Abstract
Objective. Transient B cell depletion by rituximab (RTX) has become a specific treatment of rheumatoid arthritis (RA). Although phenotypic repopulation kinetics of B cell subsets are well documented, precise molecular analyses of the reconstituting immunoglobulin (Ig) genes encoding the B cell receptor in RA are sparse.
Methods. A total of 708 individual CD19+CD27+ (memory) and CD19+CD27– (naive) B cells from 2 patients with RA were analyzed at baseline and 7 months after RTX at B cell repopulation. Ig light chain variable kappa (Vκ) and lambda (Vλ) light chain gene rearrangements were amplified, sequenced, and analyzed with a focus on receptor revision.
Results. The naive as well as the memory repertoire repopulated polyclonally with diverse use of variable light chain gene families and minigenes. During the reconstitution phase, B cells used significantly fewer Jκ distal Vκ genes (p = 0.0006), with a higher frequency of somatic hypermutation of rearrangements employing Jκ5 compared to baseline in memory B cells. The use of Vλ rearrangements in regenerating B cells was also biased toward use of Vλ genes of the proximal cassette. In general, reemerging CD27+ Ig light chain genes were substantially more highly mutated than before RTX therapy (p < 0.0001, baseline vs during reconstitution).
Conclusion. Our data indicate that RTX therapy leads to generation of distinct Vκ/Jκ and Vλ/Jλ gene repertoires consistent with replenishment of antigen-experienced B cells by germinal centers. At baseline, the imprints of receptor revision appeared to be more striking, which indicates that receptor revision is active in patients with RA and can be reduced by RTX.
- RHEUMATOID ARTHRITIS
- IMMUNOGLOBULIN LIGHT CHAINS
- RECEPTOR REVISION
Rheumatoid arthritis (RA) is an immune-mediated systemic inflammatory disorder manifested by aggressive synovitis that leads to progressive cartilage degradation and bone erosion, which causes pain, joint deformity, and functional disability1,2. Whereas different cell types play pathogenic roles in RA, B cells have drawn particular attention3. B cells do not merely secrete autoantibodies including rheumatoid factor (RF) and antibodies against citrullinated peptides, but are also involved in secretion of proinflammatory cytokines, activating autoreactive T cells, and presenting and processing autoantigens, leading to infiltration of end organs4. B cells are able to rearrange specific antigenic receptors by receptor-editing and lack of central and/or peripheral tolerance contributes to their abnormal activation and proliferation in autoimmune diseases5,6. The importance of B cells in the pathophysiology of RA has been emphasized by therapeutic approaches using the anti-CD20 monoclonal antibody rituximab (RTX). RTX results in prolonged alleviation of clinical symptoms and inflammation and reduced erosive joint destruction7. We and other investigators have reported regeneration kinetics from B cell depletion studies indicating that peripheral B cells repopulate between 6 months and 10 months after depletion with a majority of newly emerging immature B cells8,9.
The B cell repertoire is shaped by a complex set of genes that are modulated through different rearrangements, somatic hypermutation, and receptor revision. All these processes are involved in producing unique and diverse immunoglobulin (Ig) receptors and are highly regulated during ontogeny, development, and response to antigen. Selection and activation of autoreactive B cells in healthy individuals is avoided by a tolerance mechanism and the vast majority of naive autoreactive B cells are deleted through apoptosis in their premature stages10,11.
B cell antigen receptor revision involves secondary Ig gene rearrangement. Ig-transgenic mouse model studies identified receptor revision events as a relatively rare process capable of rescuing a small fraction of autoreactive B cells12. The IgV repertoire is shaped by a variety of selective and molecular events. Previous studies indicated the existence of disrupted receptor revision in RA13. However, no data are available analyzing the light chain repertoire under RTX, which might follow different expression profiles and could be exposed to distinct selection pressures.
We analyzed the molecular aspects of B cell receptor after transient B cell depletion in patients with RA, with a focus on modulation of B cell immunoglobulin Vκ and Vλ light chain gene rearrangements in CD27– naive and CD27+ memory B cells. In particular, we analyzed the mutational pattern and Ig gene repertoire before B cell depletion and during the B cell regeneration phase (7 months) after a single dose of RTX.
MATERIALS AND METHODS
Patients
Informed consent was obtained from the patients according to a protocol approved by the ethics committee of the University of Würzburg and according to the Declaration of Helsinki. Patient 1 (female, 47 years old) and Patient 2 (female, 35 years old) had active RA.
Both patients were RF-positive and were diagnosed according to the American College of Rheumatology revised criteria for RA14. They were nonresponders to standard disease-modifying antirheumatic drugs. Patients were receiving a stable dose of methotrexate before receiving 2 treatments of 1000 mg RTX, 2 weeks apart. Disease activity was determined by measuring the 28-joint Disease Activity Score, erythrocyte sedimentation rate, and C-reactive protein levels. Further disease variables are shown in Table 1. Both patients showed a clinical response during the study. Patient 2 experienced increasing disease activity at the second timepoint of analysis.
Single cell sorting
For separation of peripheral mononuclear cells, a blood sample suspension was centrifuged on a Ficoll-paque gradient (GE Healthcare, Munich, Germany) with subsequent washing steps in phosphate buffered saline (PBS). A single B cell sorting technique was performed as described15,16. In brief, B cells were stained with fluorescence-labeled monoclonal antibodies against B cell surface markers: CD19-APC and CD27-PE. Cell incubation with antibodies was performed at 4°C for 15 min in PBS/0.5% bovine serum albumin (BSA). Individual B cells were sorted into 96-well plates containing lysis buffer using a FACS DiVA process (Beckton Dickinson). Lysis buffer consisted of DTT (Roche Diagnostics, Mannheim, Germany), RNasin (Promega), Oligo dT(15) primer (Promega), BSA (Sigma-Aldrich), Triton 10% (EMD), and molecular grade H2O.
cDNA synthesis
A mixture containing nucleotides, reverse transcriptase, and RT-polymerase chain reaction buffer from the Titan One Tube RT-PCR System (Roche) was added into each well of a 96-well plate containing sorted B cells. cDNA was synthesized by incubation at 50°C for 1 h.
Nested polymerase chain reaction (PCR)
Vκ (families 1–6) and Vλ (families 1–10) were amplified by nested PCR using family-specific primers. Reaction mixture, primers, and cycling conditions were as described17,18. PCR products were separated by electrophoresis on a 1.5% agarose gel. The products were purified further with a MinElute Gel Extraction kit (Qiagen, Hilden, Germany).
Sequencing and analysis
Gel-extracted Vκ and Vλ family products were further amplified using a BigDye Terminator Cycle Sequencing Ready Reaction kit (Perkin-Elmer Applied Biosystems, Weiterstadt, Germany) and 5′ internal primers of each family followed by sequencing in an automated genetic analyzer, the ABI Prism 310 (Applied Biosystems). Sequence analysis was performed using the online program JOINSOLVER®19 (US National Institutes of Health, National Institute of Arthritis and Musculoskeletal and Skin Diseases) and productively rearranged sequences were matched to the closest germline configuration in the ImMunoGeneTics information system database. Out of frame and nonproductive sequences with a stop codon were excluded from the analysis.
Statistical analysis
Statistical analyses were performed using GraphPad Prism 3.03 (GraphPad Software, San Diego, CA, USA). Frequencies of Vκ and Vλ families as well as Jκ and Jλ gene usage were compared using 2-tailed Fisher’s exact test. Chi-square test was used to compare the mutational frequencies, RGYW/WRCY mutations, and R/S mutations, while CDR3 lengths were compared by nonparametric Mann-Whitney U test. P values < 0.05 were considered statistically significant.
RESULTS
Distribution of Vκ and Vλ family repertoire
In order to identify potential therapy-related molecular modifications in circulating B cells, we studied the Ig light chain gene usage of 2 RA patients in detail. CD19+CD27+ memory and CD19+CD27– naive B cells from both patients used light chain gene rearrangements in similar proportions in their productive repertoires, with no significant differences. Therefore, we grouped the Ig sequences from both patients and examined the changes in the light chain repertoire in terms of variable (V) and joining (J) gene usage, CDR3 length, and mutational imprints.
Before depletion, Vκ1 family members were significantly overexpressed, followed by Vκ2 and Vκ3 family members. Vκ4 family genes were observed at lower frequencies, whereas no Vκ5 or Vκ6 family members were found. During repletion after RTX, a number of changes of V family gene usage were observed. The frequency of occurrence of Vκ1 family members decreased (63.5% vs 48%; p < 0.05), while occurrence of Vκ3 genes increased (13.5% vs 26.5%; p < 0.05; Figure 1A). The frequencies of Vκ2 and Vκ4 family members before and after RTX were not statistically significantly different.
Vλ light chain analysis showed that Vλ1, 2, and 3 family members were predominantly employed by the expressed repertoire before and after depletion with Vλ2 genes, followed by Vλ1 predominantly used genes before and after therapy. Vλ6, 7, 8, and 9 family members were observed at very low percentages. Usage of Vλ families remained mainly stable between baseline and the followup after B cell depletion (Figure 2A).
Of note, the relation between Vκ and Vλ gene usage was comparable between the 2 timepoints (52% Vκ vs 50% during regeneration; 48% Vλ vs 50% during regeneration), excluding the possibility that RTX may have led to different usage of the 2 light chain gene repertoires.
Usage of Ig Vκ and Vλ minigenes by the CD19+/CD27– naive B cell repertoire
Analysis of individual VL minigene usage revealed that a wide range of Vκ genes were employed by CD27– B cells and showed a number of changes after therapy (Figure 3). Certain genes, such as Vκ1–5.03 and Vκ1–39.01, were used very often. Moreover, minigenes like Vκ1–6.01, 1D-8.01, 1D-13.01, 1–37.01, and 3–7.01 were expressed at lower frequencies before therapy, but were not found after therapy. By contrast, the genes Vκ1–13.01, 1–13.02, 1–16.01, 2–24.01, 2–30.01, 2–40.01, and 5-2.01 were not detected before depletion, but occurred in the regenerating repertoire (Figure 3A). As expected, most of the CD27– naive B cells were unmutated, with a small number of sequences carrying very low mutations (21 sequences with 1–3 mutations, 2 sequences with 6–10 mutations out of 172 sequences).
Usage of individual Vλ genes among naive B cells revealed a predominant expression of Vλ1–44.01 and 2–14.04. All genes observed before therapy were also identified in the reemerging repertoire. In addition, several genes such as Vλ1–51.01, 2–11.03, 3–9.01, 3–21.01, and 9–49.01 that were not found before therapy were identified after therapy. However, the frequency of occurrence of Vλ3–21.01 was significantly higher after RTX compared to baseline (p = 0.01). With the exception of Vλ3–21.01, there was no statistically significant difference of Vλ gene usage within naive B cells.
Usage of Ig Vκ and Vλ minigenes in CD19+/CD27+ memory B cell repertoire
Analysis of the CD27+ memory B cell compartment reflecting the superimposed picture of B cell activation, differentiation, and selection provided further interesting insights. In terms of individual Vκ gene usage, Vκ1–5.03, 1–39.01, and 4-1.01 genes were highly expressed before as well as after therapy. Several genes were found at lower or enhanced frequencies at the 2 timepoints of analysis, among which usage of 1–5.03 and 2–30.01 were significant after RTX (Figure 3).
Examination of individual Vλ genes revealed a wide array of genes constituting the repertoire before and after depletion, i.e., Vλ1–40.01, 1–44.01, 1–47.01, 2–8.01, 2–14.01, 2–23.01, 3–21.02, and 6–57.01. Some genes (Vλ1–40.03, 2–23.02, 3–9.01, 3–12.02, 3–25.01, 3–25.03, 5–39.01, and 8–61.01) that were not found in circulation before therapy were detected in the recovery phase, albeit at lower levels, except a few significant differences (Figure 3D) of occurrence of individual Vλ minigenes before and after therapy.
Imprints of potential receptor revision: usage of Jκ proximal and Jκ distal V gene cassette
Receptor revision occurs mainly at the Igκ locus through V– to Jκ secondary rearrangement. We further investigated whether the use of V genes between the proximal versus distal cassette and their rearrangements to particular Jκ elements differed, which may reflect active receptor revision. The Jκ proximal cassette is arranged when most rearrangements occur by deletion, whereas the Jκ distal cassette is arranged during recombination processes by inversion and in opposite orientation to that of the proximal cassette. Among Jκ genes, Jκ5 gene is most distally located. Analysis of Jκ gene usage (Figure 1B) revealed that Jκ2 was used most frequently before and after RTX. Moreover, RTX led to enhanced Jκ1 usage (15% vs 22%; p = 0.05) and a reduced frequency of Jκ4 gene usage (13% before vs 6% after RTX; p < 0.025). Because receptor revision leads to greater usage of Vκ genes from Jκ distal cassettes, the frequency of these genes was assessed. Although the overall use of Jκ5 gene segments (as an important indicator for active receptor revision) was not substantially different between the 2 timepoints (Figure 1B), RTX treatment led to significantly reduced usage of distal Vκ genes (2.5%) compared to baseline (14%; p = 0.0006; Figure 1C). This indicated that the imprints of receptor revision were likely more active at baseline, compared to the period after RTX. Consistent with this, the mutational frequencies between Vκ rearrangements of CD19+CD27+ memory B cells employing Jκ5 were significantly lower than those using Jκ1–4 (p = 0.02; Figure 1D).
Usage of Vλ minigenes according to their Vλ clusters and Jλ gene usage
Further studies investigated to what extent the usage of certain Vλ and Jλ genes may indicate a different role of receptor revision before and after RTX. Therefore, we analyzed Vλ gene usage in both subsets according to Vλ gene clusters. In this regard, Vλ genes are segregated into 3 clusters, A, B, and C, cluster A being the most Jλ proximal and cluster C being the most Jλ distal segment20. Notably, the frequencies of Vλ gene usage residing in cluster A, B, and C genes were comparable at baseline and during the reconstitution phase after RTX (Figure 3C, 3D). Consistent with the Vκ analysis, the usage of distal Vλ genes was reduced after RTX therapy (22% at baseline vs 16% after RTX; p = 0.147; Figure 2C), indicating that receptor revision was likely more active at baseline.
The frequency of usage of Jλ elements was also comparable at baseline and after RTX, with moderate variations as indicated by an increase of Jλ1, 2, and 3 gene usage and decreased use of Jλ7 genes (not statistically significant; Figure 2B). These data indicated that B cell depletion by RTX led to some changes in Vλ/Jλ gene usage; however, these were not significant. Moreover, analysis of the mutational frequency of Jλ1–3 containing Vλ rearrangements was comparable to those rearrangements employing Jλ7 (Figure 2D). Together, the data indicate that there were no striking changes of receptor revision of Vλ expressing B cells in these patients with RA undergoing RTX treatment.
Frequency of somatic hypermutation in Vλ and Vκ repertoire of CD27+ memory B cells
There were substantial differences in the mutational frequency of the CD19+/CD27+ memory repertoire before compared to after anti-CD20 therapy. As expected, nearly all of the CD27+ sequences were mutated, with comparable mutational frequencies of the Vκ and Vλ repertoires. Notably, CD27+ memory B cells showed an almost 2-fold increase of the mutational frequency of Vλ and Vκ rearrangements after RTX therapy (5.6%; Figure 4A) compared to that before therapy (3.2%; p < 0.0001), with Vκ 3.2% ± 0.9% at baseline versus 5.6% ± 0.1% in the repletion phase and Vλ 3.1% ± 0.1% versus 5.6% ± 0.1%. This increment of the mutational frequency was related to enhanced mutations residing in framework regions (FR 1 to 3) as well as complementary determining regions (CDR 1 and 2; Figure 4B, 4C). Further examination of the expression of mutations in the CDR and framework regions showed a reduced ratio of replacement to silent mutations (R/S) in the CDR (p = 0.001) and in the FR (p = 0.04) after treatment. No differences were observed in the frequency of transitions versus transversions or in the RGYW/WRCY hotspot motif (data not shown) targeted during B cell repletion.
CDR3 length
The overall CDR3 lengths were 30.0 ± 0.27 and 27.0 ± 0.16 (median ± SEM) within the Vκ rearrangements by naive and memory B cell rearrangements at baseline, respectively. These CDR3 lengths remained unaltered after anti-B cell therapy (30.0 ± 0.31 for naive B cells and 27.0 ± 0.26 for memory B cells). As well, the CDR3 lengths of Vλ rearrangements were similar in both naive and memory B cell subsets (33.0 ± 0.34 vs 30.0 ± 0.26 in naive; 27.0 ± 0.40 vs 30.0 ± 0.30 in memory B cells) before and after RTX. The lack of changes in CDR3 lengths suggests that selective pressures effective on the CDR3 length were less strikingly different before and after RTX therapy.
DISCUSSION
B cell depletion by the anti-CD20 antibody RTX has proven to be an efficacious and comparably safe treatment in patients with moderate to severe RA as well as other autoimmune diseases7,8,21, supporting the role of B cells in their pathophysiology. RTX-mediated B cell depletion yields a longterm effect on immunological homeostasis including distinct peripheral blood B cell subpopulations with high numbers of transitional B cells8, indicating recapitulation of B cell ontogeny. Although numeric normalization of B cell numbers takes place in the majority of patients between 1 and 2 years after a single course of RTX, a longterm delay in memory B cell subsets has been observed especially in responders8,9. Gene rearrangement analysis of IgV using single B cells allows study of different processes such as somatic hypermutation and receptor revision. Abnormalities in these events alter the composition of the peripheral B cell repertoire. Previous observations provided evidence of the influence of RTX on molecular imprints shaping the Ig heavy chain receptor15,22,23, but did not allow any evaluation of receptor revision. Receptor revision together with the imprints of somatic hypermutation, CDR3 length, and Ig switching represent results of T cell and antigen-dependent B cell activation and differentiation, which also reflect molecular signatures of immune memory. To date, modulatory effects in the immunoglobulin light chain repertoire after RTX have not been studied in detail. Therefore in this study we investigated the effects of RTX on the Vκ and Vλ light chain repertoire genes in the CD27– and CD27+ B cell compartment in patients with active RA, using single B cell analyses before and during B cell repopulation after RTX.
Different mechanisms shaping the human light chain repertoire including selection and combinational biases serve the limitation of expressed repertoire. Repertoire revision in B cells and the clonal selection processes contribute to somatic evolution of appropriate immune responses under normal circumstances6. Examination of individual Vκ minigene distribution revealed a diverse repertoire of genes, with specific Vκ1 family members being predominantly expanded in repopulating B cells. A similar overrepresentation of Vκ1 family members has also been reported in cDNA libraries studied from synovial tissue of a patient with RA in comparison to healthy donors24. A restricted repertoire of Vκ has been reported in patients with juvenile arthritis11. Interestingly, we did not find a restricted repertoire of Vκ family gene usage in our patients with RA, which may indicate an age-dependent modulation (Figure 1A, 1B). The Vκ family repertoire changed significantly after RTX (Figure 3A) as indicated by a decrease in the use of Vκ1 and an increase in Vκ3 families. Interesting as well, the B3 gene, a member of the Vκ4 family, previously shown to be overexpressed in patients with systemic lupus erythematosus (SLE)25 and linked to encode autoantibodies26, was not used by our 2 patients with RA, consistent with a differential usage of Vκ minigenes during flares of different autoimmune diseases. Analysis of the Vλ repertoire showed the majority of minigenes in equal distributions after RTX, except Vλ3–21.01, which was used significantly more (p = 0.01) in repopulated CD27– B cells, while 1–51.01 and 6–57.01 were reduced in CD27+ B cells after RTX. These findings indicate that a diverse and polyclonal Vκ and Vλ gene repertoire is emerging after RTX-induced temporal B cell depletion, suggesting a reestablishment of B cell homeostasis.
There is controversy whether defective mechanisms in secondary gene rearrangements are involved in generating autoimmunity. Studies indicate that receptor revision is a principal mechanism of B cell tolerance12 and defective central or peripheral receptor revision has been linked to autoimmunity27. The frequency of J-distal Vκ gene usage was significantly less than J-proximal Vκ genes after therapy (Figure 1C). These data indicate that imprints of receptor revision are already evident before RTX, and therapy-related changes do not support a potential defective receptor revision in patients with RA. Two different molecular mechanisms are operative on Vκ/Jκ rearrangements, one by inversion and another by deletion. Inversion is believed to be the significantly less efficient process, suggested by artificial recombination substrate studies28,29. Recombination requires rearrangement over a long stretch of DNA because the J-distal cassette is 1500 to 2000 kb from the nearest Jκ gene, with possibly less efficient recombination machinery over distances > 1000 kb. While no firm conclusion can be drawn about the exact mechanisms applied by patients with RA, the current data for significantly reduced usage of Jκ-distal genes in repopulated B cells suggest that RTX therapy normalizes B cell homeostasis by reducing the imprints of receptor revision, which may have been active particularly on VκJκ gene rearrangements. Because of the distal location of Jκ5, a diminished mutational frequency was observed in memory B cell rearrangements compared to those using Jκ1–4 (Figure 1D) at baseline. Together, the data indicate that receptor revision took place at baseline on the Vκ locus, excluding the possibility of a defective receptor revision process in these patients with RA. Moreover, receptor revision in Vκ probably occurred in secondary lymphoid organs after initiation of somatic hypermutation, as suggested by fewer mutations in rearrangements using Jκ5.
Analysis of the Vλ gene repertoire did not reveal significant differences after RTX. The overall usage of J-distal Vλ genes was decreased without reaching statistical significance (Figure 4C), indicating that the effects of RTX on Vλ gene rearrangements were less striking compared to the Vκ locus. This is consistent with the conclusion that receptor revision in Vλ occurred centrally at baseline.
We did not find increased usage of cluster C genes, in contrast to previous findings in a patient with SLE30. Increased use of cluster C genes is believed to be a predictor of extensive receptor revision in Vλ gene rearrangements because it is also observed in infants compared to healthy subjects31. Cluster C genes are most Jλ distally located with more than 600 kb upstream of Jλ1 locus. In our patients with RA, cluster A and B genes were frequently used, consistent with findings from healthy donors31. Only moderate changes were observed in the usage of Jλ genes during therapy (Figure 2B), suggesting that the basic molecular mechanisms driving recombination and selection of Vλ chain rearrangements were not substantially altered by RTX.
We noted profound changes in the CD27+ memory B cells during B cell repopulation after RTX. The mutational pattern of repopulated CD27+ memory B cells was marked by the prevalence of highly mutated Vκ and Vλ Ig gene sequences (Figure 4A). These findings are in accord with our previous findings, where we analyzed the mutational pattern of Ig-VH4 gene in a cohort of RA patients treated with RTX15, and support the concept that RTX-mediated B cell depletion modulates Ig heavy and light chains. It is assumed that repopulation of the peripheral repertoire is mainly derived from early undepleted CD20-negative pro-B cells, and permits early occurrence of highly mutated sequences during B cell repopulation. These highly mutated cells during the early regeneration phase might develop from a newly generated B cell pool that rapidly undergoes intensive somatic hypermutation and circulates in the periphery. However, part of the highly mutated memory pool may also be dependent on surviving cells, which can be found in the periphery in low frequency during the depletion phase and may generate from protected niches32,33. The increased mutational frequency by regenerated memory B cells can be explained by a marked reduction in pre-switch memory B cells and a relative increase of recirculating plasmablasts during the regeneration phase (Table 1). Detailed analyses of CDR (Figure 4B) showed that the Vλ and Vκ repertoire exhibited increased mutational frequency in the CDR2, followed by CDR1, before and after depletion. Similar observations were noted within framework regions. This indicates that B cell depletion by RTX introduced significant modulations in the regenerating Vλ/Vκ repertoire. Of note, the ratio of replacement to silent mutations (R/S) in the complementary determining regions and framework regions revealed that RTX induced modulations of R/S ratios. We observed a significantly decreased R/S mutational ratio after RTX (R/S CDR and FR in VL: p = 0.001, p = 0.04, baseline vs reconstitution phase, respectively). Positive selection of R mutations has been reported for VH and VL gene rearrangements in healthy individuals34,35,36, and therefore the reduction in R/S mutations may also be indicative of lower affinity of potential autoreactive clones that recirculate during B cell reconstitution in patients with RA.
Overall, this is to our knowledge the first study to report the regeneration pattern of Ig Vκ and Vλ light chain gene rearrangements during RTX therapy. Receptor revision was operative in these patients mainly on the Vκ locus at baseline. Transient B cell depletion introduced differential VκJκ and VλJλ minigene usage in the regenerating naive and memory B cell repertoires. Memory B cells were substantially marked by an elevated mutational frequency of the reemerging CD27+ light chain. While our findings do not support defects of receptor revision, at least on the Vκ locus in these patients with RA, the data provide evidence that anti-CD20 therapy probably alters the molecular characteristics of the V light chain gene repertoire consistent with a reduction of receptor revision.
Acknowledgment
We thank Katharina Eckert, Anette Koss-Kinzinger, Isabelle Kuntzsch, and Christian Linden for their excellent technical assistance.
Footnotes
-
A. Palanichamy and K. Muhammad share first authorship of this article.
-
Supported by Wilhelm Sanderstiftung 2006.037.2, University of Würzburg, Würzburg, and project 14, SFB 650, Charite University, Berlin.
- Accepted for publication February 23, 2012.
REFERENCES
- 1.
- 2.
- 3.
- 4.
- 5.
- 6.
- 7.
- 8.
- 9.
- 10.
- 11.
- 12.
- 13.
- 14.
- 15.
- 16.
- 17.
- 18.
- 19.
- 20.
- 21.
- 22.
- 23.
- 24.
- 25.
- 26.
- 27.
- 28.
- 29.
- 30.
- 31.
- 32.
- 33.
- 34.
- 35.
- 36.