Single‐cell multi‐omics analysis presents the landscape of peripheral blood T‐cell subsets in human chronic prostatitis/chronic pelvic pain syndrome

Abstract Cumulative evidence suggests that abnormal differentiation of T lymphocytes influences the pathogenesis of chronic prostatitis/chronic pelvic pain syndrome (CP/CPPS). Thus, understanding the immune activation landscape of CP/CPPS would be helpful for improving therapeutic strategies. Here, we utilized BD™ AbSeq to digitally quantify both the protein and mRNA expression levels in single peripheral blood T cells from two CP/CPPS patients and two healthy controls. We utilized an integrated strategy based on canonical correlation analysis of 10 000+ AbSeq profiles and identified fifteen unique T‐cell subpopulations. Notably, we found that the proportion of cluster 0 in the CP/CPPS group (30.35%) was significantly increased compared with the proportion in the healthy control group (9.38%); cluster 0 was defined as effector T cells based on differentially expressed genes/proteins. Flow cytometry assays confirmed that the proportions of effector T‐cell subpopulations, particularly central memory T cells, T helper (Th)1, Th17 and Th22 cells, in the peripheral blood mononuclear cell populations of patients with CP/CPPS were significantly increased compared with those of healthy controls (P < 0.05), further confirming that aberration of effector T cells possibly leads to or intensifies CP/CPPS. Our results provide novel insights into the underlying mechanisms of CP/CPPS, which will be beneficial for its treatment.


| INTRODUC TI ON
Chronic pelvic pain syndrome (CPPS) is characterized by pain perceived in structures related to the pelvis for at least six months without proven infection or other obvious local pathology, 1 and it is highly prevalent, affects millions of people worldwide and impairs the quality of life in a manner similar to that observed with congestive heart failure, Crohn's disease, diabetes mellitus or angina. 2 Although multimodal therapies seem to be the most successful, 3 CPPS remains a major challenge and is a source of frustration for both patients and physicians, as available treatments often fail and a 'panacea' is still lacking. Currently, the lack of a recognized model is one of the most urgent problems in the mechanistic study of prostatitis.
The most common type of prostatitis is category III, also known as chronic prostatitis (CP)/CPPS. Diverse aetiologies regarding the pathogenesis of CP/CPPS have been proposed, which suggests that immune, neurological, endocrine and psychological factors may be involved. Of the suggested theories on the aetiology of CP/CPPS, the autoimmune basis is dominant. In addition, animal models have also been used to test autoimmunity. In rats, experimental autoimmune prostatitis could be induced by immunization with a male accessory gland blend, which is a mixed organ homogenate of the prostate (ventral, dorsal and lateral), and the coagulating gland. 4 This induction is characterized by specific cell-mediated responses and the infiltration of CD4+ and CD8+ T cells. 5 The T cells sensitized by the male accessory gland are able to expand and differentiate by themselves in response to homogenates of the prostate. 6 Lymphocytic infiltration of the stroma and periglandular region of the dorsal and lateral lobes is induced after immunization with only the rat ventral prostate. 7 Notably, parallel findings are observed among humans, indicating that a similar process occurs.
A study found that seminal antigens generated from normal asymptomatic men promote the proliferation of T cells that have been extracted from patients with CP/CPPS, although this phenomenon is not seen in healthy control T cells. 8 A subsequent study found that the prostate-specific antigen (PSA) in 5/14 patients increased T-cell proliferation, whereas there was no response to prostatic acid phosphatase. 9 These studies suggest an autoimmune basis for the aetiology of CP/CPPS. Interestingly, approximately 9% of spermatozoa are described as intraprostatic upon autopsy, and they mainly exist in the peripheral zone. 10 Thus, as these non-prostatic substances, such as spermatozoa, exist in the prostate, there is support for the hypothesis that the antigens leading to autoimmune response may not only be of prostatic origin. Testing the T-cell status underlying the pathogenesis of CP/CPPS is beneficial for identifying the pathogenic antigens.
Increasing evidence suggests that the abnormal differentiation of T cells influences the pathogenesis of CP/CPPS. Understanding the immune activation landscape of CP/CPPS is helpful for developing novel therapeutic strategies. [11][12][13] Single-cell multi-omics (protein and mRNA) sequencing is a powerful tool capable of producing a complete catalog of cell types and states present within a sample. Here, we employed BD™ AbSeq on the Rhapsody™ platform in the analysis of peripheral blood T cells from patients with CP/CPPS and healthy individuals. Our findings delineate the complex array of T-cell status in human blood and suggest that effector T cells, particularly central memory T cells, T helper (Th)1, Th17, and Th22 cells, as well as Treg cells, might be the causes of CP/CPPS.

| Patient recruitment
The present study was approved by the ethics committee of the First Affiliated Hospital of Anhui Medical University. Patients were recruited from the outpatient department, and controls were healthy donors. Informed consent was obtained from the participants before experiments were performed. Two patients with CP/ CPPS and two healthy adults have included in single-cell multiomics sequencing.

| Sample preparation
Human peripheral blood samples were centrifuged in a Ficoll gradient to isolate peripheral blood mononuclear cells (PBMCs) ( Figure 1A). Then, to obtain a high-resolution identification of the T-cell subsets, PBMCs were subjected to magnetic-activated cell sorting (MACS) with human CD3 MicroBeads (130-050-101, Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer's instructions. Cell viability was evaluated by Trypan blue staining. Then, the cells were washed and resuspended in PBS for BD Rhapsody™ System multi-omics sequencing. Single-cell capture was achieved by a random distribution of a single-cell suspension across over 200 000 microwells through a limited dilution approach. Beads with oligonucleotide barcodes were added to saturation to ensure that a bead was paired with a cell in a microwell. Cell lysis buffer was used to release polyadenylated RNA molecules and hybridize RNA to the beads. Then, beads were harvested into a single tube for reverse transcription. Upon cDNA synthesis, each cDNA molecule was tagged on the 5' end with a molecular index and a label indicating its cell of origin. Libraries were prepared using the single-cell amplification workflow, which is specifically demonstrated in Figure 1B

| Data analysis
Read quality was qualified and filtered based on read length (R1 <66 or R2 <64, the ratio of R1/R2), mean base quality score (read score <20) and single-nucleotide frequency (SNF, R1_SNF ≥0.55, R2_SNF ≥0.80) ( Figure 1C). Then, the quality-filtered R1 reads were analysed to identify the cell label section sequence (CLS), common sequence Each read with a valid cell label was kept for further consideration only if over 6 out of 8 bases after the UMI were found to be Ts.
Following R1 annotation, Bowtie 2 version 2.2.9 was used to map the filtered R2 reads to the reference panel sequence. The R2 read was considered valid if all the following criteria were matched: the read aligned uniquely to a transcript sequence in the reference; the R2 alignment began within the first five nucleotides; the length of the alignment that could be a match or mismatch in the Compact Idiosyncratic Gapped Alignment Report string was >60; and the read did not align to phiX174. Reads paired with a valid R1 and valid R2 reads were retained for further analyses. Valid R1 reads required identified CLSs, a UMI sequence with non-N bases and poly (T) tail.
Valid R2 reads needed to uniquely map to a gene in panel with the correct PCR2 primer sequence at the start and to have an alignment of over 60 bases in length.
Reads with the same cell label, same UMI sequence and the same gene were collapsed into a single raw read of a molecule. The number of reads related to each raw molecule ID was reported as the raw adjusted sequencing depth. To remove the single-base substitution errors, which were identified and adjusted to the parent UMI barcode using recursive substitution error correction. Other UMI errors derived from library preparation steps or sequencing base deletions were adjusted using distribution-based error correction.
To distinguish cell labels associated with putative cells from those associated with noise, a multistep algorithm was designed for filtering cell labels. The number of reads for each cell was plotted on a log 10 -transformed cumulative curve, with cells sorted in descending order by the number of reads. In a typical experiment, a distinct inflection point was observed, as indicated by the red vertical line.
The algorithm found the minimum derivative along the cumulative reads curve as the inflection point. Cell labels to the left of the minimum second derivative were most likely derived from a cell capture event and were considered to be signals. The remaining cell labels to the right of the minimum second derivative were considered noise.

F I G U R E 1
Single-cell isolation, library preparation and bioinformatic analyses. A, The CD3+ T-cell enrichment by CD3+ T-cell enrichment kit, followed by single-cell isolation and library construction; B, the flow showed reverse transcription and library construction details; and C. bioinformatic analyses procedure

A B C
The R package Seurat was used to analyse the matrix obtained from the BD pipeline, and normalize the data, as well as reduce dimensionality and clustering, and identify differential expression. We used the Seurat alignment method canonical correlation analysis 14 for integrated analysis of datasets. For clustering, highly variable genes were selected, and the principal components based on those genes were used to build a graph, which was segmented with a resolution of 0.6. Based on the filtered gene expression matrix produced by Seurat, sample differential expression analysis was carried out using the edgeR package to obtain zone-specific marker genes. 15,16 In addition, KEGG analysis was performed based on an online database (https://david.ncifc rf.gov/).

| Flow cytometry verification
Briefly, PBMCs were acquired by gradient centrifugation. Then, the cells were incubated with various fluorescein-labelled antigens for surface staining (Table S1). PerCP/Cyanine5.

| Sample characteristics and single T-cell profile generation
We collected PBMCs from two CP/CPPS patients and two healthy controls (Table S2). The flow chart shows the detailed steps of single-cell RNA sequencing, which is described in Figure 1. Then, a CD3+ T-cell enrichment kit was employed to enrich the CD3+ T cells from PBMCs. After testing the viability of the cells with Trypan blue staining (Table S3), the live cells were used to perform library construction and simultaneous mRNA and protein quantification at the single-cell level by the BD™ AbSeq on the Rhapsody™ platform.

| Quality control
Read quality was qualified and filtered based on read length (R1 <66 or R2 <64), mean base quality score (read score <20) and SNF (R1_SNF ≥0.55, R2_SNF ≥0.80). Reads with the same cell label, same UMI sequence and the same gene were collapsed into a single raw read of a molecule. We analysed the correlation between cell and gene/UMI counts in both the healthy control and prostatitis subgroups, and according to which, the cells in low quality (determined by detected UMI or gene counts) were removed from subsequent analyses (Table S4, and Figure S1). In addition, the expression landscape of fifteen proteins among all cells is described in Figure S2.
The number of reads related to each raw molecule ID was reported as the raw adjusted sequencing depth. Single-base substitution errors, which were identified and adjusted to the parent UMI barcode using recursive substitution error correction, were removed. Other UMI errors derived from library preparation steps or sequencing base deletions were later adjusted using distribution-based error correction.
To distinguish cell labels associated with putative cells from those associated with noise, a multistep algorithm was designed for filtering cell labels. The number of reads from each cell was plotted on a log 10 -transformed cumulative curve, with cells sorted in descending order by the number of reads ( Figure S3). The R package Seurat was used to analysing the matrix obtained from BD pipeline and to normalize data; in addition, this package was used for dimensionality reduction, clustering and determination of differential expression. 14 After removing the low-quality cells, the remaining 10 000+ AbSeq profiles (merged from two cases and two controls) were used for subsequent analysis.

| Single-cell RNA-seq analysis reveals distinct subsets of cell
For clustering, highly variable genes were selected, and the principal components based on those genes and proteins were used to build a graph, which was segmented with a resolution of 0.6. Based on the filtered gene expression matrix produced by Seurat, sample differential expression analysis was carried out using the edge T package to obtain zone-specific marker genes and proteins (Table S5).
Further, we compared the gene expression differences between cells derived from prostatitis patients and healthy controls in each cluster and performed pathway enrichment analyses to identify critical pathways ( Figure S5-S7, and Table S6). These results would be helpful for future basic research to annotate the biological function of these critical candidates in regulating specific cell subsets during the pathogenesis of CP/CPPS. In addition, we also compared the DEGs between overall cells derived from CP/CPPS and healthy control subjects and found that these genes were mostly enriched in natural killer cell-mediated cytotoxicity, Th1 and Th2 cell differentiation, hematopoietic cell lineage, allograft rejection, graft-versushost disease, T-cell receptor signalling pathway, etc (Table S7 and  (Table S8 and Figures S8 and S9). Further, a flow cytometry assay is warranted to identify these specific T-cell subsets. 1.861 ± 0.1262 vs 1.263 ± 0.3152, P-value < 0.05; Figure 5).

| Flow cytometry validation
Furthermore, we also compared the variation in Treg cells in PBMCs derived from CP/CPPS patients and healthy controls and found that the proportion of Treg cells in CP/CPPS patients was significantly reduced compared with the proportion in healthy controls (Treg, case vs control: 5.16 ± 0.3192 vs 6.567 ± 0.5186, P-value < 0.05; Figure 5); this result was consistent with the findings of our previous publications. 12 In summary, our results further prove the hypothesis that abnormalities of the immune system may be one of the important factors leading to CP/CPPS.  The mRNA level of the FOXP3 gene in Treg cells was significantly lower in CP/CPPS patients than it was in healthy controls; further, serum TGF-β1 levels, rather than serum TNF-α levels, were up-regulated in CP/CPPS patients compared with healthy controls. 27 Furthermore, Breser et al 28  Although we are just beginning to dissect the significance of central memory T cell, Th1, Th17, Th22, and Treg cells and the vast amount of single-cell data uncovered by our study, we hope that our reported findings will impact the understanding of the mechanisms of CP/CPPS and that further manipulation of these cells will create better methods for restoring homoeostasis.

ACK N OWLED G EM ENTS
We thank the Novogene Co., Ltd. for single-cell multi-omics sequencing and analysis services.

CO N FLI C T O F I NTE R E S T
All authors declared no competing interests. F I G U R E 5 Flow cytometry revealed central memory T cell, Th1, Th17, Th22 and Treg proportions increased in PBMCs derived from CP/CPPS patients than healthy controls. PBMCs were incubated with various fluorescein-labelled antigens for surface staining. PerCP/ Cyanine5.5-conjugated CD3 and FITC-conjugated CD4 were used for the central memory T cell, Th1, Th17 and Th22 cell staining, and PE-conjugated CD25 and FITC-conjugated CD4 were incubated for staining Treg cells. After fixing and permeabilizing with cell fixation/ permeabilization kit, (A) for samples staining central memory T cells were incubated with PE-conjugated CD45RA and APC-conjugated CD62L; (C) for samples staining, Th1 cells were incubated with PE-conjugated IFN-γ; (E) for samples staining, Treg cells were incubated with eFluor 660-conjugated FoxP3; (G) for samples staining, Th17 cells were incubated with PE-conjugated IL-17A; and (I) for samples staining, Th22 cells were incubated with PE-conjugated IL-17A and APC-conjugated IL-22. The quantification data were presented in B, D, F, H and J. *P < 0.05; PBMC, peripheral blood mononuclear cell

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.