Abstract
While cancer is an age-related disease, most studies focus on genetically engineered younger mouse models. Here, we uncover how cancer develops as a consequence of the naturally aged immune system in mice. B-cell lymphoma frequently occurs in aged mice and is associated with increased cell size, splenomegaly, and a novel clonal B-cell population. Age-emergent B cells clonally expand outside of germinal centers driven by somatic mutations, activated c-Myc and hypermethylated promoters, and both genetically and epigenetically recapitulate human follicular and diffuse-large B-cell lymphomas. Mechanistically, mouse cancerous B cells originate from age-associated B cells, which are atypical memory B cells. Age-associated B cells secrete a spectrum of proinflammatory cytokines and activate paracrinally the expression of c-Myc in surrounding B cells. Although clonal B cells are a product of an aging microenvironment, they evolve being self-sufficient and support malignancy when transferred into young mice. Inhibition of mTOR and c-Myc attenuates premalignant changes in B cells during aging and emerges as a therapeutic strategy to delay the onset of age-related lymphoma. Together, we uncover how aging generates cancerous B cells, characterize a model that captures the origin of spontaneous cancer during aging and identify interventions that may postpone age-associated lymphoma.
Introduction
The risk of developing invasive cancer grows exponentially with age starting from midlife1. Although cancer occasionally develops at a young age, the median age of its diagnosis in the US is 66 years old2. Mortality rates for cancer and aging follow the same trend consistent with the idea that cancer is a disease of aging3. It is also well established that both normal tissues and cancers accumulate age-related mutations during aging4–6. In particular, clonal hematopoiesis of indeterminate potential (CHIP) was described as a link between aging and cancer of immune cells7,8.
However, aging has a potential to initiate cancer through a variety of non-genetic mechanisms: from epigenetic changes to microenvironment. One of the most potent oncogenes that is also implicated in the rate of aging is c-Myc9. Mice overexpressing c-Myc develop cancer10–12, whereas c-Myc deficient mice are protected from lymphoma and are longer lived13. c-Myc also drives cell competition in healthy tissues14–16. Therefore, c-Myc may orchestrate age-related clonal expansion, constituting an early step in the transition to cancer. Aberrant DNA methylation is another suspect to initiate age-related cancer. Mouse knockout models of DNA methylation regulators, Tet1/2/3 and Dnmt3a, develop myeloid leukemias or lymphomas17–19. CHIP in humans is predominantly driven by mutations in methylation regulators TET2, DNMT3A, and ASXL17,8. Human cancers often exhibit global changes in DNA methylation20. At the same time, many human and mouse tissues accumulate recurrent DNA methylation changes with age21,22, raising a possibility that aging promotes cancer through epigenetic changes alongside genetic aberrations.
Aging microenvironment is another major driver of cancer transformation23, e.g. acting through chronic inflammation24,25. Age-associated B cells (ABC) are a prominent suspect to create a pro-inflammatory environment in lymphoid tissues as they overexpress pro-inflammatory cytokines such as IFNγ, IL-6, and IL-426. ABC are atypical memory B cells that accumulate in mammalian aging and disease26–31 and express T-box transcription factor TBX2128. However, it remains unclear whether ABC contribute to age-related clonal selection and cancer.
Available information on mechanisms of age-related clonal expansion and cancer is currently largely limited to genetics due to lack of appropriate mouse models of spontaneous cancer. Commonly used C57BL/6 and BALB/c mice spontaneously develop and frequently die from B-cell lymphoma32,33 offering a tool to investigate the full complexity of the origin of age-related cancer. Here, we extensively characterized spontaneous mouse age-related lymphoma and identified potential therapeutics against it. We further demonstrate that these systems can serve as translatable models to study mechanisms by which aging favors cancer in humans.
Results
B cells increase in size and undergo clonal expansions during mouse aging
We first examined age-related changes in the spleen and blood of young (6 months old) and old (27 months old) mice. Aged mice developed splenomegaly (Fig. 1a, Fig. S1a) due to a prominent increase in cell numbers (Fig. S1b). Additionally, we found that B cells (Fig. 1b, Fig. S1f) and T cells (Fig. S1d), but not myeloid cells (Fig. S1e), profoundly increase in size with age.
To examine the etiology of enlarged B cells, we analyzed eight common populations of B cells (Fig. S1g,j,k). The proportions of follicular (FO) and marginal zone (MZ) B cells significantly decreased with age, whereas CD21-CD23- B cells increased to 80% of B cells (Fig. 1f,h, Fig. S1i), consistent with previous findings26. Of all populations, splenic and blood CD21-CD23- B cells showed the most robust increase in size (Fig. 1c), whereas the size and proportions of plasma cells and germinal center (Fig. S1h,i) B cells remained unchanged with age. Thus, accumulation of cells and increased CD21-CD23- B-cell size largely account for the observed phenotype of enlarged B cells during aging.
Analysis of progenitor B cells (Fig. S1l) confirmed an age-related increase in the ratio of Pro- and IgM+ immature B cells to B220+ cells and a drop in Pre-B cells (Fig. S1m)34. However, there was no increase in cell size (Fig. S1n) suggesting that B cells enlarge in the secondary lymphoid organs, e.g. the spleen. In addition to C57BL/6 mice, we examined BALB/c mice and found a similar spleen enlargement, an increase in average B-cell size and CD21-CD23- B-cell size, and an increase in splenic cell number with age (Fig. S2a-f). Thus, increased B-cell size, in particular of CD21-CD23- B cells, and splenomegaly are common features of mouse aging.
We next tested if enlarged B cells are premalignant. We sorted regular-sized B cells from spleens of young animals and regular-sized and enlarged B cells from old animals and sequenced variable CDR3 regions of Ig heavy chains by ImmunoSeq. We defined enlarged B cells as cells exceeding the size of young B cells based on forward scatter (FSC) in flow cytometry (Fig. 1d). Strikingly, we found that B cells develop large single clones with age representing on average 38% (Fig. S3a), and that clone size positively correlates with B-cell size (R=0.74, p=0.0015, Fig. 1e). Accordingly, CDR3 diversity profoundly decreases with age and with B-cell size increase (Fig. S3b). Clonal B cells collected from the same animal belonged to the same CDR3 clones independent of their cell size (Fig. 1f), suggesting that enlarged B cells originate from regular-sized B cells and clonally expand after VDJ rearrangement. Top clones did not carry somatic hypermutations within CDR3 regions (Fig. S3c). Together, these findings demonstrate that mouse B cells clonally expand and increase in size outside of germinal centers in the spleen during aging.
To investigate the molecular driver of B-cell clonal expansion we sequenced exomes, RNA and methylomes of the same set of samples. We first found that B cells isolated from old mice carry hundreds of somatic mutations across exomes (Fig. S3g) revealing that age-related B-cell clonal expansion relies on somatic mutations (Supplementary table 1). Only 5-10% of mutations were shared between regular and enlarged B cells, suggesting an independent path for clonal selection for enlarged and regular-sized B cells. Enlarged B cells formed larger clones as measured by variant allele fraction (VAF) (Fig. 1g). Driver somatic mutations were located in the genes commonly mutated in human lymphomas, e.g. Trp53, Pim1 and Myh11 (Fig. 1h), indicating conserved genetic mechanisms of clonal selection between human and mouse B cells. The top clones determined by CDR3 and exome sequencing were almost identical in size (Fig. S3d), consistent with the idea that the clones carrying driver mutations expand after VDJ rearrangements. B-cell size again significantly correlated with top genomic clone size (Fig. 1i), suggesting that B-cell size is a convenient marker of clonal B cells, especially when surface markers are unknown. Mutational signature deconvolution (Fig. S3f) revealed accumulation of SBS29 (73.8% of mutations), SBS42, SBS46, SBS15 and SBS5 (FIg. S3h, i) signatures in clonal B cells, as well as ID5 (Fig. S3j), DBS7, DBS11, and DBS5 (Fig. S3k), suggesting aging, impaired DNA repair, APOBEC and nucleotide modifications as sources of mutations in aging B cells. Despite large clonal expansion found in 12 mice, necropsy analysis revealed that only 3 mice had lymphoma (Fig. S4a). Thus, our analysis covered the premalignant state of B cells in addition to cancer itself.
We further assessed whether mouse premalignant B cells molecularly resemble human lymphomas. We examined gene expression patterns, promoter DNA methylation and somatic mutations in mouse clonal B cells against publicly available datasets of human blood cancer (Supplementary table 2) and aging. Clonal B cells from aged mice acquire highly similar epigenetic and genetic changes to those of human BCLs, and most strongly resemble diffuse large B-cell (DLBCL) and follicular lymphoma (FL) (Fig. 1j-l). Furthermore, analyses of spleen and blood sequences from The Genotype-Tissue Expression (GTEx)35 project (Fig. S3i) revealed a decrease in Ig diversity with age (Fig. S3m,n) and with CHIP (Fig. S3o), as well as correlation between Ig and genomic clone sizes (Fig. S3p), similar to that in mice. Overall, the data suggest conservation of molecular mechanisms of B-cell malignancy during aging between mice and humans; thus, aged mice emerge as a first translational model to study age-related origin of cancer.
Age-associated clonal B cells (ACBC) represent a novel B-cell population with a characteristic premalignant epigenetic program
PCA analyses of gene expression, promoter methylation and protein levels (Fig. 2a) revealed that most of the variation between B-cell samples is explained by clonal expansion and enlargement of B cells. The biological age of clonal B cells36, quantified by epigenetic aging clocks, significantly correlated with both clone size and B-cell size (Fig. 2b, Fig. S5b-c). Thus, clonal age-emergent B cells are substantially divergent at the molecular level and are epigenetically older than healthy age-matched B cells.
Clonal B cells further upregulated the expression of targets of c-Myc, central metabolic pathways, and G2/M checkpoint genes (Fig. S5f), as well as ribosome biogenesis and processing pathways (Fig. S5g-h), and rRNA was elevated in enlarged B cells (Fig. S5j). Upregulation of c-Myc targets and G2/M checkpoint was also validated at the protein level (Fig. S5f, Supplementary table 3). Targets of c-Myc were also consistently up-regulated in enlarged B cells (Fig. 2c), which were further validated in qRT-PCR (Fig. S5i). Altogether, our results strongly implicate c-Myc in driving clonal expansion and cell size increase of B cells with age.
While many genes became up-regulated by c-Myc in clonal B cells, many other genes were suppressed, and this pattern was consistent with the patterns of DNA methylation (Fig. S5d). In line with what is known about human cancer20,37,38, we observed a genome-wide increase in promoter methylation in clonal mouse B cells from old mice (Fig. 2d). Some of the silenced genes were tumor suppressors, like Id3 and Mitf (Fig. 2e). On the other hand, oncogene Ptpn1 was up-regulated consistent with the loss of promoter methylation (Fig. S5e), as well as oxidative phosphorylation pathway genes (Fig. S5f). Our results strongly suggest that DNA methylation contributes to clonal selection and cancer transformation in aged tissues.
To understand the origin of clonal B cells we analyzed publicly available single-cell RNA-seq datasets that previously unraveled two major clusters of B cells unique to aged spleens39,40. Both clusters were negative for CD21 (Cr2) and CD23 (Fcer2a) (Fig. S6a-b), and thus represented CD21-CD23- B cells (Fig. 1c). We identified one cluster as ABC, based on the expression of the Tbx21 and CD11c (Itgax)28 (Fig. 2f, Fig. S6c). To test which B-cell population was clonal, we estimated the clonality signature score for each cell based on bulk RNA-seq of clonal B cells (Supplementary Table 3). The other cluster of age-related B cells had the highest scores among aged B cells (Fig. 2f,g). Reconstruction of CDR3 regions of Ig kappa chains in individual cells confirmed that clonal B cells belonged to one of the age-emergent B-cell clusters (Fig. 2h). We named these cells as age-associated clonal B cells (ACBC). We further confirmed significant up-regulation of c-Myc targets in clonal B cells at the single-cell level (Fig. 2j). Notably, the clonality signature score significantly correlated with Igk clone size in both datasets (Fig. S6d).
We observed that both age-related clusters of B cells shared many deregulated genes compared to the rest of splenic B cells (FO/MZ) (Fig. 2i). Moreover, ACBC and ABC shared CDR3 Igk sequences (Supplementary Table 4). Because ABC always co-occur with the ACBC population, but not the other way around, ACBC appear to originate from ABC, providing a mechanistic link between aging and cancer.
To differentiate ABC and ACBC populations with flow cytometry, we established CD29 as a positive marker for both populations and CD24a as a specific positive marker for ACBC (Fig. 4g). qRT-PCR confirmed that our panel differentiate ABC from ACBC, as FACS-sorted ABC cells (CD19+CD21-CD23-CD29+CD24-) overexpress Tbx21, and ACBC overexpress c-Myc (Fig. S6h), which was further validated with immunofluorescence microscopy (Fig. 2m). Staining for new markers (Fig. S6e) confirmed that ACBC (CD19+CD21-CD23-CD29+CD24+) accumulated with age in the spleen (Fig. S6f), but also in blood (data not provided), and that they are larger than follicular B cells (Fig.S6g), which were further validated with immunofluorescence microscopy (Fig. 2k,l). Circulation of ABC and ACBC can explain their infiltration in peripheral tissues (Fig. S4b-e).
We conclude that clonal B cells in old mice are predominantly represented by a novel ACBC population, and that c-Myc is a potential driver of B-cell clonal expansion and cell size increase with age. We further established a panel of markers to differentiate clonal ACBC from ABC using flow cytometry. Finally, we found a significant enrichment for the ACBC signature in GTEx subjects47 with clonal B cells and enrichment for both mouse and human ABC signatures with aging (Fig. 2n). Thus, human subjects accumulate clonal B cells that are transcriptionally similar to the newly discovered mouse ACBC.
Clonal B cells are predictors of lifespan, originate with the support of an aging microenvironment and become self-sufficient over time
Longitudinal blood analysis (Fig. 3a) established that enlarged B cells accumulate in the mouse blood starting approximately at the age of 28 months and that myeloid bias started at about 25 months of age (Fig. S7a,b). The trajectory of blood age-related changes became more consistent when the mice were aligned by the death rather than age (Fig. 3b,c). This suggests a prominent role of the age-related blood composition changes in lifespan control. Indeed, mice with the higher CD21-CD23- and FSC scores had a 69-day shorter maximum lifespan (Fig. 3d).
To capture how cell composition changes during aging related to each other, we developed custom scores (see Methods) for B-cell enlargement (FSC Score), myeloid bias (Myeloid score) and CD21-CD23- B-cell bias (CD21-CD23- Score). The CD21-CD23- score correlated with the myeloid score at early ages, and correlated with the FSC score at later ages (Fig. S7d). This suggests that myeloid cells antedate the formation of CD21-CD23- B cells, and later CD21-CD23- B cells predate the enlargement of B cells.
We next established causality of age-related blood changes in vitro (Fig. 3e). In agreement with longitudinal data, follicular B cells convert into CD21-CD23- B cells in the presence of the cells from the aging spleen microenvironment (Fig. S7f), and ABC triggered c-Myc expression and cell size increase in FO B cells in vitro (Fig. 3f, S7e). Concentration of seventeen cytokines were increased in the media of ABC cells (Fig. 3g), five of which were confirmed by single-cell RNA-seq (Fig. 3h, Fig. S7g), and we named the combination of these cytokines a cancer-associated secretory phenotype (CASP) by analogy to SASP. In addition, our phosphoproteome profiling of clonal and enlarged B cells revealed activation of MAPK1 and p38a kinases (Supplementary Table 5), confirming that the extracellular stimuli from aging microenvironment are involved in age-related clonal selection and B-cell enlargement.
To test whether cancerous ACBC can survive outside of the aging environment we injected spleen extracts from mice diagnosed with lymphoma (and containing ACBC) into Rag2-/- mice (Fig. 3i). Most recipients of extracts containing cancerous ACBC experienced B-cell expansion with time, which were mostly represented by ACBC (Fig. 3j). Six mice perished as a result of the B-cell expansion (Fig. 3k).
Altogether, our data suggest that old follicular B cells turn into ABC cells under the influence of the aging microenvironment of the spleen. ABC secrete a spectrum of cytokines and activate c-Myc expression in surrounding B cells thus selecting out somatic mutations that support c-Myc overexpression. Indeed, the same genes with driver mutations in mouse lymphoma are commonly mutated in human lymphomas with MYC activation (Supplementary Table 6). That process gives rise to clonal Myc-overexpressing B cells, or ACBC (Fig. 3i), in line with the Adaptive Oncogenesis Hypothesis41,42. Over time ACBC evolve into self-sufficient malignant cells that thrive outside of the aging microenvironment.
c-Myc deficiency and rapamycin attenuate aging and clonal expansion of B cells
To test a causal role of c-Myc in the age-related clonal expansion, we analyzed c-Myc heterozygous mice (Myc+/-, Fig. 4a)13 and mice overexpressing c-Myc specifically in B cells (Eμ-Myc)11. We found that c-Myc haploinsufficiency (Fig. S8a) attenuated many age-related phenotypes of B cells: splenomegaly (Fig. S8b), increase in total B-cell size and the ACBC size (Fig. 4b,c) but not other B-cell populations (Fig, S8e,f), the accumulation of ACBC (Fig. 4d) and RNA accumulation (Fig. S8c), compared to age-matched wild type (Wt) siblings. Moreover, Myc+/- B cells preserved a more diverse Ig repertoire with age (Fig. 4e). Other immune cell types were not affected by c-Myc deficiency (Fig. S8d,j) suggesting a cell-type specific role of c-Myc for B-cell aging and malignancy. At the same time, B cells from Eμ-Myc mice had higher levels of RNA (Fig. S8k), were enlarged in the spleen (Fig. S8l) and bone marrow (Fig. S8m) and exhibited a reduced diversity of the Ig repertoire than non-carriers (Fig. S8n). Together, these data strongly suggest that c-Myc regulates cell size and fosters clonal expansion of B cells at young and old ages. We further subjected mice to rapamycin as it was recently reported to inhibit c-Myc activity43 and found that rapamycin attenuated cell size increase of ABC and ACBC (Fig. 4g) and accumulation of ABC (Fig. 4h) in 28-month-old mice.
RNA-seq demonstrated that clonal B cells from aged mice and Eμ-Myc B cells changed the same pathways, whereas c-Myc-deficient B cells were opposite to both (Fig. 4i). Genes that belong to c-Myc targets, metabolic and inflammation pathways cooperatively respond to changes in c-Myc level, and thus the majority of changes in the clonal B-cell transcriptome can be attributed to c-Myc overexpression. These data highlight the therapeutic potential of rapamycin and c-Myc deficiency against malignant transformation of aged B cells.
Discussion
To understand how cancer develops in the context of aging we studied commonly used mouse strains, which spontaneously develop B-cell lymphoma with age32,33. We found that mouse clonal B cells acquire genomic, gene expression and DNA methylation patterns of human lymphomas, in particular follicular and diffuse-large B-cell lymphomas. This unexpected result points to evolutionary conserved mechanisms of age-related B-cell malignancy at the genetic and epigenetic levels. Further studies may investigate if other spontaneous mouse cancers evolve similarly to human cancers with age.
We further discovered a population of age-associated clonal B cells - ACBC - that over time evolve into cancerous B cells. c-Myc emerged as a major driver of the B-cell clonal expansion with age as confirmed by transcriptome studies and mouse models. Our DNA methylation analyses revealed hypermethylated promoters in clonal B cells that silenced gene expression similarly to those in human cancer20,37,38. One of the silenced genes was Id3, which is a human and mouse tumor suppressor44–47, providing an example of how age-related changes in DNA methylation can contribute to B-cell malignancy. At the same time, ACBC lack somatic hypermutations in CDR3 regions and thus do not undergo Ig selection in germinal centers. Together, these results suggest that the age-related DNA methylome, somatic mutations and activated c-Myc provide a selective advantage to B-cell clones rather than the canonical Ig-selection.
ABC shared the majority of deregulated genes, secretory phenotype and Ig clones with ACBC, suggesting that age-associated B cell is the cell of origin for clonal B cells in mice. ABC accumulate in elderly patients that are at higher risk of B-cell lymphomas, in HIV-positive patients susceptible to Burkitt lymphoma48,49, and in patients with autoimmune diseases29,31 susceptible to diffuse-large B-cell lymphoma50. Thus, it is likely that ABC give rise to ACBC and, consequently, B-cell lymphoma in humans too.
Although activation of c-Myc is a consequence of genetic perturbation in many cancers51, there were no genetic alterations that can explain up-regulation of c-Myc in B cells with aging. Instead, exposure of follicular B cells to ABC was sufficient to elevate c-Myc expression, and we revealed a spectrum of cytokines associated with that. Future studies may test which component of the secretome of aging microenvironment triggers c-Myc expression in B cells thereby contributing to their malignancy. Despite ACBC arising from the aging microenvironment, some ACBC are transplantable into young immunodeficient mice, demonstrating that surrounding signals are essential to initiate rather than to support cancer.
Finally, we demonstrated that two interventions, genetic and pharmacological, that potently extend mouse lifespan (c-Myc deficiency and rapamycin)13,52 attenuate malignant phenotypes of B cells. Our longitudinal study further revealed that accumulation of clonal B cells stratifies aged mice into long- and short-lived groups. It is possible that c-Myc deficiency and rapamycin (and possibly other longevity interventions) extend mouse lifespan, at least in part, by delaying age-related lymphoma. Based on our data of conserved mechanisms of B-cell malignancy between human and mouse, mouse longevity interventions emerge as therapeutic strategies to delay B-cell malignancy in elderly patients.
Our work demonstrates the relevance of spontaneous age-related mouse cancers to human cancer and shows that aged mice is a translational model to study how aging favors cancer.
Experimental Procedures
Animals
Balb/c and C57BL6/JNia mice were obtained from the National Institute on Aging Aged Rodent Colony. Only female mice were used in the present study. 6-wk-old female Eμ-Myc mice and control non-carriers were purchased from The Jackson Laboratory (cat. #002728). Animals were euthanized with CO2. Spleens were harvested and stored in cold PBS until analysis, and liver samples were immediately frozen in liquid nitrogen and stored at -80°C. For bone marrow analyses, bones were stripped of the muscle in cold PBS, cut from both ends, and cells were aspirated with 5 ml of cold PBS. The aspirates were filtered through 40 µm Falcon Cell Strainers (Corning). Spleens from 24-27 months old Myc haploinsufficient (Myc +/-) mice and their wild type siblings were collected at Brown University and transferred on cold PBS to Harvard Medical School for further analysis. Mice were subjected to encapsulated rapamycin (Rapamycin Holdings. 126 ppm of active compound) or to encapsulating material of the same concentration (Rapamycin Holdings) in 5053 diet (TestDiet) and were fed ad libitum. For transplanted experiments, total splenocytes were cryopreserved in 10% DMSO FBS. Recovered splenocytes were counted and 2-6 mln of cells in 100 ul of sterile saline solution (Sigma) were i.p. Injected into 2-3 months old Rag2-/- mice. Rag2-/- mice were provided and housed by Manis laboratory at Karp Family Research Building of Boston Children’s Hospital. All experiments using mice were performed in accordance with institutional guidelines for the use of laboratory animals and were approved by the Brigham and Women’s Hospital and Harvard Medical School Institutional Animal Care and Use Committees.
Flow cytometry and sorting
mAbs used for staining included: anti-CD19 [6D5], anti-CD43 [S11], anti-IgM [RMM-1], anti-CD11b [M1/70], anti-B220, anti-CD80 [16-10A1], anti-CD3 [17A2], anti-Fas [SA367H8], anti-CD138 [281-2], anti-CD23 [B3B4], anti-CD21 [7E9], anti-CD24 [M1/69], anti-CD29 [HMβ1-1], anti-CD93, and anti-CD45 [30-F11] (all from Biolegend) with fluorophores as in Supplementary table 7. Dead cells were excluded by DAPI staining. Data was collected on a Cytek DXP11 and analyzed by FlowJo software (BD). Cells were sorted on BD Aria Fusion. Spleens were gently pressed between microscopy slides to get single-cell suspensions. One ml of cell suspensions were 1) incubated with 14 ml of red blood lysis buffer for 10 min on ice, 2) centrifuged at 4°C, 250 g for 10 min, 3) washed once with 1 ml of FACS buffer (1% FBS in PBS), 4) stained in 100 ul of AB solution (2 ng/ul of each AB) at 4°C for 20 min protected from light, 5) washed again, 6) filtered into tubes with cell strainer snap cap (Corning), and 7) analyzed with flow cytometry or FACS-sorted into 1 ml of FACS buffer. Surface markers of ABCs and ACBC were selected based on: 1) differential up-regulation in both populations in scRNA-seq, 2) annotation as surface protein in GO, 3) differentially up-regulated in clonal B cells in bulk RNA-seq data (Supplementary table 3). It yielded a few candidates, of which we chose CD29 (Itgb1) as a positive marker for both age-related populations, and CD24 to distinguish ACBC from ABC.
RNA sequencing and qRT-PCR
Total RNA, DNA and proteins were extracted from fresh or snap-frozen FACS-sorted 2-5 million splenic B cells using AllPrep DNA/RNA/Protein Mini Kit (Qiagen) following the manufacturer’s instructions. RNA was eluted with 42 ul of RNAse-free water. RNA concentration was measured with Qubit using the RNA HS Assay kit. Libraries were prepared with TruSeq Stranded mRNA LT Sample Prep Kit according to TruSeq Stranded mRNA Sample Preparation Guide, Part # 15031047 Rev. E. Libraries had been quantified using the Bioanalyzer (Agilent), and were sequenced with Illumina NovaSeq6000 S4 (2×150bp) (reads trimmed to 2×100bp) to get 20M read depth coverage per sample. The BCL (base calls) binary were converted into FASTQ using the Illumina package bcl2fastq. Fastq files were mapped to mm10 (GRCm38.p6) mouse genome. and gene counts were obtained with STAR v2.7.2b53. Myc targets for Figure 2c were taken from the hallmark set of genes ‘MYC_TARGETS_V1’ from msigdb database. For quantitative RT-PCR, RNA samples were normalized by DNA concentration isolated from the same sample, then 2 ul of normalized RNA were mixed with iTaq Universal SYBR (Bio-Rad) and primers for c-Myc (F: TTCCTTTGGGCGTTGGAAAC, R: GCTGTACGGAGTCGTAGTCG), Actb (F: GGCTGTATTCCCCTCCATCG, R: CCAGTTGGTAACAATGCCATGT) or Gapdh (F: AGGTCGGTGTGAACGGATTTG, R: TGTAGACCATGTAGTTGAGGTCA) for the final volume of 10 ul, and loaded into Multiplate 96-Well PCR Plates (Bio-rad). Data was acquired for 40 cycles on Bio-Rad C1000, CFX96 Thermal Cycler. Each sample was loaded in duplicate.
Whole exome sequencing
Total DNA was extracted with AllPrep DNA/RNA/Protein Mini Kit (Qiagen) following the manufacturer’s instructions. DNA was eluted with 100 ul of EB. DNA concentration was measured with Qubit using DNA BR Assay kit. Libraries were prepared with Agilent SureSelect XT Mouse All Exon according to SureSelectXT Target Enrichment System for Illumina Version B.2, April 2015. Libraries had been quantified using the Bioanalyzer (Agilent), and were sequenced with Illumina NovaSeq6000 S4 (2×150bp) (reads trimmed to 2×100bp) to get 100X throughput depth (roughly 50X on-target) coverage per sample. The BCL (base calls) binary were converted into FASTQ using the Illumina package bcl2fastq. Genomic reads were mapped to the GRCm38.p2 mouse genome assembly using BWA-MEM v0.7.15-r1140 (Li, 2013) and sorted using Samtools v1.654. Somatic mutations were called with Mutect2 from the GATK package v4.1.8.0 using liver as matched control; high quality variants were selected with GATK FilterMutectCalls55. Variants were annotated with the Ensembl Variant Effect Predictor v100.2 (with option --everything)56. Only the variants that 1) were supported by 5 reads or more (DP.ALT>4), and 2) were in positions covered by 30 reads or more (DP.ALT+DP.REF>29), were taken for VAF analysis. Signatures of somatic mutations were extracted and deconvoluted using Sigproextractor v.1.0.20 with 100 nmf replicates.
Proteomics
Tandem mass tag (TMTpro) isobaric reagents were from ThermoFisher Scientific (Waltham, MA). Trypsin was purchased from Pierce Biotechnology (Rockford, IL) and LysC from Wako Chemicals (Richmond, VA). Samples were prepared as described previously57,58. Briefly, cell pellets were syringe-lysed in 8M urea complemented with protease and phosphatase inhibitors. Samples were reduced using 5mM TCEP for 30 min and alkylated with 10 mM iodoacetamide for 30 min. The excess of iodoacetamide was quenched with 10 mM DTT for 15 min. Protein was quantified using the BCA protein assay. Approximately 50 µg of protein were chloroform-methanol precipitated and reconstituted in 100 µL of 200 mM EPPS (pH 8.5). Protein was digested using Lys-C overnight at room temperature followed by trypsin for 6h at 37°C, both at a 100:1 protein:protease ratio. After digestion, the samples were labeled using the TMTpro16 reagents for 90 mins, the reactions were quenched using hydroxylamine (final concentration of 0.3% v/v). The samples were combined equally and subsequently desalted.
We enriched phosphopeptides from the pooled TMT-labeled mixtures using the Pierce High-Select Fe-NTA Phosphopeptide Enrichment kit (“mini-phos”)58,59 following manufacturer’s instructions. The unbound fraction was retained and fractionated using basic pH reversed-phase (BPRP) HPLC. Ninety-six fractions were collected and then consolidated into 12 which were analyzed by LC-MS360.
All data were collected on an Orbitrap Fusion Lumos mass spectrometer coupled to a Proxeon NanoLC-1000 UHPLC. The peptides were separated using a 100 μm capillary column packed with ≈35 cm of Accucore 150 resin (2.6 μm, 150 Å; ThermoFisher Scientific). The mobile phase was 5% acetonitrile, 0.125% formic acid (A) and 95% acetonitrile, 0.125% formic acid (B). For BPRP fractions, the data were collected using a DDA-SPS-MS3 method with online real-time database searching (RTS)61 to reduce ion interference62,63. Each fraction was eluted using a 90 min method over a gradient from 6% to 30% B. Peptides were ionized with a spray voltage of 2,600 kV. The instrument method included Orbitrap MS1 scans (resolution of 120,000; mass range 400−1400 m/z; automatic gain control (AGC) target 2×105, max injection time of 50 ms and ion trap MS2 scans (CID collision energy of 35%; AGC target 1×104; rapid scan mode; max injection time of 120 ms). RTS was enabled and quantitative SPS-MS3 scans (resolution of 50,000; AGC target 2.5×105; max injection time of 250 ms).
Phosphoproteomics
Phosphopeptides were analyzed with FAIMS/hrMS2 using our optimized workflow for multiplexed phosphorylation analysis61,64,65. Briefly, the Thermo FAIMSpro device was operated with default parameters (inner and outer electrodes were set at 100°C, yielding a FWHM between 10 V to 15 V and dispersion voltage (DV) was set at -5000 V). Each “mini-phos” was analyzed twice by the mass spectrometer using a 2.5h method having a gradient of 6% to 30%.
Raw files were first converted to mzXML. Database searching included all mouse entries from UniProt (downloaded March 2020). The database was concatenated with one composed of all protein sequences in the reversed order. Sequences of common contaminant proteins were also included. Searches were performed using a 50ppm precursor ion tolerance and 0.9 Da (low-resolution MS2) or 0.03 Da (high-resolution MS2) product ion tolerance. TMTpro on lysine residues and peptide N termini (+304.2071 Da) and carbamidomethylation of cysteine residues (+57.0215 Da) were set as static modifications, and oxidation of methionine residues (+15.9949 Da) was set as a variable modification. For phosphopeptide analysis, +79.9663 Da was set as a variable modification on serine, threonine, and tyrosine residues.
PSMs (peptide spectrum matches) were adjusted to a 1% false discovery rate (FDR)66,67. PSM filtering was performed using linear discriminant analysis (LDA) as described previously68, while considering the following parameters: XCorr, Δ Cn, missed cleavages, peptide length, charge state, and precursor mass accuracy. Protein-level FDR was subsequently estimated. Phosphorylation site localization was determined using the AScore algorithm69. A threshold of 13 corresponded to 95% confidence that a given phosphorylation site was localized.
For reporter ion quantification, a 0.003 Da window around the theoretical m/z of each reporter ion was scanned, and the most intense m/z was used. Peptides were filtered to include only those with a summed signal-to-noise ratio ≥100 across all channels. For each protein, the filtered signal-to-noise values were summed to generate protein quantification values. To control for different total protein loading within an experiment, the summed protein quantities of each channel were adjusted to be equal in the experiment. For each protein in a TMTpro experiment, the signal-to-noise was scaled to sum to 100 to facilitate comparisons across experiments.
Spectral counts values were analyzed with R in Rstudio. Proteome and phosphoproteome data were normalized using the RLE method and log transformed using the edgeR package71. Values for phospho sites were normalized to corresponding protein level and differentially changed sites were calculated with the limma package. Ranked phospho sites were then assessed for enrichment for targets of mouse kinases using PTM-SEA resources70 and kinact software72.
Single-cell RNA sequencing analysis
To identify genes differentially expressed in this newly identified B cell cluster and other B cells, we downloaded the single cell RNA seq data from Calico’s murine aging cell atlas (https://mca.research.calicolabs.com/data, spleen single-cell count data, filtered)39. Preprocessing of the downloaded data was performed using scanpy73. We first removed cells with a high (>0.05) percentage of mitochondrial reads. Cells not annotated as B cells were also removed. We then normalized the read counts by total reads number per cell and multiplied by a rescaling factor of 10000. Normalized reads were log transformed after adding a pseudo-count of 1. We scaled the log-transformed data to unit variance and zero mean and clipped maximum value to 10. After the above data preprocessing, we selected the cells corresponding to the age-related B-cell cluster, which was C130026I21Rik+Apoe+Cr2-Fcer2a-. For each cell, we used a linear combination of the RNA level of these four marker genes to calculate a score (score = C130026I21Rik + Apoe - Cr2 - Fcer2a), and within this cluster we selected the ABC cluster that is Tbx21+ and another one that is Myc+ using the same linear system. We then used a score threshold of 2.5 to select cells in the cluster of interest. Differential expression analysis was performed between the cell cluster of interest and all other remaining cells using the rank_genes_groups function in scanpy (Wilcoxon test). Details of the analysis can be found in our jupyter notebook B_cell_scRNAseq.ipynb. To reconstruct CDR3 regions of single cells, we demultiplexed bulk fastq files into single cell fastq files with scruff package in R74, mapped each fastq to GRCm38.p6 genome and obtained gene counts with STAR v2.7.2b53, and reconstructed CDR3 regions of Ig kappa chain from individual cells using mixcr75. Clone sizes were calculated as percent of templates supporting the current Ig kappa chain to the total number of reconstructed templates of Ig kappa chain for the sample. Cells with fewer than 200 gene counts were removed. Raw gene counts were log-transformed and normalized using the edgeR package in R71. Clonality signature score was calculated for each cell as transcript level of top 50 genes minus transcript level of bottom 50 genes ranked by p-value and taken from our bulk RNA-seq regressed against clone sizes (Supplementary Table 3). tSNEs were calculated using the M3C package in R76.
Reconstruction of Ig CDR3 regions
Genomic CDR3 regions of Ig heavy chains were analyzed with Immunoseq (Adaptive Biotechnologies). CDR3 regions were reconstructed from RNA sequencing raw data using mixcr software75 with recommended settings for transcriptome data. Filtering of reconstructed regions and diversity analysis was done with VDJtools software77.
Immunofluorescence microscopy
B cells were FACS-sorted at 500 thousand cells per well and incubated with poly l-lysine treated coverslips for 1 hour in 24 well plates. Cells were permeabilized with 0.1% Triton X-100 2 times for 30 seconds, fixed in 3.7% PFA in PBS for 10 minutes and washed three times with PBS, incubated with the blocking buffer until further analysis (1% BSA, 0.1% Triton X-100 in PBS). Samples were incubated with primary antibodies overnight at 4 °C (1:100, Abcam #ab32072), then washed with PBS five times and incubated overnight with Alexa Fluor 568 - conjugated anti-rabbit secondary antibodies (1:500, Biotium cat#20098) and DAPI dye. Cells were washed with PBS and mounted with ProLong Diamond antifade (Thermo Fisher Scientific). Samples were imaged using Leica SP8 confocal microscope. Images were analyzed with ImageJ78. Raw Z-stacks were converted to the maximum intensity projection images. Nuclei and cell borders were detected using manual thresholding and “Analyze Particle” function. All crowded groups and not-round shaped cells were manually removed from the analysis.
Reduced representation bisulfite sequencing (RRBS)
Libraries were prepared and sequenced as in 22. Bisulfite sequence reads were trimmed by TrimGalore v0.4.1 and mapped to the mouse genome sequence (mm10/GRCm38.p6) with Bismark v0.15.0 79. We kept CpG sites that were covered by five reads or more. Promoter regions were determined as the [-1500, +500] bp from the transcription start site (following the direction of the transcription) taken from Ensembl annotation file (Mus_musculus.GRCm38.100.chr.gtf). The start and end positions of gene bodies were taken from the Ensembl gene predictions (Mus_musculus.GRCm38.cds.all.fa). The mean methylation levels were calculated for regions that have at least 5 covered CpG sites with average methylation level above 1%. To determine ribosomal DNA methylation (rDNAm) age, we developed a blood rDNAm clock in a similar way as described in 80. Briefly, we applied ElasticNet regression on ribosomal DNA (BK000964.3) CpG methylation levels of 153 control fed C57BL/6 blood samples with an age range from 0.67 to 35 months (GSE80672).
Longitudinal blood collection and analysis
Mice were anesthetized with isoflurane and then locally with topical anesthetic, restrained, and approximately 100 ul of blood was collected from mouse tails into EDTA-coated tubes (BD). Blood was incubated on ice until further analysis (2-3 hours), then mixed with 1 ml of red blood cell lysis buffer and centrifuged at 250g for 10 minutes at 4°C. Pellets were washed once with a FACS buffer (PBS with 1% FBS), split equally into 2 tubes and incubated for 20 minutes at 4°C with antibodies against B cell, T cell and myeloid cell markers, or follicular, marginal zone, and plasma cell markers. Stained cells were washed again, resuspended in 200 ul of FACS buffer and analyzed with FACS, with 20,000 events being recorded. Dead cells were gated by DAPI staining. Cell size was measured with forward scatter.
Longitudinal blood scores
We defined the FSC score as the difference between the last measurement of mean B cell size prior to death (if mouse died) or B cell size at the given round (if mouse was alive) and the mean B cell size of young mice at the same round. CD21.score was calculated as delta(CD21-CD23-) - delta(Follicular). Delta(CD21-CD23-) was calculated as percentage of CD21-CD23- B cells of total B cells before death (if mouse died) or percentage of CD21-CD23- B cells at the given round (if mouse was alive) minus percentage of CD21-CD23- B cells in young mice at the same round. delta(Follicular) was calculated the same way for the percentage of follicular B cells of B cells. The myeloid score was calculated as delta(Myeloid) - delta(B cells). delta(Myeloid) was calculated as percentage of myeloid cells of CD45+ cells before death (if mouse died) or percentage at the given round (if mouse was alive) minus the percentage of myeloid cells in young mice at the same round. delta(B cells) was calculated the same way for the percentage of B cells of CD45+ cells. A higher CD21.score indicates a higher proportion of CD21-CD23- B cells to total B cells and/or lower proportion of follicular B cells to total B cells. A higher myeloid score indicates a higher proportion of myeloid cells to CD45+ cells and/or lower proportion of B cells to CD45+. A higher FSC score indicates a greater increase of B-cell size.
Cell culture
Freshly FACS-sorted cells were plated into 96 wells at 400,000 cells per well for each cell type and cultured for 24-48 hours in 200 ul of RPMI medium with 10% FCS (Gibco), 2 mM glutamine (ThermoFisher), 1% oxaloacetic acid (15 mg/ml), 5 mg/ml sodium pyruvate (ThermoFisher), 1% non-essential amino acids (ThermoFisher), and 50 μM 2-ME (Sigma). Where follicular B cells were incubated alone, 800,000 cells were plated. After co-incubation, cells were centrifuged at 250 g for 10 minutes at 4°C with slow deceleration. The media were carefully removed leaving ∼50 ul, cells were washed once in 100 ul of FACS buffer, then resuspended in 100 ul of AB solution (1:100) and incubated at 4°C for 20 minutes protected from light, washed again and resuspended in 200 ul of FACS buffer, and filtered through 40 µm Falcon Cell Strainers (Corning). Follicular B cells (20-100 thousand) were gated as CD19+CD21intCD23+ and FACS-sorted into 300 ul of Trizol. RNA was extracted with Direct-zol RNA Microprep (Zymo Research).
Cytokine profiling
Thawed or fresh supernatants from cell culture experiments (see bove) were analyzed with Proteome Profiler Mouse Cytokine Array Kit, Panel A (R&D Systems) following manufacturer’s protocol, 400-800 ul of supernatants pooled from three wells were analyzed. The exposure times were 1, 10 or 60 minutes depending on the intensity of the signal. Intensity of the signals was quantified using ImageJ. Individual intensities were normalized to mean intensity of all measures within the experiment to allow combining results from three independent experiments.
Necropsy analysis
Mice were euthanized with CO2 followed by cervical dislocation. The chest and abdomen were opened, and the body was immersed into formalin solution and stored at 4°C until further analysis. For necropsy analysis all organs, including small endocrine organs, were dissected, trimmed at 5 mm thickness and embedded in paraffin blocks. Paraffin blocks were sectioned at 5 μm and stained with hematoxylin and eosin. The slides were examined blindly by a pathologist. Lymphoma was diagnosed when multiple solid tissues contained large uniform sheets of atypical lymphocytes with large nuclei. Lymphocytic hyperplasia was diagnosed when any solid tissue had small infiltrates of atypical lymphocytes with large nuclei.
Data analysis and availability
All data were analyzed and plotted with R in Rstudio. RNA sequencing, DNA methylation and proteomics data were preprocessed and analyzed for differential changes and GSEA with limma81, edgeR71 and clusterprofiler82 packages. All p-values for group means comparisons were calculated with two-tailed Student t-test, unless otherwise specified. Correlations between two variables were evaluated using Pearson’s correlation coefficient. PCA analysis was done with the factoextra package. Color schemes are from the ggsci package. Raw reads for RNA-sequencing, whole exome sequencing, and RRBS are available at SRA (PRJNA694093).
Author Contributions
AVS and JPC conceived the project, designed and performed experiments, analyzed the data and drafted the manuscript. AB, OSS, JAP, CK, APP, MM, MM, and YH performed experiments, analyzed the data and revised the manuscript. GL, SPG, and JMS provided research materials, assisted with experimental design and revised the manuscript. JPM and VNG interpreted the data, designed experiments, provided research materials, and revised the manuscript. VNG supervised the overall project.
Declaration of Interests
The authors declare no competing interests.
Supplementary Figure Legends
Acknowledgments
The authors thank members of the Gladyshev laboratory for discussion. Supported by Max Kade Foundation to JPC and by NIH grants to VNG.
Literature cited
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.
- 16.↵
- 17.↵
- 18.
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.
- 28.↵
- 29.↵
- 30.
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.
- 46.
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵