 expressing follicular helper CD4 T cells are fate committedearly and have the capacity to form memory. J Immunol 190(8):4014–4026.

Havenar-Daughton et al. PNAS | March 8, 2016 | vol. 113 | no. 10 | 2707

IMMUNOLO

GYAND

INFLAMMATION