On the differentiation of mouse IgE⁺ cells

To the Editor:

We read with interest the paper by Talay et al. in the April issue of Nature Immunology¹. In this article, the authors use a green fluorescent protein (GFP) reporter mouse strain to establish the rules that govern immunoglobulin E (IgE) responses in vivo. After carefully reading the manuscript, we believe that there are several issues that challenge the validity of their experimental model and, thus, their conclusions.

The reporter construct used by Talay et al. has two features that require careful confirmation¹. First, the construct adds an exon encoding human membrane IgE (M1′) to the existing exons encoding mouse membrane IgE. The resulting mouse-human chimeric IgE molecule is an altered IgE molecule that is longer by 52 amino acids. This humanized mouse molecule should not be the first choice of model with which to study normal mouse IgE responses. The longer mouse-human chimeric IgE molecule cannot be considered to reflect normal IgE structure, aggregation and signaling in the absence of comparative studies, which are missing in the present manuscript¹, or the previously published article by these authors describing the humanized mouse strain². Second, the authors replace the IgE 3′-polyadenylation site. The 3′-polyadenylation site in IgE is unusual in its sequence and organization, and it has been shown that a point mutation that creates a consensus site changes the ratio of membrane IgE to secreted IgE by tenfold³. This by itself could explain some of the unexpected data presented in the manuscript by Talay et al.¹. Furthermore, there is no comparison of the ratio of membrane IgE to secreted IgE generated by the targeted allele versus that generated by the normal allele.

Another potential issue with the study by Talay et al.¹ is that not all GFP⁺ cells are IgE⁺. The safest confirmation of a membrane IgE reporter mouse strain would be to correlate GFP expression with surface staining with an antibody to mouse IgE. However, such IgE staining is absent throughout this manuscript¹. The only staining with antibody to IgE (anti-IgE) in the manuscript is one panel of intracellular staining. Instead of anti-IgE, the authors use an antibody to the human domain M1′. Although the M1′ staining data in Figure 1c of the article is described as indicating that the cultured cells “expressed membrane IgE”¹. The shift in fluorescence intensity of the sample stained with anti-M1′ relative to that of the control sample is minimal. The overlay of the flow cytometry of cells from Nippostrongylus brasiliensis–infected mice with anti-M1′ shows almost perfect overlap with the that of isotype-matched control antibody on the left of the anti-M1′ curve and, in addition, 'piling up' of many GFP⁺ cells completely negative for M1′ along the vertical axis (Supplementary Fig. 4b of the article)¹. Thus, a substantial proportion of GFP⁺ cells seem to be negative for M1′ in N. brasiliensis–infected mice. These GFP⁺ cells probably express immunoglobulins of other isotypes.

A published paper by Yang et al. uses a different IgE reporter system⁴. In contrast to the model proposed by Talay et al., Yang et al. show that IgE⁺ germinal-center cells diminish rapidly during an ongoing response, whereas IgG1⁺ germinal-center cells persist⁴. Furthermore, IgE⁺ plasma cells have considerably fewer mutations than do IgE⁺ germinal-center cells, an indication that highly mutated IgE⁺ germinal-center cells are not part of the main IgE response. It remains most likely that, as proposed before, class switching to IgE is linked to a plasma-cell differentiation program and that IgG1⁺ germinal-center cells contribute to the IgE⁺ plasma-cell compartment^5,6. In fact, we have shown that mice deficient in IgG1 have a considerable impairment in their ability to generate affinity-matured IgE antibodies⁶. Although the reporter allele used by Yang et al. is less manipulated than that used by Talay et al., the ratio of membrane IgE to secreted IgE produced by the targeted allele in the mice studied by Yang et al.¹ is two- to threefold higher than that produced by the normal allele encoding IgE, and this allele is, possibly, altered even further in the mice used by Talay et al. The differences between the data obtained by Talay et al. and the data of other groups could reflect substantial differences in the activity of the longer humanized molecules used by Talay et al. relative to that of normal mouse IgE molecules.

In conclusion, we find that the report by Talay et al. is lacking in several important ways and therefore would like to bring to the attention of Nature Immunology readers the existence of alternative explanations for most of their observations.