Cross‐species comparison illuminates the importance of iron homeostasis for splenic anti‐immunosenescence

Abstract Although immunosenescence may result in increased morbidity and mortality, many mammals have evolved effective immune coping strategies to extend their lifespans. Thus, the immune systems of long‐lived mammals present unique models to study healthy longevity. To identify the molecular clues of anti‐immunosenescence, we first built high‐quality reference genome for a long‐lived myotis bat, and then compared three long‐lived mammals (i.e., bat, naked mole rat, and human) versus the short‐lived mammal, mouse, in splenic immune cells at single‐cell resolution. A close relationship between B:T cell ratio and immunosenescence was detected, as B:T cell ratio was much higher in mouse than long‐lived mammals and significantly increased during aging. Importantly, we identified several iron‐related genes that could resist immunosenescence changes, especially the iron chaperon, PCBP1, which was upregulated in long‐lived mammals but dramatically downregulated during aging in all splenic immune cell types. Supportively, immune cells of mouse spleens contained more free iron than those of bat spleens, suggesting higher level of ROS‐induced damage in mouse. PCBP1 downregulation during aging was also detected in hepatic but not pulmonary immune cells, which is consistent with the crucial roles of spleen and liver in organismal iron recycling. Furthermore, PCBP1 perturbation in immune cell lines would result in cellular iron dyshomeostasis and senescence. Finally, we identified two transcription factors that could regulate PCBP1 during aging. Together, our findings highlight the importance of iron homeostasis in splenic anti‐immunosenescence, and provide unique insight for improving human healthspan.


| INTRODUC TI ON
The immune system protects the organism by fighting infection, eliminating cancerous cells, and regulating tissue repair, etc., which is critical for the organismal homeostasis and longevity (Borgoni et al., 2021).The gradual deterioration of the immune system during aging, known as immunosenescence, may result in increased morbidity and mortality in animals (Barbé-Tuana et al., 2020).Although the importance of immunosenescence has been recognized for a long time, its underlying mechanisms remain elusive.
Lifespan varies over 100-fold across mammals, and usually positively correlates with their body mass (Kowalczyk et al., 2020).For example, the maximum lifespans of mouse and whale are reported to be ~4 and ~211 years, respectively (Keane et al., 2015).During evolution, many mammals have evolved specific adaptations of immune systems, which significantly contribute to their extreme longevity.After size correction, myotis bats, naked mole rats (NMRs), and human possess much more exceptional longevity than mice (Tollis et al., 2020).Thus, the immune systems of these long-lived mammals present unique models to study the mechanisms of antiimmunosenescence and healthy longevity.
As the largest secondary lymphoid organ of the body, the spleen plays a critical role in hematopoiesis and blood filtration (Lewis et al., 2019).Aged, infected, or dysfunctional red blood cells are removed from the circulation by splenic red pulp macrophages, and their iron is reclaimed for systemic use (Jebb et al., 2020).Besides, the spleen contains nearly one-fourth of all lymphocytes necessary for innate and adaptive immune responses (Bronte & Pittet, 2013).In this study, we make cross-species comparison of splenic immune cells between three long-lived mammals (i.e., myotis bat (Myotis ricketti), NMR, and human) and the short-lived mammal, mouse, at single-cell resolution, and we aim to (i) uncover the similarities and differences in the cellular diversity and gene expression pattern of immune cells across these species; (ii) identify marker genes that are more highly expressed in long-lived species, which may have anti-immunosenescence potentials; (iii) and finally explore their roles in anti-immunosenescence.In summary, the systematical comparison of splenic immune cells between longlived versus short-lived mammals will provide novel insights in understanding the molecular mechanisms of anti-immunosenescence and healthy longevity.

| Animal models
Myotis bats were sampled in accordance with the permits and ethical guidelines (License No. GIZ20220302) issued by Institute of Zoology, Guangdong Academy of Science.Two male and two female myotis bats were used for single-cell RNA sequencing (scRNA-seq), and four male myotis bats were used for LIP assay.Young (8-10 weeks old, male) and aged (20 months old, male) wild-type C57BL/6J mice were purchased from Casgene Biotech CO., Ltd.All mice were housed under standard environment and provided with clean water and food ad libitum.The Administrative Panel on Laboratory Animal Care at the Guangzhou Laboratory approved all the experimental procedures in this study.
Maximum longevity (years) and adult body mass (g) of mammals were downloaded from the AnAge database (Tacutu et al., 2012).
The relationships of body mass and longevity is plotted by fitting a linear regression in R v4.2.1 (https://www.R-proje ct.org/).

| The isolation of splenic immune cells
Animals were anesthetized by intraperitoneal injection of isoflurane, then the spleens were harvested and immediately transferred into 5 mL iced RPMI 1640 with 10% FBS, and then cut into small pieces (~0.2 cm 2 ) with a scissor.Splenic immune cells were released by grinding the spleen fragments through a 100μm cell strainer with a syringe plunger, or pipetting gently with a Pasteur pipette.Cells were washed with RPMI 1640 with 10% FBS and centrifuged at 450 g for 5 min at 4°C, then resuspended with 3 mL ACK Lysing Buffer (Gibco, A1049201) for 5 min at room temperature to remove red blood cells.
Cells were then washed twice and resuspended in PBS with 0.1% BSA, passed through a 40μm filter, counted and checked viability using trypan blue before use.The cells and the wash buffer should be kept on ice during the whole procedure to preserve cell viability.

| scRNA-seq and bulk RNA-seq
Single-cell libraries of bat splenic immune cells were generated with 10X genomic platform using the Chromium Next GEM single Cell would result in cellular iron dyshomeostasis and senescence.Finally, we identified two transcription factors that could regulate PCBP1 during aging.Together, our findings highlight the importance of iron homeostasis in splenic anti-immunosenescence, and provide unique insight for improving human healthspan.3' Reagent Kits v3.1 following the manufacturer's instruction.For bulk RNA-seq of different cell lines, total RNA was extracted using RNeasy Mini kit (QIAGEN, 74106), libraries were prepared using VAHTS RNA-seq V3 Library Prep Kit for Illumina (Vazyme,NR611) following the manufacturer's instruction.Quality of both scRNA-seq and bulk RNA-seq libraries was evaluated by the Fragment Analyzer (Advanced Analytical).Libraries were sequenced on an illumine No-vaSeq6000 in the PE150 mode.
We performed Illumina short-read sequencing (Illumina RNAseq) and Oxford Nanopore technologies long-read sequencing (ONT RNA-seq) on 13 tissues of adult male bats for genome annotation.
These tissues included heart, liver, spleen, lung, kidney, intestine, cochlea, stomach, pancreas, muscle, brain, eye, and testis.Meanwhile, Illumina RNA-seq data from NCBI, and well-annotated protein sequence from human, mouse, and other bats were also collected for genome annotation.Briefly, genome annotation mainly included three steps: ab initio gene prediction, protein homology-based and transcriptome-based prediction.
Transcriptome-based gene structures were modeled using PAS-Apipeline v2.5.2 (PASA-evidence) based on above results.ORFs were extracted from gene structure based on PASA-evidence and were used for training data for ab initio gene prediction.
For protein homology-based prediction, well-annotated proteins from 10 UniProt v20210405 and bat proteins from NCBI, and aligned the proteins to genome assembly with GenomeThreader 1.7.1 (gthevidence) were collected.Then we used EvidenceModeler v1.1.1 to integrate the above collected evidences into consensus gene models.The weights we used on the integration step were 12 for ONT-evidence, 11 for PASA-evidence, 10 for StringTie-evidence, 9 for Trinity-evidence, 5 for gth-evidence, and 2 for marker-evidence.
Finally, alternative splicing models and UTRs were added to EVMmodels by using PASApipeline in "genome annotation" mode.The downstream inspection and function annotation were based on the resulted PASA-models.
TransDecoder v5.5.0 (http://trans decod er.github.io)were used to predict the putative peptides for PASA-model and the results were aligned against to the Uniprot database by using blastp v2.10.0+ with E-value below 1e-6.Peptides with no hits and length below 100 amino acids were removed from the PASA-models.BUSCO v4.1.4with mammalian (odb10) dataset was used to evaluate the completeness of annotation (Manni et al., 2021).
Mus musculus was used as outgroup.For RT-qPCR, total RNA from all cell lines, mouse and bat splenic immune cells were extracted using EZ-press RNA Purification Kit (EZBioscience, B0004D).All cDNA was synthesized using HiScript III RT SuperMix for qPCR kit (Vazyme, R323).qPCR was performed using ChamQ SYBR qPCR Master Mix (Vazyme, Q311) on Quantagene q225 system (KUBO, China).All sequence of siRNA and RT-qPCR primers were listed in Table S10.

| LIP, non-heme iron assay, cell viability, and phagocytosis experiment
We first compared between bats and mice, where four male myotis bats and four mice were used for LIP assay.Then we compared between young and aged mice, where young (2 months old, male) and aged (20 months old, male) mice were fed with standard diet and high iron diet (8.3 g carbonyl iron/kg, Xietong Biotech) for 4 weeks, respectively.LIP of the splenic immune cells was evaluated as described previously.For each sample, 1 × 10 6 cells/ml in PBS with 0.1% BSA were incubated with 50 nM calcein-AM (Beyotime, C2012) for 15 min at 37°C in the dark for free iron chelation.Cells were immediately analyzed on a flow cytometer (Agilent Novocyte Advanteon), the mean florescence intensity (MFI) was measured at the FITC channel.Followed by incubation of 100 μM SIH (Blpbio, 495841) at room temperature for 30 min, cells were analyzed again.Cellular LIP level was estimated as the MFI difference (△F).
The NC and Pcbp1 knockdown immune cell lines were treated with different concentration of ferric citrate (HY-N1428C, MedChemExpress) for 4 days.Non-heme iron concentrations were determined using the Iron assay kit (Abcam, ab83366) following the manufacturer's instruction.Ferric and ferrous iron levels were calculated from the standard curve and normalized to the protein content for each sample using BCA Protein Assay Kit (C503021, Sangon Biotech).
For cell viability assay, cells from both NC and Pcbp1 shRNA knockdown groups were plated in 96-well-plate, 2500 cells/well.Cell viability was assayed using Cellcounting-Lite 2.0 Luminescent Cell Viability Assay (Vazyme, DD1101) following the manufacturer's instruction, cell proliferation ratio was calculated from Day 1 to Day 4.
After thorough washing, cells were visualized under a fluorescent microscope, then analyzed on a flow cytometer (Agilent Novocyte Advanteon) to detect the percentage of phagocytic cells.

| scRNA-seq analysis
scRNA-seq fastq files were demultiplexed to their respective barcodes using the 10 Genomics Cell Ranger mkfastq utility.Unique molecular identifier (UMI) counts were generated for each barcode using the Cell Ranger (v5.0.1) count utility.The bat reference genome was generated by ourselves.
Count data were processed with the Seurat package (v4.1.0).For quality control, genes that were expressed in fewer than three cells and cells that expressed fewer than 100 genes were excluded from analysis.Integration was performed using canonical correlation analysis (CCA), essentially according to the Standard Workflow protocol provided by the authors.The filtered gene-barcode matrix was analyzed by PCA.Then Uniform Manifold Approximation and Projection (UMAP) was performed on the top 30 principal components for visualizing the cells.Meanwhile, graph-based clustering was performed on the PCAreduced data for clustering analysis with Seurat.Clusters were identified by the expression of known cell-type markers: B cells (CD79A and CD79B); T cells (CD3E and CD3G); macrophages (CD68 and CD86); neutrophils (S100A8 and IL1R2); NK cells, (KLRC1 and NKG7); Ltf-high neutrophils: (CAMP and LTF), and erythroid cells (AHSP and TFRC).To identify differentially expressed genes between cell types across species, we used a negative binomial model as implemented in the Seurat 'FindMarkers' function, comparing each individual cell type to all other cells within the major cell type cluster (B cells, T cells, APCs, NK cell, and neutrophils).Genes were considered differentially expressed if the adjusted p value was lower than 0.05.The single-cell dataset of young and aged mice spleen was compared to get the upregulated genes in the young mouse splenic immune cells.Then we defined the upregulated genes in the long-lived species and in the young mice as candidate anti-immunosenescene genes.

| Bulk RNA-seq analysis
Reads were aligned to mm10 mouse genome (GENCODE vM23/ Ensembl 98) using STAR (Dobin et al., 2013), and reads falling on exons of genes were counted using RSEM.Differential expressed genes between different conditions was determined using DESeq2 (Love et al., 2014).Differentially expressed exons between different conditions was done using DEXSeq (Anders et al., 2012).The anti-immunosenescence scores and immunoscenescence scores of siRNA knockdown groups were calculated by AuCell (Aibar et al., 2017) using the genes upregulated in long-lived species and genes downregulated in long-lived species, respectively.Public bulk RNA-seq datasets of human splenic immune cells of different ages were downloaded from Encode.We downloaded the genes expression files from ENCSR265NZF, ENCSR700QVJ, ENCSR663IOE, EN-CSR510PSL, ENCSR910QOX, ENCSR330UMQ, ENCSR194HVU, ENCSR900SGE, and ENCSR106SZN experiment.

| Processing and analysis of public scATAC-seq datasets
scATAC-seq of CD45 + cells datasets from the spleen from young and aged male mice were downloaded from https://www.synapse.org repository (syn22255433).Fastq files were aligned to the mouse reference (mm10) and "cellranger-atac count" function (cellrangeratac, v2.0.0) was used to generate single-cell accessibility counts.
All downstream scATAC-seq analysis was conducted by using ArchR (Granja et al., 2021).Briefly, the resulting fragment files were used to create an Arrow file and create an ArchRProject after filtering the doublets.Dimensionality reduction and clustering analysis were performed by using the "addIterativeLSI" and "addClusters" function.UMAP was ran by using the "addUMAP" function and visualized by "plotEmbedding" function.Gene activity scores were generated based on the Gene Score model and clusters were annotated based on gene accessibility score of known marker genes.Pseudobulk replicates were made using the "addGroupCoverages" function.
Motif annotations were added using "addMotifAnnotations" function.Single-cell RNA data of young and aged mouse spleen from Mogilenko et al. (Mogilenko et al., 2021) were integrated analysis by "addGeneIntegrationMatrix" function.Peak-to-gene links were identified by using "addPeak2GeneLinks" function to find correlations between peak accessibility and gene expression.

| Enrichment analysis and TF analysis
Metascape was used to perform Gene ontology (GO) biological process enrichment analysis (Zhou et al., 2019).The enrichKEGG function from the clusterProfiler package (v4.4.4) (Wu et al., 2021) was used to perform KEGG pathway enrichment analysis.
For TF analysis, we used two methods to identify the candidate TFs that might regulate Pcbp1.First, Cistrome Toolkit (dbtoo lkit.cistrome.org) of existing genome-wide ChIP-seq data were used to probe which transcription factors might regulate Pcbp1 in mouse.The halfdecay distance to transcription start site was set as 10 kb.Second, SCENIC was used to identify regulons based on genes co-expression and DNA-motif analysis (Aibar et al., 2017).The single-cell data of

| Splenic immune cell atlas of long-lived and short-lived mammals
To systematically compare the similarities and differences of splenic immune cells between long-lived versus short-lived mammals, we combined scRNA-seq datasets sampled from myotis bats, NMRs, human, and mice (Figure 1a,b; Table S1).Among these datasets, bat splenic immune cells were newly generated from two adult male and two adult female myotis bats using 10X Genomics technique, while NMR, mouse, and human datasets were obtained from previous studies (He et al., 2020;Hilton et al., 2019).
Due to the lack of M. ricketti reference genome, we newly generated the bat genome by the following workflow.First, we extracted genomic DNA from muscle tissues of adult male bats and performed de novo assembly using third-generation long reads sequencing on Oxford Nanopore, then the contigs were corrected by Illumina nextgeneration sequencing (Figures 1b and S1a).Second, we preformed Hi-C sequencing to determine the chromosome borders (Figures 1c   and S1b).Third, we used 13 tissues of adult male bats to generate transcriptome data by both third-generation long reads and second-generation short reads for genome annotation (Figure S1c).
Finally, we obtained high-quality reference genome with the N50 values of 81.7 Mb for contigs and 107.7 Mb for scaffolds, and 21,273 protein-coding genes were annotated with >99.8% completeness (Figure 1d,e; Table S2).Thus, the quality of this genome was comparable to the best reference-quality genomes for bats so far (Jebb et al., 2020).We resolved the phylogenetic tree using our newly generated bat genome with reference-quality genomes of other bat species, and confirmed the right phylogenetic position of M. ricketti in Chiroptera (Figure 1f).
Based on our high-quality bat reference genome, we accurately quantified the gene expression of bat scRNA-seq dataset (Figure S1d), and integrated all single cells of these four species.After stringent quality control and batch effect correction by Harmony (Korsunsky et al., 2019), all integrated single cells could be grouped into seven

| High splenic B:T cell ratio implies immunosenescence
As shown in Figures 1j and S1g, the cellular composition of splenic immune cells varied significantly among these four species.The NK cell proportion of bat was significantly higher than those of other three species.NMR had more macrophages and neutrophils but no NK cells, which was consistent with previous results (Hilton et al., 2019).Of note, compared with long-lived species, mouse had more B cells but less T cells, namely a much higher B:T cell ratio, while the B:T cell ratio was comparable among other three long-lived species (Figure 1k).As previous studies have reported increased B:T cell ratio under pathological conditions (Becklund et al., 2016), we wondered whether the high B:T cell ratio exerted influence on aging.
We compared the proportions of five immune cell types and the B:T cell ratio between young and aged mouse splenic immune cells using another published scRNA-seq dataset (Mogilenko et al., 2021).Indeed, increased B cells and decreased T cells were detected during aging, resulting in a higher splenic B:T cell ratio in aged mice (Figure 1l).Therefore, splenic B:T cell ratio is higher in mouse than long-lived mammals and increases during aging, implying its correlation with immunosenescence.
To further reveal the heterogeneity of B cells and T cells, we re-clustered them and identified their respective subpopulations based on expression patterns of typical genes.T cells were subdivided into nine subpopulations, namely naïve CD4 + T cells, effector memory CD4 + T cells, regulatory T cells, Naïve CD8 + T cells, effector memory CD8 + T cells, GZMK + T cells, γδ T cells, NK T cells, and mucosal-associated invariant T cells (Figure S2a-c).As shown in Figure S2g, mice had more naïve CD8 + T cells, but less mucosalassociated invariant T cells in their splenic immune system compared with long-lived species.B cells were subdivided into six subpopulations, namely naïve B cells, transitional B cells, marginal zone B cells, follicular B cells, memory B cells, and plasma B cells (Figure S2d-f).
Long-lived species had a significantly higher proportion of naïve B cells than mice, whereas mice had a higher proportion of follicular B cells (Figure S2h).

| Iron-related genes exhibit prominent anti-immunosenescence feature
Next, we attempted to identity candidate anti-immunosenescence genes by comparative transcriptomic analysis of long-lived versus short-lived mammals.To reduce false positives, candidate genes should meet two criteria.(i) They should exhibit higher expression levels in all these three long-lived species compared with mouse.
Because if a certain gene is more highly expressed in only one long-lived species, it may be due to species-specific difference.
However, if a certain gene exhibits higher expression level in all the three distantly related species: bat (chiroptera), NMR (rodentia), and human (primates) than mouse (rodentia), it is more likely to involve in the common phenotype of these three long-lived species, namely, longevity, rather than species difference.(ii) The expression levels of candidate genes should significantly decrease during aging.
Following the above two criteria, we first performed differential expression analysis for every cell type by comparing each of these three long-lived species with mouse, and only selected upregulated genes that are shared by all these three long-lived species (Figure 2a,b; Table S3).Then, to check whether their expression levels significantly decrease during aging, we reanalyzed the mouse single-cell dataset including young and aged splenic immune cells (Mogilenko et al., 2021).We obtained numbers of genes conforming to the above two criteria: higher expression levels in the long-lived species than in mice and in young mice than in aged mice (Figure 2a; Table S4).Finally, we manually checked the expression patterns of these candidate genes in each cell type among these four species, and identified six genes for macrophages (PCBP1, SRSF4, SRSF9, NDFIP1, DOCK8, and RAC1), four genes for T cells (PCBP1, SUN2, SOCS1, and ISCU), five genes for B cells (PCBP1, YWHAB, PPP1CC, SNX9, and ISCU), five genes for NK cells (PCBP1, PPP1CC, ATP1A1, ARF6, and PRRC2B) (Figure 2c, heme synthesis or implement different biological redox reactions (Ryu et al., 2018).The dysfunction of PCBP1 may result in extra free iron release and increase lipid peroxidation which induces pathologies such as liver steatosis (Protchenko et al., 2021).ISCU, ironsulfur cluster assembly enzyme, encodes a scaffold protein for the assembly and delivery of iron-sulfur (Fe-S) clusters, which regulate metabolism and iron homeostasis.Decreased ISCU expression can reduce the translation of the iron storage protein FTH1 and lead to iron accumulation in liver cancer (Funauchi et al., 2015).NDFIP1 mediates ubiquitination and degradation of the ferrous iron transporter DMT1, which is responsible for importing non-heme iron into cells (Jia et al., 2015).NDFIP1 also involves in iron homeostasis by ubiq- iron-related genes play an important role in counteracting immunosenescence changes.
We also investigated the genes that increase with aging, which may also involve in immunosenescence.We identified genes that exhibit higher expression levels in mice than longlived species, and in aged than young mice.In total, 69, 11, 23, 122, and 9 candidate genes were found in T cells, B cells, NK cells, and macrophages, respectively (Figure S3c; Table S4).
Next, we filtered genes with low expression level and low fold change, and finally identified three genes for macrophages (HP, CKB, and CD79B), five genes for T cells (ITM2C, CAPNS1, SDF4, TSPAN13, and SMPD13A), three genes for B cells (UNC93B1, FC-GR2B, and NAPSA), seven genes for NK cells (ITM2C, LTB, TCF7, SNAP23, PLGRKT, CD27 and MFSD10) (Figure S3d,e).Although these genes may be potential target genes of immunosenescence, it is difficult to distinguish them between mouse speciesspecific and immunosenescence features.Therefore, we next mainly focused on candidate genes that were highly expressed in all these three long-lived species, which theoretically had much lower false positives.

| Splenic immune cells of mice have more free iron than those of bats
As iron-related genes were related to splenic anti-immunosenescence, we performed iron experiments in splenic immune cells of myotis bats and mice.We measured free iron levels in the form of labile iron pool (LIP) in splenic immune cells by cell cytometry.Compared with bat splenic immune cells, mouse splenic immune cells contained much more free iron (Figure 2d).These results suggested that mouse seems to suffer from higher level of ROS-induced damage produced by free iron, which may contribute to immunosenescence and undermine longevity (Mangan, 2021).
We also measured free iron levels in splenic immune cells of young and aged mice, but detected no differences (Figure 2e).However, after 4 weeks of high iron diet feeding, we detected higher free iron levels in aged mice than in young mice (Figure 2f).This result indicates that the ability of maintaining iron homeostasis in splenic immune cells of aged mice is reduced, which may promote immunosenescence.

| Iron homeostasis is crucial for anti-immunosenescence
To explore the important roles of these anti-immunosenescence genes in immune cells, we took macrophage as an example and performed siRNA perturbation experiments in mouse macrophage cell line, Raw264.7.Six anti-immunosenescence genes identified in splenic macrophages were knocked down separately, namely, Pcbp1, Srsf4, Srsf9, Ndfip1, Dock8, and Rac1.As shown in Figure 3a, the knockdown efficiency was pretty high, indicating the success of siRNA experiments.Next, we profiled the transcriptomes of these samples, and calculated the anti-immunosenescence and immunosenescence scores using upregulated and downregulated genes identified in splenic macrophages of long-lived species compared with mouse, respectively.Compared with negative control samples, nearly all siRNA experiment samples obtained higher immunosenescence scores but lower anti-immunosenescence scores, except Ndfip1 (Figure 3b).These results suggested that the knockdown of these anti-immunosenescence genes would drive cellular state transition from long-lived mammal mode to shortlived mammal mode.
Unexpectedly, although other four genes (i.e., Srsf4, Srsf9, Dock8, and Rac1) were not previously reported to be iron related, their knockdown also severely affected the iron homeostasis in the macrophage cell line (Figure 3c).For example, the knockdown of Dock8 increased the expression of B2m, Cyb561, Fxn, Ftl1, Hmox1, Tfrc, and Slc40a1; the knockdown of Rac1 increased the expression of genes related to iron ion homeostasis including Fxn,Fth1,Ftl1,Hmox1,Frrs1,Tfrc,Glrx3,Slc40a1,Steap3,Slc48a,Glrx3,Ciao2a,Ciao2b,Isca1,Ciapin1.Also, Slc25a37 and Slc25a28 were downregulated in the Rac1 knockdown group compared with the control group, and both genes were involved in iron import into the mitochondrion (Seguin et al., 2020).It was worth noting that genes related to iron-sulfur cluster assembly were all upregulated in the Srsf4 and Srsf9 knockdown groups, including Glrx3, Ciao2b, Ndufab1, and Ciapin1.Meanwhile, as Srsf4 and Srsf9 are crucial in alternative splice site selection during pre-mRNA splicing (Jeong, 2017), we wondered whether they also played important roles in the alternative splice of iron-related genes.As shown in Figure S4b,c, the knockdown of Srsf4 and Srsf9 indeed affected alternative splicing of some iron homeostasis-related genes.For example, they both affected Fth1 exon usage: in the Srsf4 knockdown group, the tenth, eleventh and thirteenth exon were affected, while in the Srsf9 knockdown group, the ninth and eleventh exon were affected (Figure 3d).
Importantly, we also detected regulatory relationship among these anti-immunosenescence genes.We found that Pcbp1 sat in the center and interacted with other five anti-immunosenescence genes: the knockdown of Pcbp1 resulted in the downregulation of Ndfip1, Srsf4 and Dock8, and upregulation of Srsf9 and Rac1 (Figure 3c).Besides, the protein-protein interaction network also supported their close relationship (Figure 3e).To sum up, all these six identified anti-immunosenescence genes in splenic macrophages are related to cellular iron homeostasis, and the perturbation of any of them would break the balance of cellular iron trafficking.Thus, their differential expressions between long-lived and short-lived mammals, and between young and aged mice, highlight the importance of iron homeostasis in splenic anti-immunosenescence.

| PCBP1 perturbation results in immunosenescence and iron dyshomeostasis
As PCBP1 was upregulated in all splenic immune cell types of long-lived species compared with mouse, we next mainly focused on its role in anti-immunosenescence.To validate the accuracy of scRNA-seq results, we performed qPCR experiments for PCBP1 in bat and mouse splenic immune cells, and also in young and aged mouse splenic immune cells.qPCR results showed higher expression of PCBP1 in bats than mice, and in young than aged mice, which were consistent with scRNA-seq results (Figure 4a,b).We also explored the Encode dataset and detected similar expression pattern in human splenic immune cells: the expression level of PCBP1 dramatically decreased during aging in human (Figure 4c).These results indicate that PCBP1 function is conserved between human and mouse splenic immune systems.
We then knocked down PCBP1 in multiple cell lines to verify its effect on anti-immunosenescence (Figure S5a-d).Consistent with mouse macrophage cell line (Raw246.7),Pcbp1 knockdown in mouse microglial and B cell lines (EOC20 and WEHI-231), and human T cell line (Jurkat) also resulted in higher immunosenescence scores but lower anti-immunosenescence scores (Figure 4d,e).Moreover, KEGG analysis showed that Pcbp1 downregulation in these immune cell lines all enriched the terms of aging-related neurodegenerative diseases and ferroptosis, which again confirmed the critical role of Pcbp1 in antiimmunosenescence and iron homeostasis (Figure 4f; Tables S7 and S8).
Besides, we also tested the cell viability, changes in iron levels, ferroptosis, and related immune function in immune cell lines with Pcbp1 knockdown.As shown in Figure 4g, the cell viability of T cells was largely decreased after knockdown, while there were no significant differences in the cell viability of macrophages and B cells.We measured non-heme iron in immune cell lines by colorimetric assay, and also tested iron tolerance by adding ferric citrate to the medium for 96 h of culture.Both ferrous and ferric iron levels exhibited significant differences between Pcbp1 knockdown and negative control groups, though the iron patterns varied across different immune cell lines (Figure 4h).Meanwhile, we also performed qPCR experiments for Gpx4 and Slc7a11, whose decreases imply ferroptosis.These two genes exhibited lower expression levels in macrophages of Pcbp1 knockdown group than negative control group when adding 0.1 mM ferric citrate, while in B and T cells, similar patterns appeared until ferric citrate level reached 1 mM (Figure 4i).Last but not least, after Pcbp1 knockdown, we detected reduced phagocytosis ability of macrophages (Figures 4j,k and S5e), and immune function-related genes were significantly changed (Figure S5b).Together, these results further suggest that PCBP1 perturbation would lead to immunosenescence and iron dyshomeostasis in immune cells.

| PCBP1 is also downregulated during aging in hepatic immune cells
Because spleen and liver are both crucial for iron recycling, we wondered whether hepatic immune cells also have similar expression pattern of Pcbp1 to splenic immune cells.As shown in Figure 5a, Pcbp1 exhibited resembling expression patterns in mouse splenic and hepatic immune cells, with dramatic decrease during aging.As control, there were no expression differences of Pcbp1 in pulmonary immune cells between young and aged mice.In addition, the expression level of PCBP1 was also higher in human than in mouse hepatic immune cells, which was consistent with splenic expression pattern (Figure 5b).Moreover, we reanalyzed human hepatic immune single cell data from a previous study (Guilliams et al., 2022), and detected a downregulation of PCBP1 in human hepatic macrophages during aging (Figure 5c).These results confirm important roles of spleen and liver in iron recycling, and suggest that PCBP1 function is conserved between human and mouse hepatic immune systems.
To test whether the role of Pcbp1 is specific to splenic and hepatic immune systems or also applies to other systems, we examined the expression pattern of Pcbp1 across multiple organs/ tissues using mouse aging cell atlas (Tabula Muris Senis) (The Tabula Muris Consortium et al., 2018).We found that Pcbp1 expression also decreased with aging in immune cells of thymus and pancreas, but there was no similar expression pattern in nonimmune cells (Figure 5d).This suggested that the role of Pcbp1 may be specific to the immune systems of certain organs/tissues in mice.Overall, the anti-immunosenescence feature of PCBP1 in both splenic and hepatic immune cells was conserved between human and mouse, and the decreased expression of PCBP1 during aging in immune cells of these two major iron recycling organs further demonstrated that iron dyshomeostasis would aggravate immunosenescence.
As PCBP2-4 are the paralogues of PCBP1, we also explored the expression pattern of PCBP2-4 in splenic immune system across these four species, and between young and aged mice.PCBP3 and 4 are barely expressed in splenic immune cells, while PCBP2 expression declines in mouse splenic immune cells during aging (Figure S5f).
However, PCBP2 expression was much lower in splenic immune cells of bats than human, NMRs and even mice.Thus, PCBP2 does not meet our two criteria for screening candidate genes, and we did not further investigate its role in anti-immunosenescence.

| PCBP1 can be regulated by three transcription regulators
Now that we have obtained evidence that PCBP1 is crucial for keeping iron homeostasis and resisting immunosenescence in splenic and hepatic immune cells, we are also eager to know how it is regulated during aging.First, we used Cistrome DB Toolkit to search which transcription factors (TFs) might regulate Pcbp1 (Zheng et al., 2019).Next, we used SCENIC, a tool to reconstruct gene regulatory networks based on co-expression and DNA motif analysis, to find TFs that might regulate Pcbp1 (Aibar et al., 2017).6b).
To verify the regulatory relationship between Rela, Ctcf, and Pcbp1, we knocked down these two TFs in three immune cell lines, namely, Raw264.7,EOC20, and WEHI231 (Figure 6c,d).As expected, the expression level of Pcbp1 in the experimental group was significantly lower than that in the control group.In addition, we reanalyzed mouse splenic scATAC-seq data from a previous study (Mogilenko et al., 2021).Based on the classic marker genes, we annotated four immune cell types including macrophages, NK cells, B cells, and T cells (Figure S6c,d).We found that Rela and Ctcf can simultaneously bind to the same region, and the accessibility of this region was higher in macrophages of young mice than aged mice (Figure 6e,f), though no difference was detected in other immune cell types (Figure S6e).Meanwhile, using the peak2gene function, we also identified a lncRNA, 2310040G24Rik, that could potentially regulate Pcpb1 at its distal downstream.Moreover, the accessibility of this region is higher in aged than in young mice (Figure 6g).Although the expression level of this lncRNA in mouse splenic immune cells did not significantly increase with aging, the expression of Pcbp1 was higher in the lncRNA knockdown group than in the control group, indicating its negative regulatory role of Pcbp1 (Figure 6h,i).

| DISCUSS ION
The immune system plays a critical role in identifying and destroying pathogens and nascent malignancies, which is important for organismal homeostasis and longevity (Borgoni et al., 2021).
Age-associated immune decline (i.e., immunosenescence) is concomitant with an age-dependent increase in chronic diseases in most mammalian species (Borgoni et al., 2021).A commonly used method to study immunosenescence is to make comparison between young and aged model animals (Mogilenko et al., 2021).Although such studies involving in same species with different ages can largely eliminate the effect of false positives, they are also easy to overlook some details that are biologically important but not statistically significant.Another way to tackle immunosenescence is to perform cross-species comparison between long-lived and short-lived species, such as long-lived naked mole rat or blind mole-rat versus short-lived mouse or rat (Hilton et al., 2019;Izraelson et al., 2021).However, it is difficult to distinguish between species-specific differences and true immunosenescence-related outcomes.To overcome above problems, we first build the highquality reference genome for a long-lived myotis bat, and then use three long-lived and distantly related species (i.e., bat (chiroptera), NMR (rodentia), and human (primates)) to perform comparison with the short-lived mammal, mouse (rodentia), in splenic immune cells.Furthermore, anti-immunosenescence candidates are filtered by checking their patterns in young and aged mice and/ or human.By doing so, these identified molecular clues of antiimmunosenescence is more likely to involve in the common phenotype of these long-lived species, namely, longevity, rather than species differences.
One of our noticeable findings is that the B:T cell ratio is significantly higher in mouse than other three long-lived species, and also higher in aged than young mice, implying its correlation with immunosenescence.This result is in concordance with the declined production and proliferation of naïve T cells, and the deterioration of the supportive stromal cells in lymph node upon aging (Becklund et al., 2016).Lower T cell level in mouse may render the immune system more susceptible to pathogens and age-related inflammatory factors, hence limiting their lifespan.On the other hand, previous studies found that a new subpopulation of B cells increases with age in mice, which were termed age-associated B cells (Hao et al., 2011;Rubtsov et al., 2011).These cells can be directly involved in the production of autoantibodies, causing severe autoimmunity and thus promoting aging (Rubtsov et al., 2011).In addition, age-associated B cells can secrete antibodies in response to toll like receptor stimulation, causing an inflammatory response (Hao et al., 2011).Higher B cell level in mouse may trigger higher autoimmunity and inflammatory response, hence promoting aging.
Furthermore, we identify several iron-related genes exhibiting prominent anti-immunosenescence feature: higher expression in long-lived species and decreased expression during aging.By crossspecies comparison and manually checking the expression patterns, a total of 20 candidate anti-immunosenescence genes are identified in different splenic immune cell populations.Unexpectedly, in addition to some known iron-related genes (e.g., PCBP1, ISCU, and NDFIP1), other genes that have not been previously reported to be iron related are also able to regulate iron homeostasis.For example, although there is no evidence that Srsf4, Srsf9, Dock8, and Rac1 are iron related, knocking down each of them severely affects the iron homeostasis in the macrophage cell line.Besides, we find that bat splenic immune cells contained less free iron but more protein bound iron than mouse splenic immune cells, indicating bat processes stronger ability to control oxidative stress induced by free iron, which may contribute to their longevity (Zeidan et al., 2021).
Together, these results highlight the importance of iron homeostasis in splenic anti-immunosenescence.
Among these anti-immunosenescence genes, the iron chaperone PCBP1 best meets the criteria for anti-immunosenescence.
PCBP1 displays higher expression levels in various immune cell types of the long-lived species in spleen and liver, both of which are important organs participating in iron storage and recycling, but not in lung.Also, PCBP1 expression decreases with age in both splenic and hepatic macrophages of human and mouse.Furthermore, the knockdown of Pcbp1 in multiple cell lines verify its effect on iron metabolism and anti-immunosenescence.In a recent study, PCBP1 protects hepatocytes from lipid peroxidation and oxidative-stress by controlling the labile iron pool (Protchenko et al., 2021).Moreover, PCBP1 takes part in numerous immune functions.For example, PCBP1 modulates innate immune response by promoting cGAS binding to viral DNA (Liao et al., 2021).PCBP1 is also upregulated in activated T cells and prevents the conversion of effector T cells to regulatory T cells (Ansa-Addo et al., 2020).
Therefore, the steady expression of PCBP1 in splenic immune cells may enable a healthy immune system, and thus is important for anti-immunosenescence.previous study found that the loss of Rela would cause ferroptosis in intestinal epithelial cells (Xu et al., 2020).Meanwhile, Ctcf deficiency leads to a decrease in cytochrome C and ISCU proteins, which leaded to ROS accumulation and lipid peroxidation (Roy et al., 2018).Iron accumulation and the increase of reactive oxygen species (ROS) produced by lipid peroxidation can trigger ferroptosis (Tang et al., 2021).Consistent with previous studies, comparison, immunosenescence, iron homeostasis, PCBP1, single-cell transcriptomics, splenic immune cells | 3 of 18 HE et al.

2. 6 |
Cell line culture, siRNA and shRNA knockdown, and RT-qPCR The mouse macrophage cell line Raw264.7 and the microglial cell line EOC20 were grew in DMEM (high glucose) supplemented with 10% FBS and 1% penicillin-streptomycin.The mouse B cell line WEHI231 and human T cell line Jurkat were grew in RPMI 1640 supplemented with 10% FBS and 1% penicillin-streptomycin.Transfection of siRNAs was preformed using Lipofectamine RNAiMAX Reagent (Invitrogen, 13778150) according to the manufacturer's instruction, cell samples were collected 48 h after transfection.Pcbp1 knockdown in Raw264.7,WEHI231 and Jurkat cell lines were established by transduction of shRNA lentivirus.Nontarget control (NC) and Pcbp1 shRNA sequence were cloned into pLVX-shRNA2-ZsGreen1-Puro vector.Package of lentivirus was performed by transfecting shRNA vectors into 293T cells using Lipofectamine 3000 Reagent (L3000015, Invitrogen), 48 h after transfection, the supernatants containing lentivirus were collected and concentrated using 5 × Lentivirus Concentration Reagent (C103, GenStar), then stored in aliquots at −80°C before use.
1.0) to check the average expression of PCBP1 in human liver macrophages at different ages.Single-cell data of different tissues of mouse were obtained from Tabula Muris Senis (The Tabula Muris Consortium et al. 2018).We checked the expression pattern of Pcbp1 in different tissues at different ages based on data from the microfluidic droplet approach.
mouse spleen were used to run SCENIC.The candidate transcription factors that might regulate Pcbp1 were collected.Finally, the top 20 transcription factors found in Cistrome Toolkit intersected with all TFs found in SCENIC, and the resulting transcription factors were regarding as the final candidate transcription factors for Pcbp1.

F
I G U R E 1 Cross-species expression atlas of splenic immune cells.(a) Correlation between lifespan and body mass across various mammalian species.Data were downloaded from the AnAge database.(b) Schematic of experimental design showing the workflow of this study.Single-cell datasets of human, NMR, and mouse are acquired from previous studies.(c) Hi-C map of myotis bat.(d) Genome assembly contiguity of myotis bat.(e) BUSCO assessment showing completeness of the gene annotation of myotis bat.(f) Phylogenetics tree of myotis bat among mammals.(g and h) Integrated (g) and separated (h) UMAP visualization of splenic immune cells isolated from myotis bat, NMR, human, and mouse.(i) Dotplots showing marker genes expression of each cell type.(j) Bar charts showing the cell type composition of splenic immune cells across four species.(k) B:T cell ratio in splenic immune cells across the four species.Data are mean values ± SD. *** indicates p < 0.001 (Student's t test).(l) Proportions of five immune cell types (left) and B:T cell ratio (right) in splenic immune cells compared between young and aged mouse.major clusters using graph-based algorithm in Seurat (Figures 1g,hand S1e) (Hao et al., 2021).Based on classic marker genes, we annotated these clusters as B cells (CD79A and CD79B), T cells (CD3E and CD3G), macrophages (CD68 and CD86), neutrophils (S100A8 and IL1R2), NK cells (KLRC1 and NKG7), Ltf-high neutrophils (CAMP and LTF), and erythroid cells (AHSP and TFRC) (Figures 1i and S1f ).

Figure
Figure S3a,b).Importantly, three candidate genes showed antiimmunosenescence feature in more than one cell types: PCBP1 in all splenic immune cell types, ISCU in T and B cells, and PPP1CC in B and NK cells, indicating a universal anti-immunosenescence role (Figure 2c).Besides, PCBP1, ISCU and NDFIP1 are also important iron-related genes.PCBP1 encoded protein is an important iron chaperone, which could carry and distribute free iron in the form of glutathione linked ferrous (GSH-Fe II) to other proteins for storage, uitination and degradation of the iron exporter ferroportin (FPN, encoded by SLC40A1) on cell membrane, and NDFIP1 deficiency increases ferroportin in hepatic macrophages and increase circulating iron(Traeger et al., 2021).Together, these results indicated that F I G U R E 2 Identification of candidate anti-immunosenescence genes.(a) Procedures for screening anti-immunosenescence genes.First, comparison between long-lived species (bat, NMR and human) and mouse was performed, and the shared upregulated DEGs by all these three long-lived species in different cell types were shown in venn plots (left).Then, comparison between young and aged mouse was performed to obtain the upregulated DEGs in young mouse (middle), and after filtering, the candidate anti-immunosenescence genes were finally identified (right).(b) Heatmaps showing the upregulated DEGs in long-lived species (bat, NMR, and human) compared with mouse.(c) Dotplots showing expression patterns of the anti-immunosenescence genes in macrophages, T cells, B cells, and NK cells.Comparisons were made across the four species (up), and between the young and aged mouse (bottom).(d) Free iron content ratio in mouse and bat splenic immune cells measured by LIP assay.Three C57BL/6J mice and four myotis bats were used.*** indicates p < 0.001 (Student's t test).(e) Free iron content ratio in mouse splenic immune cells measured by LIP assay.Three young (2 months old) and three aged (20 months old) C57BL/6J mice were used.ns indicates nonsignificant (Student's t test).(f) Free iron content ratio in mouse splenic immune cells after 4 weeks of high iron diet feeding measured by LIP assay.Two young (2 months old) and two (20 months old) C57BL/6J mice were used.* indicates p < 0.05 (Student's t test).

F
I G U R E 3 Anti-immunosenescence gene knockdown results in iron dyshomeostasis in macrophages.(a) siRNA knockdown of Pcbp1, Srsf4, Srsf9, Ndfip1, Dock8, and Rac1 in Raw264.7 cell line, non-target siRNA was used as negative control (NC).Bulk RNA-seq was preformed 48 h after transfection.Knockdown efficiency was evaluated by comparison between knockdown and NC groups.Three biological replicates were conducted for each group.Data are mean values ± SD. *** indicates p < 0.001 (Wald test).(b) The anti-immunosenescence (left) and immunosenescence (right) scores of anti-immunosenescence genes knockdown in Raw264.7 cell line, calculated by AuCell using upregulated and downregulated genes in long-lived species, respectively.(c) Heatmaps of iron-related genes selected from DEGs between the knockdown samples of anti-immunosenescence genes and negative control samples in Raw264.7 cell line.(d) Gene location and exon usage change of Fth1 after Srsf4 and Srsf9 knockdown in Raw264.7 cell line.The red lines and blue lines indicate the level of exon usage in the negative control and knockdown groups, respectively.The red dashed lines indicate the affected exons that involved in potential alternative splicing, labeled in purple color, with exon reads count ≥10 at FDR ≤0.05.(e) Protein-protein interaction network of the six antiimmunosenescence genes identified in splenic macrophages, with Pcbp1 sitting in the interaction center.Red arrows indicate downregulation and blue arrows indicate upregulation after Pcbp1 knockdown in Raw264.7 cell line, related to Figure 3c.| 11 of 18 HE et al.
These results indicate that Rela and Ctcf can regulate the expression of Pcbp1 in splenic immune cells, and their decreased expression during aging may result in the downregulation of Pcbp1 and splenic immunosenescence.

F
The important role of PCBP1 in anti-immunosenescence.(a-b) The expression levels of PCBP1 in splenic immune cells evaluated by RT-qPCR compared between mouse and bat (b), young and aged mice (b).Four biological replicates, each with three technical repeats, were conducted in each sample.Data are mean values ± SD. *** indicates p < 0.001 (Student's t test).(c) Bar plot showing the change of PCBP1 expression in human splenic immune cells.Data were downloaded from Encode.(d-e) The anti-immunosenescence scores (d) and immunoscenescence scores (e) of Pcbp1 knockdown in Raw264.7,EOC20, WEHI231, and Jurkat cell lines, calculated by AuCell using the genes upregulated and downregulated in long-lived species, respectively.(f) Enriched KEGG terms using all DEGs from Pcbp1 knockdown in Raw264.7,EOC20, WEHI231, and Jurkat cell lines (right), respectively.p-adjusted value <0.05.(g) Cell proliferation ratios after shRNA knockdown of Pcbp1 in Raw264.7,WEHI23,1 and Jurkat cell lines.(h) Non-heme iron contents of cells after shRNA knockdown of Pcbp1 and treatment with different concentrations of ferric citrate.(i) Gene expression quantified by RT-qPCR in cells after knockdown of Pcbp1 and treatment with different concentrations of ferric citrate.Four biological replicates, each with three technical repeats, were conducted in each sample.Data are mean values ± SD. * indicates <0.05, ** indicates p < 0.01 (Student's t test).(j and k) Phagocytosis in Pcbp1 knockdown Raw264.7 cells after fluorescent zymosan particles treatment for 3 h.Phagocytic cells visualization (j) and calculation by flow cytometry.Three technical repeats were conducted in each sample.Data are mean values ± SD. * indicates <0.05, ns indicates nonsignificant (Student's t test).F I G U R E 5 The expression pattern of PCBP1 in different tissues.(a) Violin plots displaying Pcbp1 expression level in four immune cell types (macrophages, B cells, NK cells, and T cells) from three organs (spleen, liver, and lung) of young and aged mice.(b) Violin plots displaying the expression level of PCBP1 in four immune cell types from human and mouse livers.(c) Bar plots showing the average expression level of PCBP1 in human hepatic macrophages changes with age.Data were downloaded from Liver Cell Atlas (https://liver cella tlas.org/download.php).(d) Violin plots showing the expression level of Pcbp1 in immune and non-immune cells of different mouse tissues.Data were downloaded from Tabula Muris Senis.F I G U R E 6 Identification of TFs and lncRNA that regulate Pcbp1 in mouse splenic immune cells during aging.(a) Intersecting the analysis results of regulatory TFs using Cistrome Toolkit and SCENIC.(b) Violin plots showing the expression levels of Rela and Ctcf in splenic immune cell types of young and aged mice.(c and d) RT-qPCR analysis for relative expression of Pcbp1 after Ctcf (c) or Rela (d) knockdown in Raw264.7,EOC20, and WEHI231 cell lines.Three biological replicates, each with three technical repeats, were conducted in each sample.**** indicates p < 0.0001, *** indicates p < 0.001, ** indicates p < 0.01 (Student's t test).(e) Chromatin accessibility landscape of the Pcbp1 locus generated by scATAC-seq of young and aged mouse.The binding region of Rela and Ctcf was located downstream of Pcbp1, circled by dashed red line, which showed higher accessibility in young than aged mice.(f) Binding motifs of RELA and CTCF from JASPAR database.(g) Genome track visualization of 2310040G24Rik and Pcbp1 locus.The loops represent the significance of peak-to-gene links, indicating the peak near 2310040G24Rik is associated to Pcbp1 and its accessibility is higher in aged mouse.(h) RT-qPCR results showing the expression levels of 2310040G24Rik in splenic immune cells of young and aged mice.(i) RT-qPCR analysis for the relative expression of 2310040G24Rik (left) and Pcbp1 (right) after 2310040G24Rik knockdown in Raw264.7 cell line.Four biological replicates, each with three technical repeats, were conducted in each sample.*** indicates p < 0.001, * indicates p < 0.05 (Student's t test).

Finally, we
find Pcbp1 can be regulated by two TFs, Rela and Ctcf in mouse splenic immune cells.Knocking down Rela or Ctcf significantly downregulate the expression of Pcbp1, and decreased expression levels of Rela and Ctcf are detected during aging.A , Rela, Ctcf, Stat1, and Ezh2 were identified by intersecting the outputs from these two methods (Figures6a and S6a,b; TableS9), among which Rela and Ctcf were downregulated in mouse splenic