Background
The introduction of the B-cell depleting agent rituximab constituted a major revolution in the treatment of autoimmune diseases, which renewed interest in the role of B cells in autoimmunity. In Rheumatoid Arthritis (RA), an autoimmune disease that affects the peripheral joints, pathological studies show a prominent role for B cells in at least a subset of patients. A larger influx of B cells in the synovial tissue has been associated with autoantibody positivity, but also with more radiographic disease progression [
1]. This suggests that treatment efficacy might be tightly linked to the contribution of the B-cell compartment in disease pathogenesis.
Although rituximab is clearly effective in certain autoimmune diseases, on the individual level clinical response may vary and is difficult to predict. One of the factors thought to contribute to this variability is the timing and depth of B cell depletion, which has been shown to be very patient specific [
2,
3]. However, the extent of B-cell depletion does not correlate evidently with clinical response [
4]. This might be due to the fact that our tools are relatively insensitive in monitoring B-cell levels in depleted patients, thus detecting repopulation of B cells too late to prevent disease relapse. And in fact, earlier studies confirmed that in rituximab-treated patients high sensitivity techniques are needed to successfully detect B cell signals when conventional flow cytometry fails [
2,
3]. Hence, there is a clear need for a more sensitive, quantitative diagnostic tool that is able to spot B-cell repopulation very early. If indeed this shows a link with disease progression, this might guide clinicians to adapt therapy accordingly.
In a new prospective cohort of RA patients undergoing rituximab therapy, we used adaptive immune receptor repertoire (AIRR) sequencing to analyze clonal dynamics. We confirmed previously reported effects of rituximab treatment on the peripheral blood B-cell receptor (BCR) repertoire. Furthermore, using BCR repertoire analysis we were able to find a link between BCR repopulation and clinical efficacy, thus shedding more light on the mechanism behind rituximab efficacy in RA.
Methods
Patients and samples
Thirty-one patients previously diagnosed with RA according to the 2010 ACR/EULAR criteria [
5], who were about to start with rituximab treatment were included in the ABIRISK consortium multicentric clinical study (NCT02116504) whose primary objective was to assess predictive factors of anti-drug antibodies (ADA) development (
www.abirisk.eu/) [
6]. The treatment protocol consisted of two intravenous injections of 1000 mg rituximab (Roche, Woerden, The Netherlands) 15 days apart. A second cycle of treatment was allowed after 6 months at the treating physician’s discretion (
n = 6). Concomitant medications allowed for RA treatment were Disease-modifying antirheumatic drugs (DMARDs), Non-steroidal anti-inflammatory drugs (NSAID) and corticosteroids. No other biologicals were allowed. Patient visits were at baseline and at one, three, six and twelve months after treatment for sample collection and assessment of disease activity using the Disease Activity Score 28 joints (DAS28) based on CRP, or ESR when CRP was not available. Clinical response was assessed using EULAR response criteria [
7].
Peripheral blood for BCR repertoire analysis was collected using PAXGene Blood RNA tubes (PreAnalytiX, Breda, The Netherlands) and stored at -80°C. Serum for Anti-Drug Antibodies (ADA) testing was collected in BD SST vacutainers, left to coagulate for at least 30 min, centrifuged at 1,500 g for 10 min at 4°C and then stored at -20°C.
The study protocol received ethical approval in all patient recruiting centers and was performed according to the Declaration of Helsinki. All patients gave written informed consent before participation.
Next-generation sequencing of the B-cell receptor repertoire
RNA extraction was performed using PAXgene isolation kit (Qiagen) according to manufacturer’s instructions. Amplification of the B-cell receptor repertoire was performed as previously described and reported in online Supplementary Figure S
1A [
8,
9]. In case no amplification product was obtained, the amplification was repeated with the addition of carrier RNA from the non BCR-expressing cell line HEK939T. This addition did not alter the sample’s BCR repertoire (online Supplementary Figure S
1B-C).
Processing of raw sequences and final dataset construction
Reads were processed using pRESTO [
10]. Low quality reads (phred score ≤ 25) were filtered out. IGHV and IGHJ primer sequences were masked and cut off respectively using the
MaskPrimers.py function, UMI-based consensus sequences created using
BuildConsensus.py (max.error = 0.1) and paired-end reads assembled. Unique UMI-based consensus sequences represented by at least 3 different UMIs were aligned using IMGT/HighV-QUEST [
11]. Functional rearrangements were further processed for germline reconstruction and clonal clustering using Change-O [
12].
B-cell receptor repertoire analysis
B-cell receptor (BCR) clonotypes were defined as unique IGHV-IGHD-IGHJ rearrangement at the nucleotide sequence level . Abundance was defined as the number of different UMIs associated with each clonotype, expressed as the percentage of total number of UMIs in the sample. Clonal expansion was calculated as the Gini index on the distribution of the number of unique UMIs per BCR clonotype in each sample, and Clonal diversity as the Shannon index on the distribution of the number of unique BCR clonotypes per clonal lineage in each sample [
13]. These indices were calculated using the
renyi function in the vegan R package [
14] (version 2.5-6). Analysis of somatic hypermutation was performed using the SHazaM [
12] R package (version 0.2.1).
ADA testing
Serum was tested for presence of ADA against rituximab performing a chemioluminescence drug-tolerant capture ELISA assay using a Meso Scale Diagnostic platform at the clinical immunology laboratory of GlaxoSmithKline Research and Development, Upper Merion, PA, USA.
Dealing with missing data
Three patients were excluded from the final analysis because the baseline PAXGene sample was not collected. For the remaining 28 patients (baseline characteristics in online Supplementary Table S
1), 5 follow-up samples (out of total 112) were not collected and 14 failed BCR amplification or post-sequencing quality control (online Supplementary Table S
2). For analysis of response prediction all patients were included based on the assumption that sample failure represented complete B cell depletion. These samples will be later referred to as
imputed data. This assumption is supported by the fact that available samples taken at the earlier or later timepoint indeed did show extensive BCR depletion (see Fig.
2B). No change in results was observed if these patients were, or were not included (Figs.
3 and
4 and online Supplementary Figure S
3). In case of missing DAS28-score (5% of the data), the timepoints concerned were excluded from the analysis. For the analysis of clonality, somatic hypermutation, BCR depletion and BCR repopulation after rituximab, patients were included if they had no (
n = 18) or one (
n = 5) missing follow-up timepoint (total
n = 23).
Statistics
Data are presented as mean and standard deviation (SD) or median and interquartile range (IQR) after D’Agostino and Pearson omnibus test for normality. Differences between groups were evaluated using unpaired t-test and one-way analysis of variance (ANOVA) followed by Bonferroni’s multiple comparisons post-test for normally distributed data or Mann-Whitney test and Kruskal-Wallis test followed by Dunn’s multiple comparisons post-test for not-normally distributed data. Contingency tables were evaluated with Fisher's exact test. P-values less than 0.05 were considered statistically significant. Graphpad Prism 7 software (Graph Pad, San Diego, CA, USA) was used to perform the statistical tests.
Discussion
Using UMI-based adaptive immune receptor repertoire (AIRR) sequencing in RA patients undergoing rituximab treatment we show that deletion and recurrence of the unmutated BCRs, associated with the naive repertoire, proved a sensitive marker for depletion and repopulation. Using this parameter, we find a correlation between BCR repopulation and clinical improvement shortly after, strongly suggesting that it might be the repopulation, rather than the depletion, that has a dampening role on disease activity.
As described earlier, treatment with rituximab leads to fewer, more expanded and more mutated BCR clonotypes [
3]. This is not surprising since the residual B cell population is most likely composed of plasmablasts and plasma cells which do not express CD20, the molecular target of rituximab. Therefore, these cells are not affected by this B cell depleting therapy. In other studies, phenotypic characterization of the post-depletion B cell compartment using highly sensitive flow cytometry showed the presence of a residual memory B cell population which was also the predominant re-populating fraction in patients with early relapse after B cell recovery [
2,
15,
16]. Even though our RNA-based analysis is dominated by the BCR signal coming from plasmablasts/plasma cells due to their higher RNA content (data not shown from our lab), we cannot exclude that some of the residual dominant BCR clones observed in our analysis still represent memory B cells.
In this study we monitored unmutated clonotypes in the BCR repertoire with quantitative UMI-based sequencing to sensitively follow the dynamics of B cell depletion and repopulation in rituximab treated patients. Using this method, we observed that repopulation of the BCR repertoire within 6 months of treatment did correlate with significantly better disease amelioration in the subsequent period, between month 6 and 12 post treatment, compared to patients that did achieve repopulation later or not at all (p-value < 0.01). This observation was not explained by a difference in disease activity at 6 months between the two groups (p-value = 0.92) nor by a 2nd cycle of treatment. This might indicate that it is in fact the repopulation following rituximab – rather than depletion itself - that is able to “re-set" the (pathological) B cell compartment, leading to temporal improvement of the disease activity. In this context, the ability to promptly recognize the start of B cell repopulation after depletion might thus be extremely useful in clinical practice, allowing the physicians to more closely monitor disease progression during repopulation and eventually take an informed decision on intensification of treatment.
Using the same approach, we observed that BCR
depletion does not predict clinical response at 6 and 12 months. On top of that, we showed that patients that achieve BCR depletion within one month of treatment did not show any improvement in disease activity in the same time-span. This confirms previous reports showing that timing and depth of B cell depletion does not correlate with clinical response to rituximab during 0.5 and 1 year of follow-up [
4,
20]. This lack of correlation is thus not due to a lack of sensitivity of previously used methods. The observed association of early BCR repertoire depletion with
less response at 1 and 3 months after rituximab therapy is interesting. This might be the result of the fact that in our cohort late depleting patients had a slightly higher baseline DAS28 score (late depleting: 4.6 ± 0.9 vs. early depleting: 4.1 ± 1.1;
p-value = 0.2, n.s.), thus resulting in more potential to improve in disease activity [
20,
21]. However, if indeed an influx of naïve, unmutated B cells explains the clinical effect of rituximab, an alternative explanation might be that part of the patients have a higher baseline influx of naïve, unmutated B-cells before treatment; this increase in B-cell turnover might explain both later BCR depletion with better amelioration of disease activity, and earlier BCR repopulation with better disease amelioration. In case of rituximab, this would lead to a complicated relation between anti-drug antibodies (ADA) development and clinical response. In patients newly starting on rituximab, some will develop ADA after 3 to 6 months [
6]. In these patients, ADA is likely to be correlated with earlier clearance of the drug and earlier repopulation (
p = 0.06 in our data) leading to better clinical response. However, in patients with pre-existing ADA, these antibodies will interfere with the primary effect of rituximab, i.e. B-cell depletion, and therefore lead to clinical non-response. In this situation the net effect of both mechanisms will determine the clinical response in each individual patient.
In this study we showed how our BCR-based approach could be adopted by clinician to monitor the dynamics of BCR-depletion and -repopulation during treatment with rituximab. Yet, there are several points which still need to be addressed before our approach can reach clinical practice. In first instance, this study needs be repeated in additional cohorts. This would also allow to define a more robust cut-off for BCR-depletion and BCR-repopulation. Secondly, this approach should be compared with other routinely used methodologies for the sensitive quantification of B cells during B cell depletion therapy, such as high-sensitivity flow cytometry and DNA-based NGS sequencing of immunoglobulin gene rearrangements [
2,
22]. In this respect, the use of a DNA-based rather than RNA-based approach would allow for a more precise quantification of the number of B cells rather than the number of B-cell receptor sequences without the need for a flow-based quantification. Yet the difficulty in incorporating unique molecular identifiers (UMI) in this approach makes it difficult to correct for primer-based amplification bias. Finally, to get more insight in the underlying biology, it would be interesting to further explore the phenotypic diversity of B cells carrying unmutated BCRs and the impact of these B cells on the reported correlations with clinical response.
The protocol adopted in this study has several limitations. One is the relatively short follow-up time. Since most of the patients show repopulation of the BCR repertoire at 6 months post-treatment, having the last follow-up point set at 12 months post treatment was relatively short to detect disease relapse. The second limitation is that the protocol did allow intensification of treatment in cases of insufficient response. However, clinicians did not have access to the results of the BCR repopulation analysis, and the decision to retreat with rituximab after 6 months did not correlate with early repopulation (p = 0.7). It would have been nice to correlate the BCR repertoire in paired synovial tissue samples with that in peripheral blood to study the recurrence of B cell clonotypes during disease relapse. Such a study could prove that rituximab does not eradicate pathological B cells but just prevents them to reach the site of disease activity, i.e., the synovium, therefore explaining why - despite the temporary amelioration of disease symptoms - CD20-depleting therapy does not cure RA.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.