Late B lymphocyte action in dysfunctional tissue repair following kidney injury and transplantation

The mechanisms initiating the late immune response to allografts are poorly understood. Through transcriptome analysis of serial protocol biopsies in kidney transplant recipients, we found a tight correlation between the initial response to kidney injury and a late B lymphocyte signature associated with renal dysfunction and fibrosis, suggesting a link between dysfunctional repair and immunoreactivity. To specifically investigate the immunological consequences of dysfunctional repair, we followed the mouse kidney up to 18 months after ischemia/reperfusion. Even in the absence of foreign antigens we identified a sustained immune response in conjunction with the transition to chronic kidney damage. This tissue-driven immunological process involved both the innate and the adaptive immune system and eventually induced an antigen-driven proliferation, selection and maturation of B lymphocytes into broadly-reacting antibody secreting cells. These findings reveal an unappreciated role of dysfunctional tissue repair on local immunoregulation with a particular relevance for late transplantation immunobiology.


Introduction
The immune system participates in tissue repair with contrasting effects: inflammation after injury is important to initiate the repair response but immune cells can contribute to secondary tissue damage. [1][2][3][4][5] Organ transplantation is an interesting model to investigate the interaction between tissue injury and local immune regulation. Ischemia/reperfusion injury inevitably occurs during organ transplantation and triggers the coordinated activation of the innate and the adaptive immune system of the host in a complex immunological process leading to acute allograft rejection. [6][7][8] This process has been extensively investigated in experimental models and can be effectively prevented and treated with currently available immunosuppressive drugs. 9 The early and late immune responses to allografts are distinct processes: chronic forms of rejection remain mechanistically poorly understood and are not effectively treatable. [10][11][12] As a result, long-term outcome after kidney transplantation has not substantially improved over two decades: approximately 4-5% of renal grafts are lost annually beyond the first year after transplantation, mainly because of late forms of immune-mediated injury (chronic rejection). [12][13][14] Improved pathologic and immunologic diagnostic tools (e.g. C4d stain and detection of anti-HLA antibodies) indicate a critical role for B lymphocytes and donor specific antibodies in the late immune response to allografts. 15,16 However, at a time when the criteria for chronic antibody-mediated rejection are met, even aggressive immunosuppressive therapy does not substantially improve graft survival. 10,17 The mechanisms initiating a donor-specific immune response, several months/years after transplantation and often without any clear precipitating event, are unclear and clinically important. In this study, we took advantage of recent technical advances in the characterization of the adaptive immune system and molecular processes determining the transition from acute to chronic kidney injury to investigate the impact of dysfunctional kidney repair on the late immune response following kidney injury and kidney transplantation.

Chronic kidney injury and late B lymphocyte activity correlate after kidney transplantation
Transcriptional profiling of protocol biopsies from 42 kidney allografts in the first year after transplantation enabled an examination of changing gene activity in the post-transplant kidney. We identified a cluster of strongly correlated genes associated with fibrosis (e.g. COL1A1, DPT, MMP7), immunity (e.g. CD52, CXCL10, CCL21) and B cell action (immunoglobulin genes) (Fig. 1a). In agreement with previous reports, B cell associated transcripts and immunoglobulin genes increased over time after transplantation ( Fig. 1b-d). 18,19 The sample cluster analysis based on B cell-associated genes expressed in tissue biopsies collected at 12 months following transplantation indicated that the B cell signature was restricted to a subset of patients (Fig.   1e). Independently, t-distributed stochastic neighbor embedding (t-SNE) analysis identified a group of patients separating from the remaining study population by virtue of the higher expression of genes associated with fibrosis, wound healing and immunity ( Fig. 1f-h, Suppl. Table 1). Since fibrosis and chronic inflammation are universal features of the maladaptive repair processes leading to chronic organ damage, 1,20,21 we defined this group of patients as a maladaptive injury repair group (MIR-group). The MIR-group substantially overlapped with the subset of patients classified in the cluster analysis based on B cell-associated genes (odds ratio for the overlap 45.5, P<0.0001, Fig. 1e). Thus, the fibrosis and B cell signature show a strong concordance in kidney tissue 12 months after transplantation. These findings suggest two non-exclusive scenarios: (1) the host immune response to the allograft causes kidney damage, (2) maladaptive kidney repair triggers a late B cell response targeting the injured tissue.

Kidney injury precedes B cell mediated immunity in human allografts
To find evidence to distinguish between these alternative possibilities, we examined earlier biopsies from the patient cohort to examine the transcriptional response in the same kidney over time for factors that might precede the development of chronic kidney injury and late B cell activity signatures. We compared the MIRgroup (as defined at 12 months after transplantation, Fig. 1f) with the rest of the study population referred to as the "non-MIR" group . At baseline, clinical characteristics of the patients and transcriptional profiles of MIR and non-MIR kidney biopsies were indistinguishable (Suppl. Table 2). Three months post-transplant, renal function was similar between the two groups. No histological evidence was found for lymphocytic infiltrates compatible with rejection (similar t, i and v scores according to Banff classification) 22 or chronic tissue damage (e.g. fibrosis as determined by the ci score) in the MIR group ( Fig. 2a-b). However, MIR group patients expressed higher levels of well-characterized markers of acute kidney injury and repair (e.g. LCN2, SOX9, ALDH2A1; Fig. 2c), 23,24 suggesting a more pronounced or sustained response to tissue injury.
To find evidence for subclinical rejection episodes as a potential cause of tissue injury in the MIR-group, we used an extensive list of genes associated with acute rejection that have been reported to detect rejection with greater sensitivity than conventional histology. 25 Among the 186 analyzed genes, only 10 were detectably upregulated in the MIR group (Suppl. Table 3). These genes relate to innate immunity (Fig. 2d), whereas genes more specifically linked to rejection and adaptive immunity were similarly expressed in the two groups. In contrast, among a set of 29 genes previously associated with acute kidney injury in kidney transplant recipients, 26 19 (65%) were significantly up-regulated in the MIR group and only 1 in the rest of the study population (Suppl . Table 4). Thus, kidneys exhibiting a signature of fibrosis and B cell activity at 12 months showed evidence for kidney injury, but not allograft rejection at 3 months.
One year after transplantation, patients in the MIR-group displayed a lower renal function and higher scores related to chronic kidney damage, consistent with transcriptional profiling (ci and ct score, Figs. 2e-f). In contrast, no histological evidence was observed for B cell mediated immunity as a cause of chronic kidney damage (absence of peritubular capillaritis and C4d staining, Fig. 2g). 27 Histological analysis at 12 months revealed no evidence of increased inflammation in the MIR-group, pointing to a higher sensitivity for transcriptome analysis. In summary, a transcriptional signature of acute kidney injury at 3 months was associated with chronic organ damage and dysfunction at 12 months. The clinical data suggested that fibrosis in MIR patients did not result from a pathological immune response to the antigenically distinct allograft but support an alternative hypothesis in which the late immune response reflects a dysfunctional immune response related to ongoing tissue injury.

Ectopic lymphoid tissue following renal ischemia-reperfusion injury in mice
To explore this hypothesis beyond the descriptive association provided by the clinical data, we turned to a mouse model where ischemia/reperfusion injury (IRI) led to a dysfunctional repair process that transitioned to chronic kidney disease over months, to specifically characterize the long-term immunological sequelae of maladaptive kidney repair. 28 Sixteen months after a single IRI, the tissue architecture of the kidney was markedly distorted with flattened epithelial cells, cyst formation, extensive fibrosis and inflammation, reflected in a markedly reduced glomerular filtration rate (Fig. 3a-e). 28 CIBERSORT analysis on RNAseq data over time indicated that lymphocytes were the most abundant immune cells is the kidney beyond the sixth month after injury while myeloid cell expansion characterized the acute phase (Fig. 3f). 29,30 Beyond the sixth month after injury, lymphocytes were mostly organized in large cellular clusters, primarily around small arteries, between renal tubules identified by Havcr1 and Krt20 activity, markers of unresolved tubular injury (Fig. 3g, Suppl. Fig.   1). 28 These highly vascularized ectopic lymphoid structures were populated by CD19 + B and CD3 + T cells (mainly CD4 + ) and surrounded by Lyve1 + lymphatic vessels.
Examining the cellular compartmentalization of larger cellular aggregates identified a B cell zone with germinal centers of CD19 + /CD45R + or dim B cells embedded in a network of CD21 + /Cxcl13 + follicular dendritic cells partially separated from clusters of highly proliferating Ki67 + lymphocytes, as typically observed in mature germinal centers (Fig. 3h-i). 31,32 Similar ectopic lymphoid structures were described in different clinical conditions in the kidney 33 and often develop in target tissues of autoimmune disease through the interaction of infiltrating immune cells and local stroma, directed by chemokine feedback loops. 34,35 A systematic analysis of the cytokines transcribed in the injured kidney over time highlighted the progression from acute to chronic inflammation, with the late detection of cytokines involved in the formation of ectopic lymphoid tissue (e.g. Ccl21, Cxcl12, Cxcl13; Fig. 3j). 34 At 28 days post IRI, the acute and the chronic inflammatory responses overlapped. To assess the contribution of different cellular compartments of the kidney at this critical transition point we performed RNA sequencing after Translating Ribosomal Affinity Purification (TRAP) to examine the profile of myeloid and renal stromal cell compartments (Fig. 3k). 36 Both myeloid and stroma cells showed marked translation of cytokine encoding mRNAs. However, the stroma cell compartment, which encompasses the bulk of cells directly contributing to renal fibrosis, 37 expressed a cluster of cytokines showing elevated expression sixth months after IRI, including cytokines associated with lymphocyte homing (e.g. Ccl28, Ccl21a and Cxcl12, Fig. 3j-k). Thus, adaptive immunity is an intrinsic component of dysfunctional kidney repair in the mouse kidney transitioning from acute to chronic injury in the absence of foreign antigens.

B lymphocyte accumulation in the mouse kidney transitioning to chronic injury
The immune response detected in the kidney beyond the sixth month after IRI involved both T and B cells.
The T cell compartment included a large fraction of non-conventional TCRab + CD4 -CD8 -(double negative) T cells (Suppl. Fig. 2). Double negative T cells were predominant also among the lymphocytes isolated from the kidney of aged control mice, as shown in previous reports suggesting double negative T cells as a population of kidney resident T cells. 38,39 Consistent with previous studies, 5,40 B lymphocytes were rare in the normal kidney and memory B lymphocytes infiltrated the kidney in the first days after IRI; during the following weeks, B lymphocytes expanded massively and transcriptional analysis suggested a progressive switch to a plasma cell type ( Fig. 4a-b). The direct phenotypic characterization of the infiltrating lymphocytes by flow cytometry after enzymatic tissue dissociation and magnetic cell sorting for CD45 42 16-18 months after IRI confirmed the presence of differentiated B lymphocytes, but showed that the renal B lymphocytes lacked CD138 (syndecan-1, a typical plasma cell marker, Fig. 4c-d, Suppl. Fig. 3). Instead of fully differentiated plasma cells, we consistently identified a population of CD19 +/low CD45R -B lymphocytes, that displayed CD126 (interleukin 6 receptor) and Cxcr4 (CD184) (Fig. 4e-f, Suppl. Fig. 3), the receptor for Cxcl12, a cytokine highly expressed by the renal stroma in the transition to chronic injury ( Fig. 3j-k).
B lymphocytes isolated from the kidney in the late phase after IRI were reminiscent of plasma cell precursors involved in the renal production of autoantibodies in experimental lupus models and described in patients with systemic lupus erythematosus. [43][44][45] The marked increase in immunoglobulin gene transcripts for IgM, IgG2c and IgA classes and Ig kappa light chains in whole organ RNA-profiles was consistent with the local production of antibodies in the kidney (Fig. 4g-h). Despite the high level of IgM transcripts and the recognized role of natural antibodies in the tissue injury response, 46 B-1 cells (and marginal zone B cells) were not detected in the kidney (Suppl. Fig. 4). Thus, antibody secreting cells accumulated in the kidney in conjunction with the transition from acute to chronic kidney injury.

B cell clonal expansion and affinity maturation in the injured mouse kidney
To better characterize the accumulation and the differentiation of B lymphocytes in the damaged mouse kidney we performed a B cell receptor (BCR) repertoire analysis in mice. First, we investigated the germline antibody gene segment repertoire by quantifying Ighv and Igkv transcripts in the kidney by whole organ RNA-seq. At 28 days after IRI we observed the relative enrichment of a limited number of Ighv genes, reflecting the presence of a small number of B cell. In the following months, in parallel with the massive increase of immunoglobulin transcripts, we detected the relative enrichment of V segments. The 3 most prevalent V segments accounted for 25-40% of the total Ighv transcripts and the top one for 18-35% of the Igkv transcripts at 12 months ( Fig.   5a-d). This process could only be appreciated beyond the sixth month after IRI, underlining the need for a very long-term follow-up, and suggested the expansion of a subset of B cell clones in the kidney.
The most frequent segments in both Ighv and Ihkv were present at high frequency in the 3 mice analyzed at 12 months, but were not particularly abundant in sham control mice and were not among the most frequently used V segments in mice. 47 This excludes an effect related to aging and suggest a biased usage of immunoglobulin genes. In a second step, the BCR repertoire analysis at 12 months after IRI was refined at the DNA level by Immunoseq ® , which better assays the composition of the B cell repertoire providing additional information on the single VDJ rearrangements. 48 This analysis confirmed the presence of a polyclonal B cell population in the kidney with the enrichment of a limited number of dominant clones. The amino acid sequences of the complementarity-determining region 3 (CDR3) of the immunoglobulin heavy chain in control mice did not show any evidence for a phylogenetic organization of the rearrangements (Fig. 5e). In contrast, IRI survivors presented clusters of hypermutated immunoglobulin gene rearrangements with few amino acid substitutions indicating a process of clonal expansion and affinity maturation (Fig. 5f). Thus, both histological findings and BCR analysis, provide evidence for the proliferation, selection and maturation of B lymphocytes in germinal centers within the kidney, in conjunction with the transition from acute to chronic kidney injury.

Detection of autoantibodies after ischemia reperfusion injury
The clonal expansion of B lymphocytes in germinal centers is an antigen-triggered immunological process. 32 In the absence of foreign antigens in this aseptic kidney injury model, we tested the plasma collected [16][17][18] months after IRI on a screening panel for autoantibodies. Thirty five out of 118 tested antigens showed a statistically significant increased signal to noise ratio (SNR) 16 months after IRI compared to controls (adjusted P-value<0.05, FDR 5%; Fig. 5g) without evidence for a common target antigen. This indicated that in the absence of foreign antigens, the intrarenal B cell response resulted in the production of broadly reactive autoantibodies. Since a similar process after kidney transplantation might stimulate the production of donorspecific antibodies, we retrospectively analyzed the clinical record of our study population to correlate the detection of anti-HLA antibodies with the initial injury response: among kidney recipients not HLA-sensitized at time point of transplantation. Indeed, 3 patients developed clinical evidence for late B cell mediated reactivity to the graft during further follow-up. All these patients were classified in the MIR group (Suppl. Table 3).

Discussion
B lymphocytes are involved in the acute response to ischemic injury but are not essential for acute rejection, which is mainly mediated by T lymphocytes. [3][4][5]49 In contrast, B cells are pivotal in the pathogenesis of late forms of immune-mediated graft injury. Here we show that late B cell activity in kidney allografts is tightly linked to dysfunctional kidney repair. Since in the clinical setting, this finding was associative, we turned to a mouse model to further characterize the immunological processes during the transition from acute to chronic kidney injury. A dysfunctional repair program following severe IRI transformed the mouse kidney into a tertiary lymphatic organ, which did not only host the inflammatory response typically associated to chronic organ injury, 50 but also induced an antigen-driven expansion and maturation of B cell clones. In the absence of foreign antigens, this process resulted in the production of broadly reactive autoantibodies, as previously reported in patients after heart and brain injury. [51][52][53] These findings are reminiscent of a recent characterization of the clonal evolution of autoreactive germinal centers in lupus. Once self-tolerance is broken in lupus, autoreactive B cells drive the expansion of further clones in the context of a self-sustained inflammatory feedback mechanism eventually leading to a convergent autoimmune response, with a delay of several months after the initial break of tolerance. 31 Our findings support a model wherein dysfunctional tissue repair is sufficient to initiate a similar process with clinical relevance to organ transplantation.
Taken together, our data suggest a new model for the pathogenesis of late immune responses to allografts: a maladaptive repair program following unresolved tissue injury stimulates a chronic activation of the adaptive immune system, including B lymphocytes. B lymphocytes interact with the surrounding renal stroma to create an environment supporting the recruitment of additional clones and stimulating maturation and differentiation of antibody producing cells. Over time this process triggers the production of antibodies that promote further tissue damage in a deleterious feedback mechanism (Fig. 6).
Although the numbers of patients was not sufficient to demonstrate a causal link to donor-specific immunoreactivity and to formally exclude potential confounders, the model is supported by earlier findings. Several studies characterized the broad antibody responses in conjunction with chronic rejection after kidney transplantation 54 and the acceleration of chronic rejection by intragraft lymphoid neogenesis. 55 Moreover, previous evidence for a stepwise breakdown of B cell tolerance as a preamble to chronic rejection directly parallels our observations in the mouse model. 56 Also consistent with this model are reports showing that among the anti-HLA antibodies detected 5 years following transplantation, only one third were donor-specific, and non-donor specific antibodies appeared earlier than donor-specific, but both donor-specific antibodies and non-donor-specific antibodies were associated with an adverse outcome. 57,58 Conversely, the identification of dysfunctional tissue repair as the initiator not only for fibrosis as previously proposed, 19 but also for late forms of alloreactivity provides a plausible explanation for much lower incidence of chronic antibody-mediated rejection in an organ with a high regenerative capacity, such as the liver. 11 Our model argues for a change in perspective in relation to the unresolved issue of chronic allograft dysfunction and B cell mediated immunity. In this, the production of donor-specific antibodies is a late event reflecting irreversible organ damage. Consequently, the lack of therapeutic efficacy against chronic antibody mediated rejection (including potent B cell depleting agents) is not surprising. 17 However, our data -in line with previous clinical studies -58 indicate that the progression towards chronic damage and B cell activation is predictable, thereby offering an opportunity for early diagnostic and therapeutic intervention. The cause of injury in the clinical setting is much more variable than in the experimental model, several factors might determine if the repair response will result in a sustained immune activation, e.g. recent studies in mice suggest that age might play an important role in this regard. 59 Developing strategies to attenuate the early injury response and repetitive hits during follow-up may prevent irreversible organ damage and late alloreactivity. 60 The process

Competing interests
None.

Study design, patients and clinical data
We

Tissue storage and RNA extraction
Of each renal allograft biopsy included in this study, at least half a core was immediately stored on Allprotect anesthetized with an intraperitoneal injection of a ketamine/xylazine (105 mg ketamine/kg; 10 mg xylazine/kg).
Body temperature was maintained at 36.5-37°C throughout the procedure. The kidneys were exposed by a

Cell isolation, sorting and flow cytometry
Lymphocyte were isolated from the kidney as previously described. 5 Briefly, the mouse was perfused with PBS until the kidney was visually blood-free. The kidney was removed, mechanically dissociated and then The sequencing data were aligned to mm10 genome assembly with STAR aligner (version 2.5.0b). The mapping index was generated using GENCODE release M4 (GRCm38.p3) gene annotations. In addition to read mapping, STAR was also used to remove duplicates, generate read-count tables and wiggle files.
Differentially-expressed genes were called for each time point with DESeq2. Cluster analysis of cytokine transcripts was performed with Gene Cluster (version 3.0) and visualized with Treeview (version 1.1.6r4).

Autoantibodies
Autoantibody reactivity against a panel of autoantigens was determined in mouse plasma samples at the UT Southwestern Medical Center. The samples were incubated with autoantigen array and the autoantibodies binding to the antigens on the array was be detected with Cy3 labeled anti-IgG and Cy5 labeled anti-IgM to generate Tiff images. Genepix Pro 6.0 software was used to analyze the image. Signal-to-noise ratio (SNR) was used as a quantitative measure of the ability to resolve true signal from background noise. A higher SNR indicates higher signal over background noise. SNR equal or bigger than 3 was considered true signal from background noise. Quality control and normalization using variance stabilizing normalization were performed according to the standard procedure at UT Southwestern Medical center.

Statistical analyses
RPKM and eGFR values were compared by t-test or by Mann-Whitney U tests, a two-tailed P value <0.05 was considered significant. Categorical analyses were performed with Fisher's exact or Chi-square tests.
Differential gene expression analyses were performed with EdgeR, 6

CIBERSORT and Immunoseq
CIBERSORT analyses were performed on RNAseq data by using the analytical tool developed by Newman and al. (cibersort.stanford.edu). 7 LM22 was used as a reference signature for analysis on human data. The mouse specific reference profile developed by Chen et al. was applied for studies in mice. 8 For Immunoseq® analysis mouse renal tissue was analyzed by Adaptive biotechnologies (Seattle, USA). 9 The data were analyzed with the immunoSEQ Analyzer (https://www.adaptivebiotech.com/immunoseq).

Software
Gene correlation analysis and feature correlation heatmaps were performed on the PIVOT platform, developed by the Kim Lab (http://kim.bio.upenn.edu/software/pivot.shtml). 10