Single-cell transcriptomics reveals expansion of cytotoxic CD4 T cells in supercentenarians, Hacker News

Significance

Exceptionally long-lived people such as supercentenarians tend to spend their entire lives in good health, implying that their immune system remains active to protect against infections and tumors. However, their immunological condition has been largely unexplored. We profiled thousands of circulating immune cells from supercentenarians at single-cell resolution and identified CD4 T cells that have cytotoxic features. This characteristic is very unique to supercentenarians, because generally CD4 T cells have helper, but not cytotoxic, functions under physiological conditions. We further profiled their T cell receptors and revealed that the cytotoxic CD4 T cells were accumulated through clonal expansion. The conversion of helper CD4 T cells to a cytotoxic variety might be an adaptation to the late stage of aging.

Abstract

Supercentenarians , people who have reached 110 y of age, are a great model of healthy aging. Their characteristics of delayed onset of age-related diseases and compression of morbidity imply that their immune system remains functional. Here we performed single-cell transcriptome analysis of 61, 202 peripheral blood mononuclear cells (PBMCs), derived from 7 supercentenarians and 5 younger controls. We identified a marked increase of cytotoxic CD4 T cells (CD4 cytotoxic T lymphocytes [CTLs]) as a signature of supercentenarians. Furthermore, single-cell T cell receptor sequencing of 2 supercentenarians revealed that CD4 CTLs had accumulated through massive clonal expansion, with the most frequent clonotypes accounting for 15 to 35% of the entire C D4 T cell population. The CD4 CTLs exhibited substantial heterogeneity in their degree of cytotoxicity as well as a nearly identical transcriptome to that of CD8 CTLs. This indicates that CD4 CTLs utilize the transcriptional program of the CD8 lineage while retaining CD4 expression. Indeed, CD4 CTLs extracted from supercentenarians produced IFN-γ and TNF-α upon ex vivo stimulation. Our study reveals that supercentenarians have unique characteristics in their circulating lymphocytes, which may represent an essential adaptation to achieve exceptional longevity by sustaining immune responses to infections and diseases.

We profiled fresh PBMCs derived from 7 supercentenarians (SC1 – SC7) and 5 controls (CT1 – CT5, aged in their 50 s to 80 s) by using droplet-based single-cell RNA sequencing technology ( (× Genomics) (26,27) (Fig. 1 (A)andSI Appendix, Fig. S1 (A)). The total number of recovered cells was 61, 202 comprising 41, 208 cells for supercentenarians (mean: 5, 887 cells) and 19, 994 cells for controls (mean: 3, 999 cells), which is in the normal range of median gene and unique molecular identifier (UMI) counts per cell reported in the (XQC database)http: // 10 xq c.com/index.html) (Fig. 1 (B)andSI Appendix, Fig. S1 (B)). Based on their expression profiles, we visualized the cells in 2D space using t-distributed stochastic neighbor embedding (tSNE), a method for nonlinear dimensionality reduction. Using aK– means clustering algorithm, we found (distinct clusters representing different cell types)Fig. 1 (C)and(SI Appendix) , Fig. S1 (C) andD). We identified the major cell types comprising PBMCs, including: T cells (TC1 and TC2 clusters) characterized by (CD3) and T cell receptor (TRAC) expression; B cells (BC cluster) characterized by (MS4A1) ( (CD) ) and (CD) expression; natural killer cells (NK cluster) characterized by (KLRF1) expression; 2 subsets of monocytes (M 14 and M 16 clusters) characterized by (CD) andFCGR3A( (CD) ) expression, respectively; and erythrocytes (EC cluster) characterized byHBA1(hemoglobin alpha locus 1) expression (Fig. 1 (D)andSI Appendix, Fig. S1 (E)). We also found 3 small clusters, annotated as MKI 67proliferating cells (MKI, marker of proliferation Ki – 67 positive), dendritic cells (DCs), and megakaryocytes (MGKs), based on the expression of established marker genes (S Appendix, Fig. S1 (F)). Although there are some batch effects leading to local enrichment of specific libraries on tSNE plots (SI Appendix , Fig. S1 (D)), all of the 10 clusters are not library specific, but consisted of cells from more than 11 Different donors.

Significant Reduction of B Cells.

In previous FACS analyzes using cell-surface markers, various age-associated population changes were observed in human PBMCs, such as B cell reduction (15) and loss of naïve CD8 T cells (18). To understand whether supercentenarians follow the common population changes, we compared the percentages of the immune cells in PBMCs between the supercentenarians and controls. Among the identified cell types in our single-cell transcriptome analysis, B cell numbers were significantly increased in the supercentenarians compared with the controls ( (P)=0.00 25, Wilcoxon rank sum test ( (Fig. 2) A). The median percentage of B cells in the 7 supercentenarians (2%) was far below that in the controls (11%) and the reference values ​​reported in a previous cohort study (28); in contrast, the populations of the other cell types were relatively stable and did not significantly change compared with the controls ( (Fig. 2) ************************************************************************************************************ (A) and(SI Appendix) , Fig. S2A). The reduction of B cells was validated by FACS analysis of 4 supercentenarians (SC1 – SC4) and 3 controls (CT1 – CT3), which showed low levels of CD3^–and CD 19 ^{(B cell populations in supercentenarians)Fig. 2 (B)andSI Appendix, Fig. S2 (B)). We also confirmed that the percentages of major cell types (B cells, T cells, natural killer cells, and CD 14monocytes) in PBMCs were consistent with those measured by FACS using canonical markers ( (Fig. 2) ************************************************************************************************************ (C) andSI Appendix, Fig. S2 (B)). We further clustered the B cells into 3 distinct subtypes (BC1, BC2, and BC3) by usingK– means clustering (SI Appendix, Fig. S2 (C)). BC1 corresponds to naïve B cells due to the presence of (IGHD) , an Ig isotype expressed before class switching, and absence of the activation markerCD 27. BC2 corresponds to quiescent memory B cells, characterized by expression of (CD) , (IGHG1) , and (IGHA1) (SI Appendix, Fig. S2 (D)). BC3, which accounts for a small fraction, albeit one with contributions from all donors, shows distinct features of plasma cells such as high levels of immunoglobulins (IGHAand (IGHG) ), expression of (CD) , and loss of (MS4A1) ( (CD) ) (SI Appendix, Fig. S2 (D) andE). Among these 3 B cell subtypes in PBMCs, the percentage of naïve B cells was significantly lower in supercentenarians compared with the controls ( (P) ************************************************************************************************************* (=0.0) , Wilcoxon rank sum test), and the percentage of memory B cells also tended to be lower in supercentenarians but the difference was not significant ( (P)=0. 073) (SI Appendix, Fig . S2 (F)).}

Expansion of Cytotoxic T Cells in Supercentenarians.

In contrast to the profound reduction of B cells, the T cell fraction remained stable at around 40% of PBMCs according to both the transcriptome data (TC inFig. 2 (A)) and the FACS analysis (CD3 ^{(CD)–inFig. 2 (C)). However, 2 T cell clusters, TC1 and TC2, were imbalanced between supercentenarians and controls: TC1 was significantly diminished ( (P)=0.00 25, Wilcoxon rank sum test), whereas TC2 was significantly expanded (P=0.00 25) in supercentenarians (Fig. 3 (A)). To better understand this T cell-specific population shift, we extracted all of the cells from TC1 and TC2 for further analysis using the Seurat R package (version 2.3.0) (29). A clustering algorithm based on shared nearest neighbor modularity optimization implemented in Seurat produced 2 major clusters: Seurat_TC1 and Seurat_TC2, corresponding to the original TC1 and TC2 clusters ( (Fig. 3) ************************************************************************************************************ (B) andSI Appendix (Fig. S3) ************************************************************************************************************ (A)). We then compared these 2 clusters and identified 332 differentially expressed genes, of which the most significant gene distinctively expressed in Seurat_TC2 was (NKG7) , a component of granules in cytotoxic lymphocytes. In addition, the top 20 most significant genes included multiple genes encoding cytotoxic effector molecules responsible for the perforin / granzyme apoptosis pathway, such asGZMH,GZMB, (GZMA) , andPRF1(Fig. 3 (C)andSI Appendix, Fig. S3 (B)). In contrast, Seurat_TC1 was characterized by expression of (CCR7) andSELL(encoding CD (L), which are required for lymph node migration (SI Appendix, Fig. S3 (C)). These genes are normally expressed in naïve and central memory T cells, but not in cytotoxic effector memory T cells (30), indicating that the primary factor separating the 2 clusters is cytotoxicity. Perforin / Granzymecells were predominantly found in the supercentenarians ( (Fig. 3) ************************************************************************************************************ (D) ), whereas CCR7noncytotoxic cells were more abundant in the controls (SI Appendix , Fig. S3 (D)). We then examined how many of the 4 cytotoxic genes ( (GZMH) , (GZMB) , (GZMA) , and (PRF1) ) showed detectable expression in each single cell. As expected, for both the supercentenarians and controls, the vast majority of cells in the noncytotoxic cluster (Seurat_TC1) expressed either 0 or 1 cytotoxic gene (s)  ( (Fig. 3) ************************************************************************************************************ (E) ,Left). In the cytotoxic cluster (Seurat_TC2), cells that expressed all 4 genes were abundant in supercentenarians but rare in controls, indicating that the level of cytotoxicity per cell might be higher in supercentenarians ( (Fig. 3) ************************************************************************************************************ (E) , (Right) ). Cytotoxic T cells were significantly expanded in supercentenarians ( (P)=0.00 25, Wilcoxon rank sum test), reaching 80% of T cells in some individuals (Fig. 3 (F)). This was in sharp contrast to controls where cytotoxic T cells made up ∼ 10 to 20% of the total T cell population.}

Expansion of Cytotoxic CD4 T Cells in Supercentenarians.

In general, cytotoxic T cells are CD8and noncytotoxic helper T cells are CD4, with both being derived from double positive thymocytes (31). Therefore, a simple interpretation of our results is that there is an increase in CD8T cells in supercentenarians. However,CD8Aand (CD8B) , which encode the 2 components of CD8, were expressed only in a subset of cytotoxic T cells, whereas (CD4) and (TRDC) (T cell receptor delta constant) were expressed in the other subsets, suggesting the presence of 3 subsets of cytotoxic T cells: CD8 cytotoxic T lymphocytes ( CTLs), CD4 CTLs, and cells T cells (Fig. 4 (A)). To investigate cytotoxic T cells other than CD8 CTLs, we manually defined CD4 CTLs and γδ T cells based on ranges of (CD4) ,CD8, andTRDCexpression ( (Fig. 4) ************************************************************************************************************ (A) , (Bottom Right) and(SI Appendix) , Fig. S4 (A)). Previous studies reported that CD4 CTLs account for a tiny fraction of CD4T cells in PBMCs (eg, mean 2.2% in (healthy donor) (32)). Here, the supercentenarians show significantly higher levels of CD4 CTLs (mean, 25 .3% of total T cells (than in the controls (mean, 2.8%) ( (P)=0.00 25, Wilcoxon rank sum test), as well as higher levels of CD8 CTLs than in the controls ( (P)=0.00 25), whereas the population of γδ T cells was average in size and comparable to that in the controls ( (P)=0.2) (Fig. 4 (B)andSI Appendix, Fig. S4 (B)). To validate the expansion of CD4 CTLs, we performed FACS analysis of 6 supercentenarians (SC1 and SC5 – SC7 [studied above] and SC9 and SC 10), 1 semisupercentenarian (over (y old; SC8), and 5 controls (CT4 and CT5 [studied above] and CT6 – CT8) (S Appendix, Fig. S1 (A)) using antibodies against CD3, CD4, CD8, and GZMB. According to the CD4 / CD8 staining profile (gated on CD3), the T cells in the supercentenarians were not predominantly CD8 ^{(T cells)Fig. 4 (C)andSI Appendix , Fig. S4 (C)). We then asked how many of the CD4T cells retained in supercentenarians were cytotoxic by using the CD4 / GZMB staining profile. Remarkably, CD4 (GZMB) T cells were quite abundant in the supercentenarians, in which at least 10% (mean, 30. 1% (of T cells are CD4 CTLs in all tested supercentenarian samples ( (n)=7) (Fig. 4 (D)). The percentages of CD4 CTLs (CD4 (GZMB) T cells) in the total T cell populations were significantly higher in the centenarians than in the controls (P=0.0 18, Wilcoxon rank sum test) (Fig. 4 (E)andSI Appendix, Fig. S4 (D)). Furthermore, GZMBcells were more abundant than GZMB–cells in both CD4 and CD8 T cell populations in 5 out of 7 tested (semi) supercentenarians but none of the controls, indicating expansion of CD4 CTLs as well  as CD8 CTLs (SI Appendix, Fig. S4 (E)). The percentages of CD4 CTLs correlated well between single-cell RNA-Seq and FACS analyzes according to the comparison of the 6 commonly analyzed samples (4 supercentenarians and 2 controls) ( (Fig. 4) F). Thus, the high level of CD4 CTLs in supercentenarians was supported by 2 independent methods. Finally, we assessed protein levels of 2 cytotoxic molecules, perforin and granulysin, together with granzyme B in 1 of the supercentenarians (SC2) using FACS. According to the GZMB / PRF1 and GZMB / GNLY staining profiles (gated on live CD3 (CD4) CD8–), the CD4 (GZMB) T cells were predominantly perforin positive, but not necessarily granulysin positive ((SI Appendix) , Fig. S4 (F)), suggesting that the composition of cytotoxic granules might be different in The CD4GZMBPopulation.}

Analysis of T Cell Subsets.

The Seurat R package (version 2.3.0) was used to analyze T cell subsets (TC1 and TC2). The outputs ofcellranger countwere loaded using theRead (X) function. Cells clustered in TC1 and TC2 by the Cell Ranger analysis pipelines were extracted, and principal components were calculated usingRunPCAfunction. The first 16 principal components, based on the manual inspection of the elbow plot (PCElbowPlot), were used for cell clustering (using theFindClustersFuncti on with resolution 0. 05) and tSNE visualization (usingRunTSNE). Differentially expressed genes were identified using theFindAllMarkersfunction, and the top 20 genes were visualized in a heatmap using the (DoHeatmap) function. CD4 CTL, CD8 CTL, and γδ T cell clusters were manually defined in the interactive mode of the tSNE plot by using theTSNEPlotfunction with do .identify=TRUE, based on the expression of marker genes. Wilcoxon rank sum test was applied to compare percentages of T cell subtypes between the supercentenarians and controls using the wilcox.test function in R.

Analysis of T Cells in Young Donors from a Public Dataset.

The single-cell RNA-Seq data of PBMCs derived fro M 45 healthy donors were generated and released by van der Wijst et al. (33). Freely available UMI count data for T cells were downloaded and visualized in 2D space (tSNE) based on the expression profile using the Seurat R package (version 2.3.0). Age information of individual donors was extracted from Supplementary Table 8 of the original paper (32).

Pseudotime Analysis.

Monocle 2 (version 2.4.0) was used to estimate a pseudotemporal path of T cell differentiation (35). Cells clustered in TC1 and TC2 by Cell Ranger analysis pipelines were loaded to create a Monocle object using thenewCellDataSetfunction implemented in Monocle 2. The cells were ordered in pseudotime along a trajectory usingreduceDimensionwith the DDRTree method andorderCellsfunctions. Mean log-normalized expression values ​​of selected marker genes were calculated for each bin from 0 to 12 pseudotime points.

Antibodies and Flow Cytometric Analysis.

Cryopreserved PBMCs were thawed and suspended in FACS buffer (1 × Hank’s balanced salt solution with 2% FBS and 0.2% NaN3). Monoclonal antibodies specific for human CD3ε (UCHT1 and HIT3a), CD4 (RPA-T4), CD8 (RPA-T8), CD 19 (HIB 19), CD 14 (M5E2), CD 16 (B 73 1), CD (B 159), GzmB (GB 11), Perforin (δG9), and Granulysin (RB1), were purchased from BD Pharmingen. Cell numbers were counted using a Countess II Automated Cell Counter. For intracellular staining, cells were fixed and permeabilized with IntraPrep Permeabilization Reagent (Beckman Coulter) according to the manufacturer’s protocols. Cells were analyzed using FACSAria III and FACSAria SORP cell sorters (BD Biosciences) with FlowJo Software (version 10 4.2).

RosetteSep Human CD4T Cell Enrichment Mixture with SepMate – 15 (STEMCELL Technologies) was used to remove non-CD4T cells from fresh whole blood. A single-cell transcriptome library was prepared from the enriched CD4T cells by using the Chromium Single Cell 5ʹ Library Kit (10 × Genomics (with 50 ng of cDNA amplified product. A single-cell TCR library was prepared using Chromium Single Cell V (D) J Enrichment Kits, Human ( (× Genomics). The libraries were sequenced with paired-end 150 – bp reads on the Illumina HiSEq. 2500 platform. Analysis pipelines in Cell Ranger version 3.0.2 (updated version was used for the 5 ′ single-cell and TCR libraries from version 2.1.0 used for the 3 ′ single-cell libraries) were used for the sequencing data processing. TCR data were processed by running cellranger vdj with – reference=refdata-cellranger-vdj-GRCh 38 – alts-ensembl -2.0.0 to assemble TCR alpha and beta chains and determine clonotypes. Transcriptome data were processed by running cellranger count with – transcriptome=refdata-cellranger-GRCh 38 – 1.2.0. The Seurat R package (version 2.3.0) was used for cell clustering (FindClusters) and tSNE visualization (RunTSNE). Three control datasets of T cell clonotypes analyzed by the same 10 × Genomics kits were downloaded from the 10 × Genomics websites below (a simple registration is needed).

• T cells:http: // cf. 10 xgenomics.com/samples/cell-vdj/3.0.0/vdj_v1_hs_pbmc2_t/vdj_v1_hs_pbmc2_t_clonotypes.csv

• CD4 T cells:http: // cf. 10 xgenomics.com/samples/cell-vdj/2.2.0 /vdj_v1_hs_cd4_t/vdj_v1_hs_cd4_t_clonotypes.csv

• CD8 T cells :http: // cf. 10 xgenomics.com/samples/ cell-vdj / 2.2.0 / vdj_v1_hs_cd8_t / vdj_v1_hs_cd8_t_clonotypes.csv

Cell Culture and Stimulation with PMA and Ionomycin.

(Cryopreserved PBMCs and CD4) T cells were thawed at 37 ° C and washed once with X-VIVO (Lonza). The cells were centrifuged at 500 ×gfor 5 min, resuspended in X-VIVO 20 at a concentration of 1.0 × 10⁶cells / mL, and incubated at 37 ° C in 5% CO (2) ***************************************************************************************************************************************************************************************************************************************************************************** (for 16 h. The cells were stimulated with cell activation mixture containing PMA and ionomycin (BioLegend) and brefeldin A (eBioscience), and incubated for 6 h. The stimulated cells were harvested, washed once with 500 ofL of PBS with 2% FBS , and treated with Human BD Fc Block (BD Biosciences). Dead cells were stained with Fixable Viability Dye eFluor (eBioscience). Cell-surface proteins were stained with specific antibodies against CD3ε (UCHT1), CD4 (RPA-T4), and CD8 (RPA-T8). The cells were fixed and permeabilized with Fixation / Permeabilization solution and BD Perm / Wash buffer (BD Biosciences). Intracellular molecules were stained with specific antibodies against GzmB (GB 11), IFN-γ (4S.B3), and TNF-α (Mab 11) or isotype controls, IgG1 × (×) ), IgG1 M (MOPC – 21), and IgG2b ((28⇓⇓⇓⇓⇓⇓⇓–36). Cells were analyzed using FACSCanto II (BD Biosciences) with FlowJo Software (version (*********************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************************. 4.2). All FACS antibodies were purchased from BD Pharmingen.

The statistical significance of differences between supercentenarians and controls were determined by a 2-sided Wilcoxon r ank sum test using the wilcox.test function in R. (P) values ​​are indicated w ith an asterisk (************************************************************************************************************ (P)

Data Availability.

Raw UMI counts and normalized expression values ​​for single-cell RNA-Seq are publicly available at (http://gerg.gsc.riken.jp/SC 2018 /. Individual sequencing read data will be available on request under the condition of approval of the ethics committee of Keio University and material transfer agreement.

Acknowledgments

We thank all the donors who participated in this study. We also thank RIKEN GeNAS for the sequencing of the single-cell libraries, Dr. Shigeki Hirabayashi for advice on collecting human blood samples, and Dr. Monique van der Wijst for advice on using their single-cell RNA-Seq dataset. This work was supported by a research grant from the Japanese Ministry of Education, Culture, Sports, Science and Technology to the RIKEN Center for Integrative Medical Sciences and research grants for Keio University Global Initiative Research Projects.

Footnotes

• Author contributions: KH , T. Sasaki, GP, AM, IT, YA, N. Hirose, and PC designed research; K.H., T.K., T.I., N. Hayatsu, Y.M., H.Y., T.T., T. Sasaki, T. Suzuki, Y.O., H.S., J.W.S., A.M., I.T., H.O., Y.A., and N. Hirose performed research; K.H., M.V., G.P., and P.C. analyzed data; and K.H. and M.V. wrote the paper.

• The authors declare no competing interest.

• This article is a PNAS Direct Submission.

• Data deposition: Raw UMI counts and normalized expression values ​​for single-cell RNA-Seq are publicly available at (http://gerg.gsc.riken.jp/SC) .

• This article contains supporting information online at www.pnas.org/lookup/suppl/doi: 10. 1073 / Pnas. 1907883116 / – /DCSupplemental.