Chronic Rejection Triggers the Development of an Aggressive Intragraft Immune Response through Recapitulation of Lymphoid Organogenesis

Microarchitecture of the inflammatory infiltrate during chronic rejection

Over a period of 3 y, 26 explanted nonselected renal grafts were analyzed. The individual characteristics of these patients are presented in Table I. In contrast with kidneys removed due to immediate primary surgical failure (n = 3), or nonimmune failure (angiosarcoma, n = 1; or recurrence of the renal disease, n = 2), all the 20 chronically rejected renal grafts were infiltrated by immune effectors. The possibility that inflammatory infiltration was related to CMV-induced tubulointerstitial nephritis was ruled out for 25/26 samples by the negativity of CMV QRT-PCR (Table I).

In the vast majority of chronically rejected renal grafts (19/20), the interstitial chronic inflammatory infiltrate displayed a highly organized microarchitecture similar to that of secondary lymphoid tissues (Fig. 1A). In particular, B cells were packed into nodular structures evocative of B cell follicles. The phenotypic characterization of B lymphocytes revealed the heterogenous nature of these nodules, as already reported in the synovia of rheumatoid arthritis patients (15). Two types of nodules could be identified; nodules composed of a uniform CD20^pos B cell population expressing IgD and Bcl-2 (Fig. 1B, left panels) were similar to primary follicles; whereas nodules with a core of CD20^posIgD^negBcl-2^neg B cells, highly expressing Bcl-6 that had pushed aside the CD20^posIgD^posBcl-2^pos B cells (Fig. 1B, right panels), resembled secondary follicles, namely, germinal centers. The ratio between these two types of structures differed between samples, and the number of eGCs did not increase with the quantity of primary nodules.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Persistent immune activation triggers the development of heterogeneous nodular B cell structures in inflamed tissues. A, Systematic histological analysis was carried out on sections of kidney grafts explanted because of terminal failure. Representative findings characterizing chronically rejected grafts are shown. H&E staining showed the nodular organization of the inflammatory infiltrate. PNAd and podoplanin immunostaining revealed the high endothelial venule and lymphatic networks, respectively, surrounding the nodules. The spatial organization of cell populations was analyzed by immunohistochemistry with Abs directed to FDC, CD3 (T cells), and DC-LAMP (mature dendritic cells) (original magnification ×200). B, Nodules were composed of a core of CD20^pos cells. Most of the nodules were composed of a homogenous B cell population expressing IgD and Bcl-2 but no Bcl-6 (left column, hereafter referred to as primary follicle-like nodules). In certain nodules (right column, hereafter referred as to eGCs), this B cell population was driven toward the periphery by CD20^pos IgD^neg Bcl-2^neg cells expressing Bcl-6 now occupying the center of the nodule (original magnification ×400). a, the same arteriole on the successive serial sections; FDC, follicular dendritic cells.

eGC formation during chronic inflammation requires the expression of genes involved in lymphoid organogenesis

We reasoned that the development of eGC during chronic inflammation could follow the same biological program as that identified in the embryo during the ontogeny of secondary lymphoid organs. The expression levels of a selected set of genes (Table II), reported to play key roles in lymphoid organogenesis (16–22), were quantified by QRT-PCR in each tissue sample, and in six normal kidneys, all of which were devoid of inflammation. Data were processed by hierarchical clustering to identify samples sharing a similar pattern of expression. Normal kidneys were characterized by a low level of expression for all lymphoid organogenesis genes. Strikingly, this pattern was shared by the three grafts explanted due to primary surgical failure (DTR11, DTR12, and DTR20) as well as by the three grafts explanted due to nonimmune mediated failure (DTR2, DTR10, and DTR16). All these samples were grouped in the C1 cluster (Fig. 2A). In the 20 remaining chronically rejected grafts, the level of expression of lymphoid organogenesis genes was heterogeneous, and samples could be split into three clusters: C2, C3, and C4 (Fig. 2A). In contrast to the samples of the C1 cluster, those of the C2, C3, and C4 clusters all expressed a common core of genes (CXCL13 and CXCR4), the expression of which was associated with the recruitment of B cells organized into primary follicles (Fig. 2B). C3 and C4 additionally expressed CCL19, CCL21, CCR7, LTα, LTβ, and CXCR5. Finally, CXCL12 and LTβR were expressed only in samples from the C4 cluster, this latter therefore grouping together the samples that expressed the complete set of lymphoid organogenesis genes. The expression of these additional genes in the C3 and C4 clusters correlated with the detection of eGC by immunohistochemical analysis (Fig. 2C). The conversion of primary into secondary follicles was more efficient in the C4 samples because the number of eGC was higher in the C4 than in the C3 samples (1 ± 0.71 versus 0.50 ± 0.62; Fig. 2C), despite the fact that they evolved from a reduced quantity of primary nodules (1.00 ± 0.41 versus 1.70 ± 1.03; Fig. 2B).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

Lymphoid neogenesis is a recapitulation of the embryonic lymphoid organogenesis program. A, For each kidney, mRNA was extracted from two distinct tissue fragments and reverse transcribed. For each gene involved in lymphoid organogenesis, real-time QRT-PCR was run five times on each cDNA preparation. Expression levels of these genes were normalized to that of the GAPDH housekeeping gene. For each gene, the 10 values obtained for a single kidney were averaged and the data set was computed to clusterize tissues according to their expression pattern (Ward hierarchical clustering). Individual samples are listed in rows, the genes in columns, and increasing expression levels are encoded from dark blue to bright red. On the top and on the right of the color map, dendrograms list each observation, and indicate the cluster it is in and when it entered its cluster. Four clusters C1–C4 were identified. B and C, The number of primary follicle-like nodules (B) and eGCs (C) was evaluated blindly by a pathologist (N.P.) using a semiquantitative scale ranging from 0 (samples without follicle/eGC) to 3 (samples with >5 follicles/eGCs per field) (original magnification ×50). For each sample, 10 sections from four distinct paraffin-embedded blocks were analyzed. For the analysis, histological quantifications were grouped according to the gene expression-defined clusters (mean ± SD). D, AID protein expression analysis was performed by immunohistochemistry. This staining, performed on the successive serial sections used for the immunohistological study presented in Fig. 1B (a, the same arteriole as in the sections of Fig. 1B), shows that AICD expression is restricted to the germinal center B cell population (CD20^posIgD^negBcl2^neg Bcl6^pos) (original magnification ×400). AID was not expressed by the CD20^posIgD^posBcl2^posBcl6^neg B cells, either in the periphery of eGCs or in primary follicle-like nodules. E, The expression level of AID was analyzed by real-time QRT-PCR with the same methods described in the Fig. 2A legend (mean ± SD). F, The principal component analysis was used to reduce the multidimensional gene expression data set to lower dimensions for analysis, by retaining those characteristics of the data set that contributed most to its variance. The projection of the data on the first and second principal components (canonical 1 and 2) efficiently discriminated the four clusters when individual samples were projected orthogonally. Each circle represents the multivariate mean for each cluster and the size of the circle corresponds to a 95% confidence limit. Nonintersecting circles indicate that clusters are significantly different; however, the difficulty to discriminate between the C2 and C3 clusters suggests that these two clusters are closely related. The composition of the eigenvectors, indicated in the Table below the graph, revealed that the chemokines CCL21 and CXCL13, and the LTβR, were the genes with the largest eigenvalues, namely, the genes that contributed most to dividing the samples into the four clusters. LT, lymphotoxin; NoK, normal kidney.

Germinal center reaction is characterized by the expression of activation-induced cytidine deaminase (AID), the key enzyme controlling class switch recombination and somatic hypermutations. In accordance with recently published data (12, 23), AID protein was not detectable in naive B cells forming the primary follicles, but was readily evidenced in eGC B cells (Fig. 2D). Because this enzyme is mandatory for the production of high-affinity Abs endowed with adequate effector functions, the gradual increase in AID expression observed from C1 to C4 (Student t test = 5.52; p < 0.001) indicates that this gene expression-based clusterization defines functional clusters with respect to the maturation of the local humoral immune response (Fig. 2E).

To identify the genes whose expression contributed the most to distribute the samples into the four clusters, we performed a principal component analysis on the realtime QRT-PCR dataset. This analysis revealed that CXCL13, CCL21, and LTβR were the genes with the largest eigenvalues. Whereas the dendrogram scale, which illustrates the way each sample enters its cluster (Fig. 2A, right dendrogram), suggested a greater level of similarity between C1 and C2 on one hand, and C3 and C4 on the other, the discrimination between samples belonging to the C2 and C3 clusters appeared difficult along the first two eigenvectors (canon 1 and 2; Fig. 2F), indicating that C2 and C3 are highly related.

Lack of completion of the biological program results in impeded local B cell maturation

The maturation of B cells was analyzed by flow cytometry according to the Bm1–Bm5 classification, which is based on the expression of CD38 and IgD (24). This classification allows for the identification of the successive cell development stages, from naive B cells to differentiated memory B cells.

The proportion of CD19^pos cells in the inflammatory infiltrate (Fig. 3A) was concordant with the histological quantification of primary nodules (Fig. 2B) and further documented the paucity of the B cells in the C1 cluster samples.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Incomplete recapitulation of the program impedes local B cell maturation. Tissue samples were processed to isolate the inflammatory infiltrate which was then analyzed by polychromatic flow cytometry. A, The percentage of CD19^pos B cells among lymphocytes (identified on the FSC/SSC morphology gating) is provided for each cluster. Note that the inflammatory infiltrate in the samples from the C1 cluster was minimal and B cells were barely detectable (mean ± SD). B, B cell maturation stages were identified on the basis of the IgD/CD38 staining of CD19^pos cells in the C2 (left), C3 (middle), and C4 (right) clusters (dot plots from three representative samples). The Bm1 and Bm2 naive cells are IgD^posCD38^neg and IgD^posCD38^dim, respectively. The Bm2′ and Bm3δ4δ pregerminal center B cells are IgD^posCD38^bright; the Bm3 and Bm4 germinal center B cells are IgD^negCD38^bright. The early and late Bm5 are, respectively, IgD^negCD38^dim and IgD^negCD38^neg. The gating that was used excluded the CD38^ultrabright from the present analysis. The bubble graphs below the dot plots represents the average percentage of each type of B cell (mean ± SD). C, The percentage of CD27^pos cells among CD19^pos cells was used to quantify memory B cells. In the samples from the C1 cluster, the paucity of the infiltrate made the assessment of memory B cells unreliable. nd, not determined.

Regarding the maturation of B cells, a gradual increase in the proportion of Bm5 memory B cells and a symmetric reduction of Bm1/Bm2 naive B cells was observed from the C2 to C4 samples (Fig. 3B). In the samples of the C4 cluster, in which the complete set of lymphoid organogenesis genes was expressed, the B cell maturation program appeared to be proceeding toward the memory stage unhindered. In contrast, B cell maturation in the C3 cluster was blocked at the Bm2′/Bm3δ4δ pregerminal stage, because a marked reduction in Bm3 and Bm4 germinal center B cells was observed (Fig. 3B). In the C2 cluster B cells maturation was impeded, resulting in naive B cells accumulation (Fig. 3B). Bm5 B cells proportion was only slightly reduced in the C2 cluster, suggesting that a direct recruitment of memory B cells from the periphery is also possible.

The Bm1–Bm5 classification has recently been challenged by the demonstration that some Bm1 cells express the CD27 memory marker (25) therefore indicating that the CD19^posIgD^posCD38^neg population includes both naive Bm1 cells (CD27^neg) and IgD^pos memory B cells (CD27^pos) (26). To exclude the possibility that memory B cells were underestimated when identified by their IgD/CD38 expression, we compared the proportion of CD19^pos CD27^pos cells. We found a gradual increase in CD27^pos B cells, from C2 to C4, further validating the results obtained with the Bm1–Bm5 classification (Fig. 3C).

Finally, kidneys from the C2 cluster can be regarded as inflamed tissues in which the maturation of naive B cells is the less efficient; whereas, the C4 cluster gathers samples in which naive B cells are the most efficiently converted into memory B cells (Student t test = 3.23; p = 0.001). We therefore conclude that the progression of B cells through the successive maturation stages correlates with the gene expression profile of the inflamed tissue.

Clusters correlate with the intensity of the local immune responses

During germinal center reaction, high-affinity B cells are converted into plasmacells, which secrete large amounts of Abs. Plasmacells, easily identifiable by the expression of Syndecan-1 (CD138) and the presence in their cytoplasm of large, eosinophilic, homogenous inclusions, known as Russell bodies, were evidenced in the periphery of eGC (Fig. 4A). In accordance, plasmacells, defined as CD38^ultrabrightCD138^pos, were also detected in the samples from the C3 and C4 clusters (Fig. 4B), namely, the two clusters comprising the samples in which 1) lymphoid organogenesis genes were highly expressed (Fig. 2A), 2) eGCs were detected histologically (Fig. 2C), and 3) the local maturation of B cells was the most efficient (Fig. 3B). This strongly suggests that the gene expression profile of the inflamed tissue determines the progression of B cells through the successive maturation stages. To test this hypothesis, the production of Ig was quantified in tissue-culture conditioned media. Quantification of the various classes and subclasses of Ig produced locally was performed by ELISA in the supernatants after 5 d of tissue culture. IgA and IgM were not detected (data not shown). In contrast, a local production of IgG (Fig. 4C), which increased gradually from the C1 to the C4 samples, was found (Student t test = 2.26; p = 0.024). There was no difference regarding IgG subclass distribution in the samples from the various clusters (Fig. 4D). One limitation of using ELISA to detect IgG on tissue-culture supernatants is the impossibility of discriminating between IgG, which might be trapped within the tissue and subsequently released, from IgG produced by infiltrated plasmacells. Indeed, a small amount of IgG was detected in the C1 cluster samples (Fig. 4C), in which plasmacells were absent (Fig. 4B). We therefore performed an ELISPOT assay on representative samples from the C3 and C4 clusters that demonstrated the presence of IgG-producing cells among the inflammatory infiltrate (Fig. 4E). The increased number of spots observed on mitogen stimulation in the C4, as compared with the C3 cluster (Fig. 4E), indicated that the number of memory B cells was higher in the C4 samples, in accordance with the results of flow cytometric analysis (Fig. 3B, 3C).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

Complete recapitulation of the program leads to development of fully functional eGCs. A, Plasmacells were evidenced in the periphery of eGC by the expression of CD138 (left panel) and the presence in their cytoplasm of large pathognomonic, eosinophilic, homogenous inclusions, known as Russell bodies, on H&E staining (right panel) (original magnification ×200). B, The percentage of CD38^ultrabrightCD138^pos plasmacells infiltrating the graft analyzed by flow cytometry is represented. C and D, For each explanted kidney, tissue cultures of 24 cortical fragments were performed in 24-well plates in 1 ml serum-free RPMI 1640. Supernatants were collected after 5 d. The content of total IgG was measured in duplicate in two randomly selected supernatants by ELISA (C) (mean ± SD). The subclasses of IgG were quantified in duplicate in the same supernatants by ELISA (D). The values are expressed as the percentage of the total amount of IgG. E, Quantification of IgG-producing cells (plasmacells and stimulated memory B cells) was carried out by an ELISPOT assay on two randomly selected samples from the C3 and C4 clusters. Given that CD38^ultrabrightCD138^pos cells were not detected by flow cytometry in the samples from the C1 and C2 clusters, ELISPOT assays were not run on cells from these samples. Representative findings are shown. Cells extracted from explanted kidneys were cultured for 6 d in complete medium with or without a B cell mitogen mix (PWM, SAC, CpG). Cells were plated 6 h in cascade 1/4 dilutions in multiscreen 96-well filter plates coated with anti-human IgG Abs. Spots were revealed with biotinylated goat anti-human IgG Fc, HRP-conjugated Avidin D, and 3-amino-9-ethylcarbazole. Unstimulated wells allow for evaluation of plasmacells; mitogen-treated wells allow for that of memory B cells.

Clusters correlate with the aggressiveness of the local alloimmune response

For 18 of 26 of the explanted kidneys, 24 randomly selected cortical fragments per kidney were tissue cultured and the supernatants were collected after 5 d. The quantification of anti-HLA Abs in these supernatants by luminex assays allowed us to characterize the local alloimmune response as follows: 1) the density of the response was defined as the percentage of wells that were positive for either HLA I or HLA II; 2) the diversity was the ratio of the number of specificities over the number of HLA mismatches for each donor/recipient couple; and 3) the intensity was the highest mean fluorescence intensity (MFI).

As expected, there were no alloantibodies in the tissue supernatants from the C1 cluster. In contrast, a local production of antidonor Abs was detected in the tissues of the three remaining clusters. The density, diversity, and intensity of the local humoral response against the target Ags (Table III) were maximal in the C4 cluster and minimal in the C2 cluster (Student t test = 4.62; p < 0.001). The diversity and intensity of the local alloimmune response in the C2 and C3 clusters were similar. This is consistent with the close relation between of these two clusters, as demonstrated by the principal component analysis (Fig. 2F). These two clusters only differed in density.

Local humoral response is uncoupled from the systemic response

Alloantibodies produced in “canonical” secondary lymphoid organs (i.e., the spleen and the lymph nodes) reach the graft through the circulation. Analysis of the serum therefore offers a unique opportunity to compare the local (i.e., intragraft) and the systemic (i.e., elicited in the secondary lymphoid organs) humoral alloimmune responses. The luminex assays performed on the sera collected at the time of the explantation demonstrated a disconnection between systemic and local humoral responses. Indeed, Ab specificities found in tissue-culture supernatants were almost systematically distinct from those detected in the sera (Fig. 5A, Table III). Also, the local response appeared more diverse, especially in the C4 samples (Fig. 5A, Table III).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

Local and systemic humoral alloimmune responses are uncoupled. A, Anticlass I and anticlass II HLA Abs were identified by the luminex single Ag kit (see Table III legend) in the supernatants of tissue cultures (grey bars) and in the sera obtained immediately before detransplantation (white bars). The black bars represents the specificities detected simultaneously in the serum and the supernatants of tissue cultures. Because this part of the study was initiated after DTR8, data concerning DTR1 to DTR8 are lacking. B, The duration of graft function, set as the number of days between transplantation and the return to dialysis, was plotted for the four clusters of DTRs (mean ± SD).

Functionality of intragraft lymphoid tissue and graft survival

Because we observed that grafts of the C2 cluster had functioned 7.8 y on average and those of the C4 cluster had lasted only half that time (3.5 y; Fig. 5B); whereas we found no statistically significant difference between the clusters for the clinical characteristics that could explain this difference (Table I), we put forward the hypothesis that the functionality of intragraft lymphoid tissue could negatively impact graft survival. A first analysis documented a statistical difference in graft survival between the four clusters (χ² = 30.08; p < 0.0001). However, this difference was likely related to the very short survival of samples from the C1 cluster (three samples of this group were explanted for primary surgical failure). Therefore, in an attempt to more specifically explore the influence of gene expression-based clusterization on chronic rejection kinetic, we performed a subgroup analysis on chronically rejected grafts only (samples from C1 cluster were excluded). Interestingly, this subgroup analysis (χ² = 5.96; p = 0.0507) almost reached the statistical significance despite the very low number of patients of the study.