CCR5 deficiency impairs CD4+ T cell memory responses and antigenic sensitivity through increased ceramide synthesis

In CD4+ T cells, CCR5 is not only a coreceptor for HIV-1 infection, but also contributes to their functional fitness. Here we show that by limiting GATA-1-induced transcription of specific ceramide synthases, CCR5 signaling reduces ceramide levels and thereby increases T cell antigen receptor (TCR) nanoclustering in antigen-experienced mouse and human CD4+ T cells. This activity is CCR5-specific and independent of CCR5 costimulatory activity. CCR5-deficient mice showed reduced production of high affinity class-switched antibodies, but only after antigen rechallenge, which implies an impaired memory CD4+ T cell response. This study identifies a CCR5 function in the generation of CD4+ T cell memory responses, and establishes an antigen-independent mechanism that regulates TCR nanoclustering by altering specific lipid species.


INTRODUCTION
The C-C motif chemokine receptor 5 (CCR5) is a seven-transmembrane G protein-coupled receptor (GPCR) expressed on the surface of several innate and adaptive immune cell subtypes, including effector and memory CD4 + T lymphocytes (Gonzalez-Martin et al., 2012). CCR5 acts also a necessary coreceptor for infection by HIV-1. An HIV-resistant population served to identify a 32-bp deletion within the CCR5 coding region (ccr5Δ32), which yields a nonfunctional receptor (Blanpain et al., 2002). Since ccr5Δ32 homozygous individuals are seemingly healthy, a radical body of thought considers that CCR5 is dispensable for immune cell function.
Experimental and epidemiological evidence nonetheless indicates that CCR5 has an important role in innate and acquired immune responses. CCR5 and its ligands C-C motif ligand 3 [CCL3; also termed macrophage inflammatory protein (MIP)-1α], CCL4 (MIP-1β), CCL5 [regulated upon activation, normal T cell expressed and secreted (RANTES)] and CCL3L1 have been associated with exacerbation of chronic inflammatory and autoimmune diseases. Despite varying information due probably to ethnicity effects (Lee et al., 2013, Schauren et al., 2013, further complicated in admixed populations (Toson et al., 2017), epidemiological studies support the ccr5Δ32 allele as a marker for good prognosis for these overreactive immune diseases (Vangelista & Vento, 2017). In contrast, ccr5Δ32 homozygotes are prone to fatal infections by several pathogens such as influenza, West Nile, and tick-borne encephalitis viruses (Ellwanger & Chies, 2019, Falcon et al., 2015, Lim & Murphy, 2011. The mechanisms by which the ccr5Δ32 polymorphism affects all these pathologies have usually been linked to the capacity of CCR5 to regulate leukocyte trafficking. For example, CCR5 deficiency reduces recruitment of influenza-specific memory CD8 + T cells and accelerates macrophage accumulation in lung airways during virus rechallenge (Dawson et al., 2000, Kohlmeier et al., 2008; this could lead to acute severe pneumonitis, a fatal flu complication. CCR5 nonetheless has migration-independent functions that maximize T cell activation by affecting immunological synapse (IS) formation (Floto et al., 2006, Franciszkiewicz et al., 2009, Molon et al., 2005 as well as T cell transcription programs associated with cytokine production (Camargo et al., 2009, Lillard et al., 2001. CCR5 and its ligands are also critical for cell-mediated immunity to tumors and pathogens, including HIV-1 (Bedognetti et al., 2013, Dolan et al., 2007, González-Martín et al., 2011, Ugurel et al., 2008. Whereas the role of CCR5 in T cell priming is well-established, its involvement in memory responses has not been addressed in depth. Only a single report suggested CCR5 involvement in CD4 + T cell promotion of memory CD8 + T cell generation through a migration-dependent process (Castellino et al., 2006). It remains unknown whether CCR5 endows memory T cells with additional properties. One such property is the elevated sensitivity of effector and memory ("antigen-experienced") CD4 + and CD8 + T cells to their cognate antigen compared to naïve cells (Huang et al., 2013, Kersh et al., 2003, Kimachi et al., 1997. This sensitivity gradient (memory>>effector>naïve) in CD8 + T cells is linked to increased valency of preformed TCR oligomers at the cell surface, termed TCR nanoclusters (Kumar et al., 2011). This antigenindependent TCR nanoclustering (Lillemeier et al., 2010, Schamel & Alarcon, 2013, Schamel et al., 2005, Schamel et al., 2006, Sherman et al., 2011 enhances antigenic sensitivity by increasing avidity to multimeric peptide-major histocompatibility complexes (Kumar et al., 2011, Molnar et al., 2012 and by allowing cooperativity between TCR molecules (Martín-Blanco et al., 2018, Martínez-Martín et al., 2009. TCRβ subunit interaction with cholesterol (Chol) and the presence of sphingomyelins (SM) are both essential for TCR nanoclustering (Beck-Garcia et al., 2015, Molnar et al., 2012. Replacement of Chol by Chol sulfate impedes TCR nanocluster formation and reduces CD4 + CD8 + thymocyte sensitivity to weak antigenic peptides (Wang et al., 2016). Whether antigen-experienced CD4 + T cell sensitivity is linked to TCR nanoscopic organization and the homeostatic factors that regulate TCR nanoclustering remains unexplored.
Given its costimulatory role in CD4 + T cells, we speculated that CCR5 signals would affect the antigenic sensitivity of CD4 + memory T cells. To test this hypothesis, we analyzed the function of in vivo-generated memory CD4 + T cells in wild-type (WT) and CCR5 -/mice, and the effect of CCR5 deficiency on CD4 T cell help in the T-dependent humoral response. We found that CCR5 is necessary for the establishment of a functional CD4 memory response through a mechanism independent of its co-stimulatory role for the TCR signal. We show that CCR5 deficiency does not affect memory CD4 T cell generation, but reduces their sensitivity to antigen. Our data demonstrate an unreported CCR5 regulatory role in memory CD4 + T cell function by inhibiting the synthesis of ceramides, which are identified here as negative membrane regulators of TCR nanoscopic organization.

CCR5 deficiency impairs the CD4 + T cell memory response
To determine the role of CCR5 in CD4 + memory T cell generation and/or function, we adoptively transferred congenic CD45.1 mice with lymph node/spleen cell suspensions from OT-II WT or CCR5 -/mice (CD45.2), and subsequently infected them with OVA-encoding vaccinia virus; five weeks post-immunization, we analyzed spleen CD45.2 + donor cells from OT-II mice. CCR5 expression on OT-II cells affected neither the total number of memory CD4 + T cells (Fig.1A We also studied T cell-dependent B cell responses in WT and CCR5 -/mice after immunization with the hapten 4-hydroxy-3-iodo-5-nitrophenylacetyl coupled to ovalbumin (NIP-OVA; Fig.1F). We detected no difference in the percentage or absolute number of T follicular helper (Tfh) cells (CD4 + , CD44 hi , CXCR5 + , PD1 + ) between WT and CCR5 -/mice at seven days post-immunization ( Fig. 1G-I). At day 30, half of the mice were boosted with the same NIP-OVA immunogen (OVA/OVA) and the other half received NIP conjugated with another carrier protein (OVA/KLH); levels of NIP-specific high and low affinity immunoglobulins (Ig) were analyzed 15 days later. Comparison of the humoral responses between OVA/OVA-and OVA/KLH-immunized mice would assess the effect of memory CD4 + T cells specific for the first carrier protein on the humoral response to NIP. There were no differences in high/low affinity NIP-specific IgM production between WT and CCR5 -/mice with either immunization strategy (Fig. 1J, K). CCR5 deficiency markedly impaired the generation of high-affinity class-switched anti-NIP antibodies specifically in OVA/OVAimmunized mice (Fig. 1J, K). Since class switching was similar in WT and CCR5 -/-OVA/KLH-immunized mice, our results suggest that CCR5 deficiency reduces the generation of highaffinity class-switched immunoglobulins due to deficient memory CD4 + T cell function.

The CCR5 effect on antigen-experienced CD4 + T cells is cell-autonomous
To test whether the in vivo memory defect associated to CCR5 deficiency was intrinsic to CD4 + T cells, we activated OT-II WT and CCR5 -/spleen T cells with OVA323-339 antigen for three days; after antigen removal, we cultured cells for four days with IL-2 or IL-15. OT-II cells that differentiated in exogenous IL-2 expressed CCL3, CCL4, CCL5, and a functional CCR5 receptor, as determined by their ability to migrate in a CCL4 gradient ( Fig. S1A-C).
Like CD8 + T cells (Richer et al., 2015), OT-II cells cultured with IL-15 showed a memorylike phenotype; they were smaller than IL-2-cultured cells and retained CD62L with reduced activation marker expression (CD25, CD69, CD44) compared to IL-2-cultured T cells ( Fig.   2A). Findings were similar in OT-II WT and CCR5 -/cells (Fig. 2B), which reinforced the idea that CCR5 is not involved in CD4 + T memory cell differentiation. Restimulation of IL-2-or IL-15-expanded OT-II lymphoblasts with the OVA323-339 peptide nonetheless indicated that CCR5expressing cells showed strong proliferation and higher IL-2 production at low antigen concentrations than CCR5-deficient cells ( Fig. 2C-F), indicative of an increased number of cells responding to antigenic stimulation. CCR5 might thus increase the antigenic sensitivity of antigen-experienced CD4 + T cells in a cell-autonomous manner.

CCR5 modulates TCR nanoclustering in antigen-experienced CD4 + T cells
The high antigenic sensitivity of antigen-experienced CD8 + T cells was partially attributed to increased TCR nanoclustering (Kumar et al., 2011). To determine whether CCR5 deficiency influences TCR organization, we used electron microscopy (EM) to analyze surface replicas of OT-II WT and CCR5 -/naïve cells and lymphoblasts after labeling with anti-CD3ε antibody and 10 nm gold-conjugated protein A. We found no differences in TCR nanoclusters between OT-II WT or CCR5 -/naïve cells, which had a small percentage of TCR nanoclusters larger 8 than 4 TCR in both genotypes (Fig. 3A). In contrast, there was a significant increase in TCR nanocluster number and size in WT compared to CCR5 -/lymphoblasts ( Fig. 3B, C). As predicted, there was a gradient in TCR nanoclustering of naïve << IL-2-< IL-15-differentiated OT-II WT cells (Fig. S1D), which coincided with increased antigenic sensitivity of the IL-15expanded cells (Fig. S1E). These findings thus reinforce the IL-15-induced memory-like phenotype versus the IL-2-induced effector-like phenotype, and link TCR nanoclustering with increased sensitivity in antigen-experienced CD4 + T cells. The difference in TCR nanoclustering between WT and CCR5 -/cells was nevertheless similar in IL-2-and IL-15expanded lymphoblasts, which indicates that CCR5 affects TCR nanoclustering in lymphoblasts independently of the cytokine milieu.
Using a Monte-Carlo simulation, we applied data from surface replicas of naïve and IL-2expanded OT-II lymphoblasts to determine whether the experimental frequency of cluster size was due to random distribution of gold particles. In all cases, the cluster distributions observed experimentally differed significantly from pure random proximity between clusters (Fig. S2).
To define the differences between OT-II WT and CCR5 -/cells, we used a model that accounts for receptor clustering dynamics (Castro et al., 2014), a Bayesian inference method that estimates the so-called clustering parameter, b. Based on this model, we concluded that the probability of a chance nanocluster distribution similar to that observed for naïve and activated OT-II WT and CCR5 -/cells approaches 0% (Fig. 3D, E). Posterior distribution analysis also showed that whereas the clustering parameter was very similar between naïve OT-II WT and CCR5 -/cells (Fig. 3D), there was clear separation in lymphoblasts (Fig. 3E). These analyses provide a mathematical framework that validates the TCR nanoclustering differences between WT and CCR5 -/cells, as determined by EM.
The differences in TCR oligomerization between OTII WT and CCR5 -/lymphoblasts were also studied using blue-native gel electrophoresis (BN-PAGE) (Schamel et al., 2005, Swamy & Schamel, 2009). Cell lysis with digitonin, a detergent that disrupts TCR nanoclusters into their monomeric components, showed that WT and CCR5 -/lymphoblasts expressed comparable TCR levels, as detected with anti-CD3ζ antibodies (Fig. 3F). Cell lysis with Brij96, which preserves TCR nanoclusters, showed a notable reduction in large TCR complexes in CCR5 -/compared to WT lymphoblasts (Fig. 3F). Two independent techniques thus support a CCR5 role in TCR nanoscopic organization in antigen-experienced CD4 + T cells.
To determine whether CCR5 controls TCR nanoclustering in in vivo-generated memory T cells, we analyzed TCR distribution in surface replicas of CD4 + memory T cells purified by negative selection from OVA/OVA-immunized WT and CCR5 -/mice (Fig. S3). CD4 + memory cells from CCR5 -/mice showed fewer, smaller TCR nanoclusters than those from WT counterparts ( Fig. 3G), which indicates that CCR5 promotes formation of large TCR nanoclusters in endogenously generated CD4 + memory T cells.

CCR5-induced TCR nanoclustering is independent of its co-stimulatory activity
Since CCR5 has co-stimulatory functions in CD4 + T cell priming (González-Martín et al., 2011, Molon et al., 2005, it is of interest to know whether defective TCR clustering in CCR5 -/lymphoblasts is due to suboptimal primary activation of these cells. To address this question, we treated OT-II WT cells with the CCR5 antagonist TAK-779 at various intervals throughout culture, and analyzed TCR nanoclusters in IL-2-expanded T lymphoblasts. TAK-779 addition during the priming phase (blockade of CCR5 co-stimulatory function) decreased the percentage of large TCR nanoclusters compared to untreated controls (Fig. 4A). TAK-779 treatment did not alter TCR clustering in OT-II CCR5 -/cells (Fig. S4), which indicates that the TAK-779 effect on OT-II cells is CCR5-specific.
To avoid interference with the CCR5 co-stimulatory activity, we primed OT-II WT cells in the absence of the inhibitor, and added TAK-779 only during IL-2-driven expansion of the CD4 + lymphoblasts. In these conditions, TAK-779 also reduced the percentage of large TCR nanoclusters (Fig. 4B), which indicates that the CCR5 signals that control TCR organization are independent of those involved in its co-stimulatory function.
We next explored whether other chemokine receptors involved in T cell activation control TCR nanoclusters in CD4 + T cells. CXCR4 is a paradigmatic chemokine receptor that also provides costimulatory signals (Kumar et al., 2006, Smith et al., 2013. We primed OT-II WT cells in the presence of the CXCR4 antagonist AMD3100, and analyzed TCR nanoclusters in IL-2-expanded T lymphoblasts. Vehicle-and AMD3100-treated cells showed similar TCR nanocluster distribution (Fig. 4D), which implies that CXCR4 blockade does not interfere with TCR nanoclustering.

CCR5 deficiency increases ceramide levels in CD4 + T cells
We analyzed CCR5 regulation of TCR nanoclustering in CD4 + T cells, and found no differences between OT-II WT and CCR5 -/cells in TCR/CD3 chain mRNA levels or in cell surface expression of the TCRα chain (Fig. S5). These data suggest that the reduction in TCR clustering in CCR5 -/cells is not due to decreased TCR expression.
TCR nanoclustering is dependent on plasma membrane Chol and SM (Molnar et al., 2012), two lipids also necessary for CCR5 signaling (Mañes et al., 2001). OT-II WT and CCR5 -/lymphoblasts expressed comparable levels of total Chol and SM species (Fig. 5A, B). OT-II CCR5 -/lymphoblasts nonetheless showed a significant increase in most ceramide (Cer) species and their dihydroCer (dhCer) precursors (Fig. 5C, D). These differences were not observed in naïve OT-II WT and CCR5 -/cells (Fig. S6A), indicative that the Cer increase was specific to antigen-experienced cells. The increase in Cer species in CCR5 -/lymphoblasts was not linked to enhanced apoptosis compared to WT cells (Fig. S6B).
We sought to validate the CerS isoforms upregulated by CCR5 deficiency at the protein level. In accordance with mRNA analyses, CerS2 protein levels were significantly higher in CCR5 -/than in WT lymphoblasts (Fig. 5F); CerS3 and CerS4 were undetectable or only barely detectable by immunoblot. This is consistent with the fact that CerS2 has the highest expression level and the broadest substrate specificity in other cell types (Laviad et al., 2008). Chromatin immunoprecipitation (ChIP), followed by amplification of a region of the CerS2 promoter enriched in CpG islands, showed that binding of the transcriptional activation marker acetylated histone H3K9 (H3K9Ac) was higher in CCR5 -/than in WT lymphoblasts (Fig. 5G). Moreover, blockade of CCR5 signaling with pertussis toxin (PTx; an inhibitor of the Gαi subunit) also increased CerS2 mRNA expression (Fig. 5H).
To further study CCR5 transcriptional regulation of CerS, we scanned for transcription factors with putative binding sites in the CerS2, CerS3 and CerS4 promoters, which are transcriptionally upregulated in CCR5 -/lymphoblasts, but not represented in the CerS6 promoter, which is not CCR5-regulated. We selected two regions; region 1 comprised -5 kb to the 5'UTR, and region 2 encompassed the 5'UTR to the first coding exon (Fig. 5I). This bioinformatic approach identified GATA-1 and NF-IB (nuclear factor-1B) as putative transcription factors involved in the differential expression of the CerS2 isoform (Fig. 5J, K).
We focused on GATA-1, since it is implicated in the differentiation of some CD4 + T cell subtypes (Fu et al., 2012, Sundrud et al., 2005. Immunofluorescence analyses showed increased nuclear levels of the phosphoSer142-GATA-1 form in OT-II CCR5 -/compared to WT lymphoblasts (Fig. 5L, M), which correlated with enriched GATA-1 binding to the CerS2 promoter in CD4 + CCR5 -/lymphoblasts (Fig. 5N). These results suggest that high Cer levels in CCR5 -/lymphoblasts are a result of transcriptional induction of CerS2 through GATA-1.

Ceramide levels control TCR nanoclustering
We used a synthetic biology approach to determine whether ceramide content affects TCR nanoclustering. Large unilamellar vesicles (LUV) were prepared at different molar ratios of PC, Chol, SM and Cer (Fig.6A), and then reconstituted with a streptavidin-binding-peptide-tagged TCR purified in its native state from murine M.mz-SBP T cells (Swamy & Schamel, 2009).
The proteoliposomes were analyzed by BN-PAGE after solubilization in 0.5% Brij96 to maintain TCR nanocluster integrity, or in 1% digitonin to disrupt TCR clusters. As anticipated (Molnar et al., 2012, Wang et al., 2016, TCR was monomeric in PC-containing LUV, whereas it formed nanoclusters when reconstituted in PC/Chol/SM liposomes (Fig. 6B, C). The inclusion of ceramides in these LUV (PC/Chol/SM/Cer liposomes) reduced TCR nanoclustering in a dose-dependent manner. This effect was not due to differential TCR reconstitution in Cer-containing LUV, since digitonin treatment rendered equivalent levels of monomeric TCR in each condition (Fig. 6B). These data suggest that Cer membrane content impairs TCR nanoclustering.
To test whether this effect also occurs in live cells, we treated OT-II WT lymphoblasts with recombinant sphingomyelinase (SMase), which hydrolyzes SM to ceramide (Kitatani et al., 2008). SMase treatment of OT-II blasts increased Cer levels robustly (Fig. 6D), but did not compromise cell viability (Fig. S8). Analysis of membrane replicas from these cells showed that SMase treatment reduced the number of high valency TCR nanoclusters compared to controls (Fig. 6E), which indicates that high Cer levels hinder TCR nanoclustering in CD4 + T cells.

CerS2 silencing restores TCR nanoclustering after CCR5 functional blockade
To correlate increased CerS2 expression with the impaired TCR nanoclustering in OT-II CCR5 -/-T cells, we attempted to silence CerS2 expression by lentiviral transduction of primary lymphoblasts with short-hairpin (sh) RNA for CerS2 or control (shCtrl). In the most successful experiments, we were only able to transduce ~20% of the lymphoblasts, which did not lead to solid CerS2 mRNA silencing (Fig. S9A,B). Despite the low efficiency, antigenic restimulation tended to promote stronger responses in shCerS2-than in shCtrl-transduced cells (Fig. S9C,D).
The low efficiency also precluded analysis of TCR nanoclusters in membrane replicas, as transduced cells could not be distinguished from non-transduced cells.
To overcome these difficulties, we used the 2B4 CD4 + T cell line. We verified that 2B4 cells expressed CCR5, and that TAK-779 treatment increased CerS2 levels and impaired TCR nanoclustering (Fig. S10). The data suggest that CCR5 effects on TCR nanoclustering and CerS2 induction are not exclusive to the OT-II system, and that TAK-779-treated 2B4 cells mimic the functional findings in OT-II CCR5 -/lymphoblasts. 2B4 cells were transduced efficiently by lentiviruses and, after three days of antibiotic selection, 100% of the cells expressed the shRNA; this led to strong silencing of CerS2 mRNA and protein in shCerS2compared to shCtrl-transduced cells ( Fig. 6F-H). Analysis of TCR organization showed recovery of large TCR nanoclustering in TAK-779-treated, shCerS2-transduced cells compared to controls (Fig. 6I); after restimulation with plate-bound anti-CD3ε antibody in the presence of TAK-779, CD69 upregulation was higher in CerS2-deficient than in shCtrl-cells (Fig. 6J).

CCR5 modulates TCR nanoclustering in human CD4 + T cells
Finally, we tested whether CCR5 deficiency also impairs TCR organization in human CD4 + T cells. Approximately 1% of the Spanish population bears the ccr5Δ32 polymorphism in homozygosity (Mañes et al., 2003). Purified CD4 + T cells from healthy WT or ccr5Δ32 homozygous donors were activated with anti-CD3 and -CD28 antibodies for three days, and 14 maintained for five additional days with IL-2. We found that ccr5Δ32 lymphoblasts had a lower percentage of large TCR nanoclusters than WT cells (Fig. 7A); concomitantly, the fraction of monomeric TCR was increased in the former. Sphingolipid analysis of these CD4 + lymphoblasts showed an increase in saturated 24-carbon Cer (C24:0) and its precursor (dhCer C24:0) in cells derived from ccr5Δ32 donors, whereas SM levels were comparable between both genotypes (Fig. 7B). This increase in Cer levels was associated with upregulation of CerS2 mRNA in ccr5Δ32 lymphoblasts compared to WT controls (Fig. 7C); expression of other enzymes involved in Cer metabolism was unchanged in both genotypes (Fig. S11). As found in mouse CCR5 -/lymphoblasts, antigen-experienced human CD4 + T cells from ccr5Δ32 homozygotes also show defective TCR nanoclustering associated with increased Cer levels and upregulated CerS2. Moreover, they indicate that these CCR5 effects are not restricted to specific T cell clones, but can be observed in a polyclonal T cell repertoire.

DISCUSSION
Here we show that CCR5 signaling is largely dispensable for memory CD4 + T cell differentiation, but provides specific signals that improve the functional fitness of memory cells after antigen re-encounter. The CCR5 signals optimize TCR nanoclustering and antigen sensitivity by triggering a CD4 + T cell-specific transcription program that regulates Cer metabolism. This CCR5 program operates in murine and human CD4 + T cells, which suggests physio-pathological relevance.
A central observation of our study is that CCR5 expression enhances the degree of TCR nanoclustering in resting antigen-experienced T cells, both in vitro and in vivo. The presence of TCR nanoclusters in resting T cells was shown by BN-PAGE, EM, and super-resolution microscopy (Hu et al., 2016, Jung et al., 2016, Pageon et al., 2016, Schamel et al., 2005. We demonstrate here differential TCR nanoclustering in WT and CCR5-deficient cells using two complementary approaches (EM and BN-PAGE), based on different conceptual principles. In EM, TCR nanoclusters were defined as gold particle aggregates at less than 10 nm distance from one another. Previous analyses showed that this criterion permits identification of TCR nanoclusters formed by TCR-TCR interactions (Kumar et al., 2011); these tightly-associated TCR nanoclusters would allow inter-TCR cooperativity for pMHC binding (Martín-Blanco et al., 2018). We therefore intentionally considered TCR not to be in the same nanocluster if the gap between them was >10 nm; this excludes considering more loosely associated TCR as nanoclusters, but the strict definition allowed association of TCR nanoclusters to a T cell biological function.
It is also important to clarify that gold particle counts do not necessarily correspond to the number of TCR molecules in a nanocluster. BN-PAGE defines neither the exact size nor the abundance of TCR nanoclusters. Direct comparison of CCR5 -/with WT cells using both methods nonetheless allowed us to detect relative differences and determine the promoter effect of CCR5 in TCR nanoclustering. Application of Monte Carlo simulations further indicated that the nanoclusters observed in EM are not the result of random proximity of gold particles. The estimated clustering parameter (b) for randomly distributed particles was virtually zero.
Since CCR5 provides positive signals during activation of naïve CD4 + T cells (González-Martín et al., 2011, Molon et al., 2005, Nesbeth et al., 2010, we attempted to clarify whether TCR nanoclustering impairment in CCR5 -/cells is solely an effect of this defective priming. This is unlikely, since TCR clustering was inhibited when CCR5 was blocked during expansion of fully-activated WT lymphoblasts. This effect on TCR nanoclustering during lymphoblast expansion was modest compared to that observed in the priming phase, but is probably the result of insufficient CCR5 inhibition during lymphoblast expansion. CCR5 is not only upregulated shortly after activation, but is maintained in memory CD4 + T cells, which are highly susceptible to infection by R5-tropic HIV-1 strains (Nie et al., 2009). Our results showed increased CCR5 mRNA expression during lymphoblast expansion. These lymphoblasts also expressed CCR5 ligands, suggesting autocrine CCR5 stimulation during this phase. We thus propose that TCR nanoclustering is regulated by CCR5 signals transduced during lymphoblast differentiation rather than during priming.
Another feature that distinguishes CCR5 effects on priming and on TCR nanoclustering is the role of CXCR4 in these events. During priming, CCR5 and CXCR4 are recruited to and accumulate as heterodimeric complexes at the IS of CD4 + T cells; AMD3100 (a CXCR4 antagonist) prevented not only CXCR4 but also CCR5 accumulation (Contento et al., 2008), which indicates necessary cooperation between CCR5 and CXCR4 for full T cell activation. In contrast, AMD3100 did not affect TCR nanoclusters in CCR5-expessing cells, which suggests that CXCR4/CCR5 heterodimer signaling is not essential for TCR nanoclustering in lymphoblasts. CCR5 homo-and heterodimers are thought to associate differently to Gα subunits; homodimers signal through the PTx-sensitive Gαi, whereas heterodimers generate PTx-resistant responses (Mellado et al., 2001). CCR5 costimulatory signals in the IS are PTxresistant (Molon et al., 2005), consistent with CCR5/CXCR4 heterodimerization during priming. PTx potentiates transactivation of the CerS2 promoter (Fig. 5H), however, which suggests involvement of the CCR5-induced Gαi pathway in TCR nanoclustering. CCR5 thus appears to trigger distinct signaling pathways for co-stimulation and TCR nanoclustering in CD4 + T cells.
Cholesterol and SM are two lipids essential for CCR5 signaling and TCR nanoclustering (Mañes et al., 2001, Molnar et al., 2012. In resting T cells, these receptors nonetheless partition in different membrane phases, liquid-ordered (lo) for CCR5 (Molon et al., 2005) and liquiddisordered (ld) for TCR (Beck-Garcia et al., 2015). This differential phase segregation argues against direct CCR5/TCR interaction as a mechanism that influences TCR nanoclustering. Our results suggest instead that increased levels of long-chain Cer species cause defective TCR nanoclustering in CCR5 -/lymphoblasts. Indeed, elevation of Cer levels in TCR-reconstituted proteoliposomes and in live cells by SM hydrolysis impaired nanoscopic TCR organization.
Although Cer levels were higher in lymphoblasts than in naïve cells, which supports a role for Cer in T cell activation (Sofi et al., 2017), CCR5 deficiency further increased Cer levels specifically in lymphoblasts. The Cer increase in CCR5 -/lymphoblasts did not cause spontaneous apoptosis (Fig. S6B), which coincides with the non-apoptotic and preventive effects of long-chain Cer in this process (Stiban & Perera, 2015). We propose that in activated CD4 + T cells, autocrine CCR5 activation downgrades the nuclear entry and binding of activated GATA-1 to specific CerS promoters, such as that of CerS2; this restrict Cer biosynthesis and provide a membrane environment permissive for TCR nanoclustering (Fig. 8). Given that WT and CCR5 -/lymphoblasts have similar SM levels, our results pinpoint de novo Cer biosynthesis as an important metabolic checkpoint for CD4 + T cell memory function.
Several settings can be hypothesized that explain Cer effects on TCR nanoclustering. In model membranes, Cer have strong segregation capacity, which might affect their lateral organization (Goñi & Alonso, 2009). When the Chol concentration is saturating, however, as is the case of cell membranes and our proteoliposomes, Cer-enriched domains are not formed, due to the ability of Chol and Cer to displace one another (Castro et al., 2009). This argues against the idea that Cer impairs TCR clustering by promoting a general reduction in membrane lateral diffusion. Our mathematical model also predicted that cluster distribution is independent of TCR diffusivity. Interaction of the TCRβ subunit with Chol/SM complexes is critical for TCR nanoclustering (Molnar et al., 2012, Wang et al., 2016. High levels of Cer or their precursors (dhCer) could transform Chol/SM into Chol/SM/Cer domains with specific physicochemical properties, which might hinder TCR nanocluster formation physically or thermodynamically. For instance, high dhSM levels inhibit CCR5-mediated HIV-1 infection by rigidifying CCR5-containing lo domains (Vieira et al., 2010). Chol/SM interaction depends on the hydrogen bond generated by the amide group of the SM molecule and the 3-hydroxyl group of Chol (Ramstedt & Slotte, 2002), but the SM amide group can also form hydrogen bonds with the Cer hydroxyl group (Garcia-Arribas et al., 2016). It is thus possible that, rather than forming SM/Chol/Cer domains, small increases in Cer levels would increase the mutual displacement of three lipids. This could lead to replacement of SM/Chol by SM/Cer complexes with gel-like biophysical properties (Sot et al., 2008).
Our results indicate that the CCR5 effect on TCR clustering is biologically meaningful. In a first model, we show that the responses of CCR5-deficient memory CD4 + T cells generated by vaccination were impaired after ex vivo stimulation. In a second model that involves T:B cell cooperation, we show that CCR5 deficiency impaired class switching of high affinity antibodies after re-exposure to a T cell-dependent antigen. Affinity maturation and class switching depend on recruitment of Tfh cells to GC (Vinuesa et al., 2016). This Tfh cell confinement is a result of CXCR5 expression and the downregulation of other homing receptors, including CCR5 (Crotty, 2011), which could explain the lack of difference in class switching between OVA/KLH-immunized WT and CCR5 -/mice. There were also no differences between WT and CCR5 -/effector Tfh cells after the first OVA immunization. Once GC resolve, however, some Tfh cells are reported to enter the circulation as Tfh central memory-like cells (Vinuesa et al., 2016). These circulating, antigen-experienced Tfh cells express CCR5 and are very susceptible to HIV-1 infection (Xu et al., 2017). We found that the frequency of large TCR nanoclusters increased in memory T cells from WT compared to CCR5 -/-OVA/OVA-immunized mice, which suggests increased antigenic sensitivity.
We hypothesize that following re-exposure to antigen, CCR5-expressing memory pre-Tfh cells will have a more efficient response than CCR5-deficient cells, which would support robust antibody responses after their differentiation to GC-Tfh cells. In humans, functional CCR5 deficiency does not cause strong immune suppression, but ccr5Δ32 homozygosity was associated with four times more fatal infections than average during the 2009-2011 influenza season in Spain (Falcon et al., 2015) and fatal infections by the West Nile Virus in the US (Lim & Murphy, 2011). Our results provide a conceptual framework on which to base clinical trials to evaluate CD4 + T cell memory responses in CCR5-deficient humans, and suggest caution regarding the risks associated to genetic ablation of CCR5 as a preventive strategy to block HIV-1 infection.

MATERIAL AND METHODS
Resource Identification Portal accession numbers for antibodies, cell lines, animals and other reagents used in the study are provided as supplementary material.

Antibodies and reagents
Antibodies used for characterization of mouse cells by flow cytometry were anti-Vα2TCR-PE  (rs333) was genotyped by PCR (AriaMx Real-time; Agilent Technologies) as described (Mañes et al., 2003).

Flow cytometry
For cell surface markers, cell suspensions were incubated (20 min, 4ºC) with the indicated fluorochrome-labeled or biotinylated monoclonal primary antibodies in phosphate-buffered saline with 1% BSA and 0.02% NaN3 (PBS staining buffer). For intracellular labeling, cells were fixed and permeabilized with IntraPrep (Beckman Coulter), followed by intracellular staining with indicated antibodies. Cells were analyzed on Cytomics FC500 or Gallios cytometers (both from Beckman-Coulter) and data analyzed using FlowJo software.

Immunization and adoptive transfer
Spleen and lymph node cell suspensions from OT-II WT or CCR5 -/cells were adoptively transferred (5 x 10 6 cells/ mouse) into CD45.1 mice. The following day, recipient mice were infected intravenously with rVACV-OVA virus (2 x 10 6 pfu). Mice were sacrificed 35 days later, and splenocyte suspensions obtained as described above.

Immunogold labeling, replica preparation, and EM analysis
Immunogold-labeled cell surface replicas were obtained as described (Kumar et al., 2011).
Briefly, T cells were fixed in 1% paraformaldehyde (PFA) and labeled with anti-mouse CD3 mAb (145-2C11) or anti-human CD3 mAb OKT3, followed by 10 nm gold-conjugated protein A (Sigma-Aldrich). Labeled cells were adhered to poly-L-lysine-coated mica strips and fixed with 0.1% glutaraldehyde. Samples were covered with another mica strip, frozen in liquid ethane (KF-80, Leica), and stored in liquid nitrogen. Cell replicas were prepared with a Balzers 400T freeze fracture (FF) unit, mounted on copper grids, and analyzed on a JEM1010 electron microscope (Jeol, Japan) operating at 80 kV. Images were taken with a CCD camera (Bioscan, Gatan, Pleasanton, CA) and processed with TVIPS software (TVIPS, Gauting, DE). Gold particles were counted on the computer. When distance between gold particles was smaller than their diameter (10 nm), they were considered part of the same cluster.

Preparation of large unilamellar vesicles (LUV) and TCR reconstitution
LUV with a custom lipid composition were prepared by the thin film method (Molnar et al., 2012), followed by extrusion through polycarbonate membranes with a pore size of 200 nm (21 times) and 80 nm (51 times). The diameter of the resulting LUV was determined by dynamic light scattering (Zetamaster S, Malvern Instruments). The LUV preparation (2 mM) was mixed with purified TCR (0.1 µg) in 100 µl saline-phosphate buffer with 0.02% Triton X-100, and 40 µl 0.01% Triton X-100 was added. Samples were agitated (30 min, 4°C), and detergent removed by adsorption to polystyrene BioBeads SM-2 (3 mg; Bio-Rad; overnight, 4°C).

SMase treatment
All experiments (sphingolipid quantification, apoptosis, and TCR nanoclustering) were performed by incubating OT-II WT and CCR5 -/cells (0.2 x 10 6 ) with recombinant sphingomyelinase from Bacillus cereus (0.5 U/ml; 1 h, 37ºC) in serum-free medium. Cells were washed and processed immediately for EM analysis or for sphingolipid quantification as above.

Quantitative RT-PCR analyses
Total RNA was extracted from human or murine cells using the RNeasy Mini Kit (Qiagen), and cDNA synthesized from 1 µg total RNA (High Capacity cDNA Reverse Transcription Kit, Promega
Transduction efficiency was analyzed after 24 h by FACS. In the case of 2B4 cells, transduced cells were selected with puromycin (2 µg/ml) for 3 days prior to analyses.

Cell migration assay
OT-II WT or CCR5 -/lymphoblasts (10 6 ) were added to the upper chamber of a transwell (3 µmpore; Corning) and allowed to migrate towards 100 mM CCL4 for 4 h. Migrating cells were quantified by flow cytometry (Cytomics FC500).

Mathematical and Bayesian analyses
To analyze cluster size distribution, we used a standard χ 2 test to compare the fraction of clusters of a given size (1, 2, 3, etc.) in each dataset. In all plots, "Random" refers to synthetic distributions of receptors generated randomly.
To quantify the mechanistic relevance of cluster size between random distributions of clusters and clusters in WT and in CCR5 -/-CD4 + T cells, we used a Bayesian inference model on top of a mechanistic model (Castro et al., 2014). The model assumes that TCR aggregates by incorporating one receptor at a time, with on and off rates that depend on the diffusion properties of the receptor on the membrane, but not existing cluster size. That is, The "affinity" of the process is given by b = q+/q-, which we also refer to as the clustering or affinity parameter. In the steady state, we can calculate analytically the fraction of clusters of a given size n: with The model was fitted using the Bayesian JAGS code (Kruschke, 2014)  At the molecular level, the kinetic rates q + and qcan be expressed in terms of the diffusion rates of the receptors, kd +/-, the receptor size (a), the mean distance between receptors (s), and the correct receptor-receptor binding rates, k +/-, through the equations (Lauffenburger & DeLisi, 1983): with The clustering parameter b is independent of the TCR diffusivity (as D is canceled), so the observed TCR nanoclustering differences for WT and CCR5 -/cells would be due to TCR-TCR interactions, as previously reported (Beck-Garcia et al., 2015).

Statistical analyses
For comparison between two conditions, data were analyzed using parametric Student's t-tests, unpaired, or paired when different treatments were applied to the same sample. Multiple parametric comparisons were analyzed with one-way ANOVA with Bonferroni's post-hoc test.
The χ 2 test was used to analyze overall distribution of gold particles. All analyses were performed using Prism 6.0 or 7.0 software (GraphPad). Differences were considered significant when p <0.05.

Data availability
The authors confirm that all relevant data and materials supporting the findings of this study are available on reasonable request. This excludes materials obtained from other researchers, who must provide their consent for transfer.

Code availability
The Bayesian JAGS code generated in the study is provided as supplementary information.
Online supplemental material Figure S1 shows the expression of CCR5 ligands and CCR5 in lymphoblasts, the functionality of CCR5 in these cells, as well as the analysis of TCR nanoclustering in OT-II WT lymphoblasts expanded in IL-2 or IL-15. Fig. S2 shows the comparison of experimental and synthetic TCR nanocluster distribution. Fig. S3 shows a representative FACS analysis of CD4 + memory T cells purified by negative selection from OVA/OVA-immunized WT and CCR5 -/mice. Fig.   S4 shows the lack of effect of a CCR5 antagonist (TAK-779) on TCR nanoclustering in CCR5deficient lymphoblasts. Fig. S5 shows that CCR5 deficiency does not affect TCR expression in in vitro-generated lymphoblasts or in vivo isolated memory CD4 + T cells. Fig. S6 shows sphingolipid levels in naïve WT and CCR5-deficient OT-II cells, and representative dot plots of WT and CCR5-deficient OT-II cells stained with apoptotic markers. Fig. S7 shows relative mRNA expression levels of enzymes involved in ceramide metabolism. Fig. S8 shows that SMase treatment does not induce apoptosis. Fig. S9 shows the effect of CerS2 silencing in primary CD4 + lymphoblasts. Fig. S10 shows that CCR5 blockade with TAK-779 impairs TCR nanoclustering in the CD4 + T cell hybridoma 2B4. Fig. S11. shows relative mRNA expression levels of enzymes involved in ceramide metabolism in primary human CD4 + lymphoblasts. Table S1 shows the list of primers used in the study. Table S2 shows the Resource Identification Portal accession numbers or catalog numbers of the reagents used in the study. It is also provided as supplemental material the Bayesian code for the R-language used in the study.

AUTHORS' DECLARATION OF INTERESTS
The authors declare no competing interests