CD21^low B cells in common variable immunodeficiency do not show defects in receptor editing, but resemble tissue-like memory B cells

To the editor:

We read with great interest the article by Isnardi et al¹  wherein the authors showed very similar phenotypic and genotypic features of CD21^−/lo B cells as previously reported.²  In view of the polyreactivity, increased autoreactivity, and strongly reduced calcium response and proliferative capacity of CD21^−/lo B cells—reminiscent of anergic B cells in mouse models—the authors propose that CD21^−/lo B cells represent anergic human B cells.¹  Moreover, they showed that the majority of highly HEp2-reactive CD21^−/lo B-cell clones express kappa (κ) light chains and suggested that the receptor editing using lambda (λ) light chains might not be induced in those clones serving as a mechanism to silence autoreactive B cells. We would like to suggest a different interpretation of the origin of CD21^low B cells.

Anergy in CD21^low B cells in patients with common variable immunodeficiency (CVID) is not complete. In the original model of hen egg lysozyme–transgenic mice,³  anergy in B cells represents a mechanism of central tolerance caused by B cell receptor (BCR)–induced hyporesponsiveness, which is associated with down-regulated BCR, poor calcium response, defective up-regulation of CD86, failure of antibody production, and impaired proliferation.^3,4  Although poor Ca⁺⁺ response and impaired proliferation are both features that CD21^low B cells share with anergic mouse B cells, CD21^low B cells of CVID patients have not down-regulated their surface BCR (Figure 1A-B) and are able to up-regulate activation markers like CD86 (unpublished data) as well as ICOS-L, TACI, and Fas surface expression upon BCR stimulation.¹  Even more importantly, CD21^low B cells produce higher levels of IgM than naive B cells after stimulation in vitro.²  In addition, a comparison of gene-expression profiles of mouse anergic B cells^3,4  and human CD21^low B cells²  showed only little overlap: only 3 (11.5%) of 26 genes describing the signature of anergic mouse B cells were identically expressed in CD21^low B cells. In contrast, these cells closely resemble²  the recently characterized tissue memory-like B cells in tonsils⁵  and circulating “exhausted” CD21^low B cells in HIV patients, suggesting rather a mechanism of activation-driven peripheral exhaustion underlying this anergic phenotype.⁶  This would also better explain the accumulation of CD21^low B cells in the bronchoalveolar fluid of CVID patients or the synovial fluid of rheumatoid arthritis patients.² 

In addition, in our cohort of CVID patients we cannot confirm a defect in secondary editing and λ light chain usage. We actually found that CD21^low B cells preferentially express λ-chains compared with the naive B cells of CVID patients, which was also the case in 1 of the 3 CVID samples shown in Figure 2D of Isnardi and colleagues' article.¹  In our experience the κ/λ ratio in CD21⁺ naive B cells (1.24 ± 0.03; n = 4) of CVID patients was within the normal range (1.0-2.1), whereas CD21^low B cells showed a significantly decreased usage of κ-chains (0.69 ± 0.10; n = 4; P = .0019) compared with naive B cells of the same patients (Figure 1C-D). In addition, our analysis of κ recombination excision circles (KREC)⁷  had revealed a more extensive proliferative history of CD21^low B cells compared with naive B cells.²  We appreciate the beautiful demonstration of increased autoimmune reactivity of CD21^low B cells by Isnardi et al,¹  but our data do not support an editing defect in central B-cell selection as the underlying etiology.

According to our data it seems more likely that CD21^low B cells of CVID patients are related to tissue-like memory B cells in tonsils of immunologically healthy individuals⁵  and therefore resemble tissue associated innate-like B cells, which reveal partial anergy in vitro due to their “exhausted” preactivation state in vivo.