Long-lived IgE plasma cells persist in secondary lymphoid tissues using a Navitoclax-sensitive survival program

Summary:

Long-term allergies exhibit enduring concentrations of circulating IgE. Here we examined the lifespan of IgE antibody-secreting cells (ASC), to determine whether the IgE that sustains allergies receives contributions from long-lived cells or relies more heavily on constant ASC production. In mouse aeroallergy, IgE ASC localized to lungs, mediastinal lymph nodes, spleen, and bone marrow (BM). IgE ASC production continued for months after allergen exposure ceased. We identified long-lived IgE ASC residing predominantly outside BM, with a half-life exceeding 49 days; in contrast, most IgE ASC had a three-day half-life. Long-lived IgE ASC matured phenotypically, became quiescent, retained surface B cell receptors, but had low expression of the BM homing receptor CXCR4. They were hierarchically more reliant on the Navitoclax-sensitive anti-apoptotic molecules BCL2, BCLXL and BCLW than MCL1. Thus, continual production of short-lived IgE ASC and retention of long-lived IgE ASC outside BM together drive IgE persistence, perpetuating allergic disease.

eToc blurb:

Ding et al. show that allergy-associated IgE plasma cells exhibit limited accrual in bone marrow, instead residing in other tissues for extended periods. These long-lived cells are rare and sensitive to navitoclax rather than deletion of the anti-apoptotic protein MCL1,which effectively kills other plasma cells.

Graphical Abstract

INTRODUCTION

IgE antibody is a critical mediator of allergic diseases, binding mast cells and basophils to elicit their degranulation upon exposure to allergen, triggering allergy symptoms¹. While serum IgE can persist for months in the absence of known allergen exposure^2,3, the lifespan of IgE antibody-secreting cells (ASC) is unclear. Two competing, yet not mutually exclusive models have been proposed to explain the maintenance of IgE titers in the absence of known allergen re-exposure, these being the continuous production of short-lived IgE ASC, or the persistence of long-lived IgE ASC. A small amount of data supports the existence of long-lived IgE ASC. As examples, first, measurement of IgE antibody titers as an indirect measure of IgE ASC persistence suggests that the cells are maintained for many months in the context of helminth infection⁴, and potentially years in humans⁵ and similarly, IgE titers can remain elevated in periods of allergen avoidance². Second, individuals treated with the IL-4Rα-inhibiting drug dupilumab⁶, a treatment presumed to prohibit de novo IgE ASC formation and thus leave only long-lived IgE ASC in place⁷, retain 5–10% of starting IgE over the course of a year, suggesting rare long-lived IgE ASC may exist⁸. Third, in a chronic allergic airway disease model in IgE reporter mice, reporter positive ASC reside in bone marrow (BM) 9 weeks after the cessation of allergen challenges⁹. Fourth, indirect assessments show that dramatically reducing the B cell population does not expedite the decline of serum IgE titers⁹, and inhibiting the signaling pathways that drive IgE ASC production leaves some residual IgE ASC in BM¹⁰, both tests supportive of the existence of long-lived IgE ASC. However, it is difficult to know or confirm that IgE ASC production is completely prevented with such treatments. Further, most BM ASC turn over every few weeks^11,12, so residence there does not necessarily equate to longevity, meaning IgE ASC could be invariably short-lived, despite being found in BM.

In contrast to the evidence for long-lived IgE ASC, several pieces of data indicate that IgE ASC have poor survival^13,14, leading to the idea that they are continuously generated. IgE B cell receptor (BCR) expression is a driver of apoptosis^15,16, and ligation of the IgE BCR limits the survival of IgE ASC in vitro and in vivo¹⁷. Indeed, a large fraction of^14,18 or potentially all¹⁹ IgE ASC are short-lived cells and longevity may be entirely impossible for IgE BCR-bearing cells. This is supported by activation of the Bcl2l11 (Bim)-dependent death cascade upon ligation of the IgE BCR¹⁸ and a low ability to respond to CXCL12^15,20, the chemokine mediating recruitment to BM where ASC are thought to receive key signals that support long-term survival²¹. The IgE BCR also undergoes autonomous antigen-independent signaling^16,22, meaning even without antigen exposure, these outcomes may associate with IgE BCR expression. Although noting that only reasonably short-term models have been used to assess IgE ASC surface BCR expression, at least at the plasmablast (PB) stage they have high levels of IgE BCR^17,23,24, so it could be considered anomalous for IgE ASC to survive long-term. Human in vitro cultures suggest that IgE ASC fail to undergo terminal differentiation from the PB stage to a mature, quiescent PC state and further exhibit high levels of apoptosis, indicating constrained survival of human IgE ASC²⁵. Similarly, IgE ASC from nasal polyps and blood from allergic individuals are transcriptionally immature based on their PB state or retention of genes encoding MHC-II, CD19 and BAFF-R^26,27. BM IgE ASC exhibit low expression of Mki67 (Ki-67), suggesting they can mature beyond the PB state in that tissue^{10,28(preprint)}. However, the cells maintain high expression of the death receptor FAS, suggesting they are poised for death, and further exhibit signs of endoplasmic reticulum stress^10,28, meaning their survival may be limited relative to other proven long-lived ASC populations. Importantly, however, a subpopulation of IgE ASC exhibit high expression of the pro-survival protein gene Bcl2¹⁰, suggesting survival mechanisms may be engaged that support the cells long-term, despite exhibiting divergent transcriptomes from the more abundant and known isotypes of persistent ASC. Thus, all available data considered, there are transcriptional indications of longevity in a subset of IgE ASC, but ultimately no definitive evidence for or against a transcriptionally and phenotypically mature, long-lived IgE ASC population. A key open question is what distribution of lifespans IgE ASC have.

The immature and stressed state of IgE ASC evokes questions about the survival mechanisms supporting ongoing IgE responses because their survival program is undefined. Mcl1 deletion combined with inhibition of BCL2, BCL_XL and BCL_W with the BH3 mimetic ABT737 results in dramatic reductions of splenic and BM ASC²⁹. MCL1 is the main survival protein for ASC³⁰, however, our understanding of ASC survival programming is incomplete. ASC in spleen, which have relatively lower transcription of Mcl1 are more resistant to its deletion than BM ASC^29,30, and while deletion of Mcl1 impacts total BM ASC abundances^29,30, a fraction of BM ASC exhibit sensitivity to ABT737³¹, suggesting alternate survival programs for BM ASC sub-populations. As IgE ASCs were not investigated in these prior studies, nor were long-lived ASC definitively identified, both the IgE ASC survival program and that of long-lived ASC in general, remain to be determined.

To resolve whether continuous production of short-lived IgE ASC or retention of long-lived IgE ASC best explains the maintenance of IgE titers, we used genetic timestamping to analyze the production and persistence of IgE ASC in house dust mite (HDM)-driven allergic airway disease. IgE ASC were continually produced as short-lived cells, with the majority localizing in secondary lymphoid tissues (SLT) and showing evidence of poor BM tropism: only rarely did the cells obtain longevity, and even these localized atypically, residing in SLT and comparatively infrequently in BM. Unexpectedly, rather than relying on MCL1, persistent IgE ASC showed a BCL2, BCL_XL or BCL_W dominated survival program. These findings show that BH3 mimetics are effective at attenuating cellular IgE responses and implicate both continual production of short-lived IgE ASC and retention of longer-lived IgE ASC as important for the etiology of IgE-mediated disease.

RESULTS

IgE PC detection in multiple tissues after chronic allergen exposure

We followed an established protocol that generates putatively long-lived IgE ASC^9,10 in which mice are subjected to three-times weekly intranasal HDM exposure for 7 to 15 weeks, then measured total serum IgE, total serum IgG1, and HDM-specific IgG1 amounts by ELISA, and HDM-specific IgE responses by basophil activation test (Figure S1). Seven weeks of three-times weekly HDM (Figure S1A) was sufficient to generate total IgE and IgG1 amounts elevated above naïve controls, and the IgE amounts persisted above background for 114 days after challenge cessation (Figure S1B and S1C). Similarly, fifteen weeks of challenge (Figure S1D) generated elevated total IgE and IgG1 titers at the endpoint (Figure S1E and S1F, respectively). Basophils derived from mice receiving acute HDM treatment (Figure S1G and S1H) required around 10-fold more HDM for activation compared to one week after chronic challenge (Figure S1I and S1J), showing an increased IgE response with extended challenge. The HDM-specific IgE sensitization and an HDM-specific IgG1 titer elevation persisted for at least 9 weeks after cessation of treatment (Figure S1K and S1L). These data show that 7 to 15 weeks of three-times weekly HDM challenge generates sustained IgE titers.

IgE ASC can reside in lung, secondary lymphoid tissues and BM^10,32, but the relative contribution of each to total IgE production after HDM sensitization was unclear. We therefore sought to define the tissue sources of the produced IgE. We analyzed various tissues after 15-weeks of HDM treatment for Venus^bright IgE ASC using Verigem IgE reporter mice, mice in which Venus fluorescent protein reports membrane IgE positive cells^23,33. We examined tissues where allergen is drained after intranasal administration, namely the nasal-associated lymphoid tissue (NALT), posterior cervical (pc) lymph nodes (LN), lungs, and right mediastinal LN (medLN), as well as the non-draining tissues, spleen, femoral BM, parathymic (pt) LN, and superficial cervical (sc) LN. This revealed small populations of Venus^bright IgE ASC in lung, medLN, spleen and femoral BM of HDM-treated mice, while Venus^bright cells were not found after HDM treatment in the other tissues and naïve controls lacked Venus^bright IgE ASC in all tissues except for a single event in NALT (Figure S2A). We therefore focused our analyses on the select tissues in which populations of Venus^bright cells were identified in HDM-sensitized mice.

IgE ASC were present in femoral BM but were more abundant in medLN and spleen after chronic HDM sensitization

To confirm the above findings, we next modified an established protocol²³ to detect IgE ASC without the Venus reporter. Therefore, we used WT C57BL/6 or BLTcre mice (in which Prdm1 promoter-driven TdTomato (TdT) expression identifies ASC³⁴) combined with a stringent flow cytometry protocol for IgE detection relying on dual surface (s) and intracellular (ic) IgE expression in the cells (Figure S2B–L, details in STAR Methods section).

In C57BL/6 mice after 15 weeks of HDM treatment, we examined lung, medLN and BM for IgE_(ic+s)⁺ ASC, identifying WT ASC as CD138⁺CD98⁺ cells in medLN and BM, and CD98⁺ in lung. IgE_(ic+s)⁺ cells were present in all three tissues two days after cessation of allergen challenge (Figure 1A). The abundance of IgE_(ic+s)⁺ ASC differed by tissue (Figure 1A) and the same treatment scheme using BLTcre hCD4 reporter mice³⁵ showed IgE_(ic+s)⁺ ASC in similar representations to the C57BL/6 mice in medLN and BM, and further confirmed IgE_(ic+s)⁺ ASC were also present in the spleen (Figure 1B). IgE_(ic+s)⁺ ASC made up on average 2000 cells per medLN, 200 per lung, typically fewer than 100 per femur pair, and ≈ 1000 per spleen (Figure 1C and 1D). Thus, in line with earlier findings^9,10, BM harbored IgE ASC following chronic HDM treatment, but importantly, our data reveal medLN and spleen as major reservoirs.

Fig. 1: Tracking the IgE response in tissues.

(A, B) IgE ASC as the proportion of live events or (C, D) enumerated in different tissues gated as CD11c^loIgD^loFcεRIα^loCD98^brightIgG1^negIgE_(ic+s)⁺ in lung and additionally CD138⁺ in medLN and BM in C57BL/6J mice, and as CD98^brightTDT^brightIgD^loIgG1^negIgE_(ic+s)⁺ in BLTcre hCD4 reporter mice. (E) IgG1 ASC as the proportion of live events in various tissues, gated as CD11c^loIgD^loFcεRIα^loCD98^brightIgG1_(t)⁺IgE_(ic)^neg in lung and additionally CD138⁺ in medLN and BM. (F) The ratio of IgE:IgG1 ASC across different tissues. Data shown are from one experiment to show the normal IgE response in the BLTcre hCD4 reporter line without reference to hCD4 expression (B, D), one of two similar experiments (A, C, E), or combined from two (F) experiments with six mice per group per experiment. Symbols show individual mouse responses and lines denote arithmetic (A, E) or geometric (B-D) means or (F) median. * P <0.05 by (A, E) Student’s t-test, (B) One-Way repeated measures ANOVA with Tukey’s post-test, (C, D) One-Way repeated measures ANOVA with Tukey’s post-test after log-transformation, or (F) a Friedman with Dunn’s post-test. X denotes no events detected. See also Figure S1.

We also enumerated the IgG1 response by staining for total (t) IgG1 in ASC (Figure 1E). For each IgE_(ic+s)⁺ ASC we detected in medLN, spleen and BM, we detected 50, 80 and 500 IgG1_(t) ASC, respectively (Figure 1F). These data show that, compared to IgG1 ASC, IgE ASC are poor at seeding in BM and preferentially localize to SLT.

IgE ASC rarely persist after cessation of allergen exposure

The abundance of IgE ASC in medLN and spleen suggested that ongoing production might have played an important role in maintaining sensitization, while their presence in BM may have indicated a role for long-lived cells. We sought to resolve definitively which of these possibilities best explained ongoing IgE production using genetic timestamping of ASC.

We timestamped using the BLTcre.Mcl1^fl/+ strain (BLTcre crossed with an Mcl1 deletion-mediated human CD4 [hCD4] reporter), in which ASCs are indelibly labeled with surface hCD4 following Cre activation by tamoxifen (Tam) administration³⁵. Hemizygosity had no discernable impact on IgE PC numbers after sensitization to HDM (Figure S3A–S3F), nor did BLTcre-driven Cre activity diminish the T cell response (Figure S3G–S3O), confirming this as an appropriate system in which to study IgE PC persistence.

We surveyed timestamped ASC abundance and decay after 7-weeks of 3x weekly HDM challenge (Figure 2A). The starting number of IgE_(ic+s)⁺ ASC across spleen, medLN and femurs two days after the final HDM treatment (four days after Tam gavage) was 1500 (95% C.I.: 690 to 3280; Figure 2B) and the number of IgG1_(t)⁺ ASC was 85000 (95% C.I.: 48500 to 149300; Figure 2C). hCD4 representation among IgE_(ic+s)⁺ ASC and IgG1_(t)⁺ ASC declined with time, suggesting ongoing production of ASC (Figure 2D and 2E). Both total and hCD4⁺ IgE_(ic+s)⁺ ASC declined in number, the latter by 90% in 9 weeks (Figure 2B and 2F). Extending the analysis time course to 77 and 114 days post HDM showed that hCD4⁺ IgE_(ic+s)⁺ ASC could still be detected, although only statistically significantly above the background at the day 77 timepoint (Figure 2F), while hCD4⁺ IgG1_(t)⁺ ASC were reduced to relatively the same extent, yet remained above the background up to day 114 (Figure 2G). The distribution of hCD4⁺ ASC demonstrated that the medLN and spleen together held more hCD4⁺ IgE_(ic+s)⁺ ASC than femoral BM (Figure 2H) while hCD4⁺ IgG1_(t)⁺ ASC were reciprocally and substantially enriched in BM (Figure 2I).

Fig. 2: Turnover of ASC after seven weeks of intranasal HDM exposure.

(A) Treatments given during the seven-week model. (B) Total IgE and (C) IgG1 ASC detected in BM, spleen and medLN of mice (numbers across organs collated) gated respectively as CD98⁺TDT⁺IgD^loIgG1_(t)^negIgE_(ic+s)⁺ and CD98⁺TDT⁺IgD^loIgG1_(t)⁺IgE_(ic)^neg. Percentage values indicate the frequency relative to the geometric mean on day 2. (D) hCD4 expression among total detected IgE, and (E) IgG1 ASC. (F) The number of hCD4⁺ IgE and (G) hCD4⁺ IgG1 ASC over time. (H) Tissue distribution of hCD4⁺ IgE and (I) hCD4⁺ IgG1 ASC. Data shown represent geometric mean + geometric SD factor (B, C, F, G) or arithmetic mean + SD (D, E, H, I) of groups of n=5–7 mice per timepoint. Backgrounds were set by n=3 (day 2, 7, 21, 42, 63, 77) or n=4 (day 114) BLTcre.Mcl1^+/+ (BLTcre only) mice. The background in F and G is shown by the dashed line. (D, E) * P <0.05 by two-way ANOVA with Sidak’s post-test relative to the day 2 timepoint. For F, G, * denotes statistical elevation (P <0.05) above the background by two-way ANOVA with Sidak’s post-test after log-transformation. Similar observations were made in a second similar experiment with 15 weeks of 3x weekly HDM challenge (data not shown). See also Figure S2 and S3.

We conclude that in response to HDM allergen, de novo differentiation after treatment cessation accounts for maintenance of the IgE and IgG1 ASC over time because most of the cells are hCD4^neg several weeks after exposure cessation. However, ASC of both isotypes persist, indicating that sensitization is an amalgam of IgE ASC retention and continuing production, weighted here toward continuing production.

Calculated half-lives for short- and long-lived ASC

We next attempted to elucidate the lifespan structure of IgE ASC. As ASC have been reported to exhibit different half-lives in SLT and BM^11,36, we fitted the tissue types independently. We tested three models. The first tested was a single exponential, which ascribes a single half-life to all ASC within a population. A single exponential decay with a 2.8 day half-life, estimated from a previous work examining SLT IgE ASC decay over 10-days²³, generated fits incongruent with the experimental data, and performed poorly by Akaike Information Criterion (AIC)³⁷, indicating the cells in the HDM model were unlikely to have this distribution of lifespans (Figure S4A). We conclude that IgE ASC are not invariably short-lived.

To further investigate the lifespan distribution, we allowed the half-lives to float in the single-exponential model. This generated best fits of 24 (95% C.I. 18–36) days and 55 (95% C.I. 33–180) days in SLT and BM respectively (Figure S4B), but still with comparatively high AIC values and poor fits to the first timepoint in both tissues so was rejected as a structure. The second structure tested was the steady-state lognormal, wherein lifespan is described by a lognormal distribution with unsynchronized birth times, allowing an interpretation that the ASC lifespan exists on a continuum rather than in discrete short- and long-lived states¹². However, the steady-state log-normal was similarly rejected as it generated a heavily skewed structure, with an implausible mean lifespan of 0.1 days (the lower bound of the fitting range) for IgE ASC in BM and a wide standard deviation in SLT and BM compartments to fit the data set (Figure S4C). The third fit was a double-exponential, whereby cells are divided into sub-populations characterized by a short half-life, t₁, and a long half-life, t₂, and a fraction denoting the proportion long-lived, f_LL. This gave more physiological parameters for the fit than the lognormal and performed well by AIC (Figure S4D).

We were encouraged by the goodness-of-fit of the double exponential so asked if this provided a parsimonious explanation of the totality of the data, including the IgG1 ASC in SLT and BM in a combined analysis of all four compartments (IgE SLT, IgE BM, IgG1 SLT, IgG1 BM). We applied fits to the four compartments independently (Table S1) to account for potential differences in decay associated with isotype^11,38,39, or by constraining parameters between them (Figure 3A–D) and tested the suitability of the double-exponential, again against the single exponential and steady-state lognormal models.

Fig. 3: Retention and turnover of ASC in SLT and BM.

Circles and error bars indicate the data and standard error in the mean (s.e.m.) and dashed lines indicate the model fits. Fitting was performed and plotted on a log₁₀ scale to illustrate the tail of the distribution and allow comparable contribution to the s.s.e. at later time points. (A) IgE SLT, (B) IgE BM, (C) IgG1 SLT and (D) IgG1 BM ASC fit by a double-exponential model with two common half-lives across compartments. Fits are to the mice shown in Fig. 2. See also Figure S4 and Table S1.

Of all models considered (Table S1), the best by AIC was the double-exponential model with short- and long-lived half-lives of 3.1 [2.1–5.2] days and 88 [49–$\propto$] days, respectively, equal across compartments, but with $f_{LL}$ variable between them (Figure 3A–D and Table S1). In the combined double-exponential analysis, we were also able to calculate the transition time $\left(t_{\mathit{trans}}\right)$ as the point at which the relative proportions in the measured population switch from short-lived dominant to long-lived dominant. The transition occurred in a similar timeframe across compartments (IgE BM, 17 days; IgE SLT 26 days; IgG1 BM 26 days; IgG1 SLT 33 days). Thus, a double-exponential, comprising major short-lived and smaller long-lived ASC populations, explains the IgE and IgG1 ASC lifespan distribution. Short-lived ASC exhibit a half-life of ≈3-days and long-lived ASC persist for 49-days or more. Apparent differences in compartment turnover was best explained by higher proportions of long-lived ASC in BM compared with SLT. Ostensibly, this is consistent with recent modelling work arguing that fine-tuning of a gene-regulatory network drives criticality and bifurcation of the ASC population into sub-populations with brief and extended lifespans⁴⁰.

Persistent IgE ASC are mature, quiescent PC

Having generated definitive evidence of persistent IgE ASC, we asked if they exhibited signs of maturation. ASC form as I-A⁺SLAMF6⁺ PB, and after entering quiescence, lose I-A and then also downmodulate SLAMF6¹². To assess if this pattern held true for IgE ASCs, we evaluated IgE_(ic+s)⁺ ASCs in the medLN after an acute HDM response (treatment for three consecutive days), seven days after the first dose of HDM, and found > 99% of IgE_(ic+s)⁺ ASC (< 7-day old) were I-A⁺SLAMF6⁺ cells (Figure 4A), confirming a surface phenotype as found for ASC expressing other BCR isotypes¹². After timestamping, among hCD4⁺ IgE_(ic+s)⁺, I-A⁺SLAMF6⁺ cells were less frequent, with the hCD4⁺ IgE_(ic+s)⁺ ASC increasingly enriched for I-A^loSLAMF6^lo cells with time, reaching 70% by day 63 post cessation of chronic HDM treatment (Figure 4B and 4C). Therefore, IgE ASC that persist can mature phenotypically with age, similar to other ASC populations.

Fig. 4: Maturation profile of IgE ASC.

(A) The day 7 cell surface phenotype of medLN IgE ASC after receiving acute HDM treatment (i.n. 50 μg HDM on days 0, 1 and 2). (B, C) Molecular profile of medLN hCD4⁺ IgE ASC at various times after cessation of HDM treatment after 15 weeks of 3x weekly HDM challenge, with Tam exposure occurring 48 h before the final HDM i.n. treatment (‘day 0’). Flow cytometry plots are concatenated from 5 (day 2, 7 and 21) or 4 (day 42 and 63) BLTcre.Mcl1^fl/+ mice from one experiment and numbers show percentages and parenthesized numbers show absolute event counts in the plot. (C) Numerated frequencies from B, bars show mean + SD for each phenotype from each group from the individual mice. (D) Surface IgE BCR expression on total IgE_(ic)⁺ events (gated as CD98⁺TDT⁺IgD^loIgG1_(t)^negIgE_(ic)⁺) by phenotype 48 h after the final HDM treatment in the chronic treatment model. Flow cytometry plots show concatenates of n=5 mice from one of three similar experiments 48 h post-final HDM exposure in the chronic treatment model, with symbols representing median fluorescence intensity (MedFI) of the concatenates of three separate experiments. (E, F) Ki-67 in total and IgE_(ic)⁺ ASC 3-weeks after final HDM exposure. Data are shown 21 days after the final HDM instillation, received 3x weekly for 15 weeks, and the flow plot is concatenated from the n=5 mice shown in the graph. Similar data were generated in IgE_(ic+s)⁺ ASC 9-weeks post final Tam exposure from the same experiment. * P < 0.05 by (D) one-way repeated measures ANOVA or (F) two-way repeated measures ANOVA, both with Tukey’s post-test. See also Figure S5.

Surface BCR downmodulation is thought to be a feature of IgG but not IgM or IgA ASC maturation^41,42 and should correlate with the loss of I-A and SLAMF6, but this progression has not been analyzed for IgE ASC. Indeed, there was small-scale down-modulation of IgE BCR among maturing IgE ASC, with IgE_(ic)⁺ events in the I-A⁺SLAMF6⁺ compartment expressing high amounts of IgE_(s), whereas those in the I-A^lo compartments had 50% lower IgE_(s) expression (Figure 4D). In a further analysis, we gated IgD^loCD98⁺TDT⁺ IgE_(ic)⁺ ASC for negativity of all other isotypes, and looked at the retained hCD4⁺ fraction of cells for IgE_(s) BCR expression 114 days after timestamping (Figure S5A and S5B). In this analysis, a small proportion of the IgE_(ic)⁺ ASC were IgE_(s)^neg, and these tended to have lower Igκ_(s) amounts than the IgE_(ic+s)⁺ events, indicating there may be a minority of IgE ASC that lose surface BCR. These findings lead us to conclude that despite some down-modulation, the majority of IgE ASC maintain BCR expression as they age.

In light of a report showing that IgE ASC fail to mature readily past the PB stage in vitro²⁵, there was a possibility that IgE ASC were maintained by continuous or periodic proliferation. In our timestamping system, hCD4 positivity would be inherited by the progeny of timestamped ASC, so to resolve proliferation status, we additionally co-stained for Ki-67, a marker of cell cycle activity. In this Ki-67 analysis, we did not include IgE_(s) staining in the panel and examined the total IgE_(ic)⁺ ASC fraction instead. Three weeks post HDM treatment, and 23 days post-Tam treatment, a fraction of total IgE_(ic)⁺ and hCD4^neg IgE_(ic)⁺ ASC were Ki-67⁺, although in 2/5 mice, Ki-67⁺ IgE_(ic)⁺ cells were undetected (Figure 4E and 4F). In contrast, hCD4⁺ ASC were deficient in Ki-67⁺ cells (Figure 4E and 4F). We conclude that a small number of the IgE ASC in medLN, spleen and BM detected after cessation of HDM treatment become mature PC. We further conclude that the presence of Ki-67⁺ IgE ASC three weeks after challenge cessation is consistent with ongoing formation after HDM treatment is stopped.

A further mark of aged ASC relative to more newly formed ones is expression of high amounts of the CXCL12 receptor, CXCR4⁴³. As we observed a deficiency in the capacity of IgE ASC to seed in BM (Figure 1F) and considering reports of IgE⁺ cells exhibiting poor CXCL12 chemotaxis and an association of IgE BCR expression with altered CXCR4 expression in vitro^15,20, we questioned if CXCR4 expression was altered on IgE ASC in vivo. To test this, we treated a cohort of mice with HDM for seven weeks, gavaged with Tam and analyzed CXCR4 expression amounts on IgE and IgG1 PC three weeks later (schematic in Figure S5C, Group 1). The endpoint, three weeks after Tam gavage was seven half-lives of the short-lived population (Figure 3), such that > 99% of the remaining cells marked with hCD4 could be considered in the long-lived state. MedLN, splenic and BM IgE_(ic+s)⁺ and IgG1_(t)⁺ ASC expressed detectable CXCR4 amounts (Figure 5A – 5C). Overall, there was a general profile wherein the IgG1⁺hCD4⁺ PC had higher average CXCR4 amounts than the IgE hCD4⁺ PC, reaching statistical significance in spleen (Figure 5A–5C and Figure S5D). Also in spleen, hCD4⁺ IgE_(ic+s)⁺ ASC had slightly higher CXCR4 amounts than hCD4^neg IgE_(ic+s)⁺ ASC, consistent with maturation (Figure 5B). However, despite the elevations, the amounts were lower than those on hCD4⁺ IgG1_(t)⁺ PC (Figure 5B), which may in part explain the deficiency in BM seeding of IgE ASC. We conclude that IgE ASC mature as do other long-lived ASC populations, but show deficiencies in CXCR4 expression that may limit their propensity to home to and seed in BM survival niches.

Fig. 5: IgE PC have comparatively low CXCR4 expression.

(A) MedLN, (B) Spleen and (C) BM IgE and IgG1 PC were examined three weeks after Tam gavage and compartmentalized into hCD4⁺ (≥3 wk old) and hCD4^neg (younger than 3 wk) and compared for surface CXCR4 expression. PC were gated as non-autofluorescent live IgD^loCD98⁺ and then IgE_(ic+s)⁺IgG1^neg or IgG1⁺IgE_(ic)^neg for IgE and IgG1 PC respectively. Non-PC were gated as non-autofluorescent live CD98^loTDT^lo. Flow plots are concatenated from the n=4 (medLN) or n=5 (Spleen, BM) mice in which hCD4⁺ IgE ASC were detected in that tissue at endpoint in the experiment, and symbols show the individual mice and lines show group medians from one of two similar experiments (schematic is Group 1 in Figure S5C). Mice in which no hCD4⁺ IgE ASC were detected were excluded from analysis of a given tissue. * P <0.05 by Two-Way repeated measures ANOVA with Tukey’s post-test. See also Figure S5.

IgE PC can die upon BCR ligation¹⁷. In the experiment shown in Figure 5, we included a cohort that received three additional weeks of HDM after timestamping (schematic shown as Figure S5C, Group 2) such that by comparing hCD4⁺ IgE PC abundance in the group with continued exposure we could ascertain what the outcome of re-dosing was on extant IgE PC.

The background representation of hCD4 without Tam exposure in BLTcre.Mcl1^fl/+ mice was ≤1 hCD4⁺ IgE PC event per tissue, translating to a low background detection limit (Figure S5E). Three weeks after timestamping, when mice received ongoing HDM exposure, hCD4⁺ IgE_(ic+s)⁺ PC counts, both in medLN and combined across medLN, spleen and BM, were similar to mice in which HDM exposure was ceased (Figure S5E). This indicated that the extant IgE ASC were unaffected by the ongoing HDM exposure, although the reasons for this are unknown. In contrast, more newly formed (hCD4^neg) IgE_(ic+s)⁺ ASC were significantly more abundant in medLNs of mice that received continued exposure (Figure S5E). We conclude that HDM exposure through the airway promotes IgE ASC production in the local draining LN without altering the abundance of extant IgE ASC.

Unlike other ASC, persistent IgE ASC are more reliant on BCL2, BCL_XL and BCL_W than on MCL1

In our BLTcre hCD4 reporter system, activation of Cre recombinase brings a downstream-inserted hCD4 gene under control of the Mcl1 promoter⁴⁴. The hCD4 expression thus indicates Mcl1 promoter activity. We found that long-lived IgE_(ic+s)⁺ ASC expressed lower amounts of hCD4 than did long-lived IgG1_(t)⁺ ASC (Figure 6A). With MCL1 being a key survival protein for PC³⁰, this raised the question of whether IgE ASC may be constrained in survival because they have an imbalance in MCL1:pro-apoptotic proteins. Alternatively, low transcription of one survival protein may have indicated stronger reliance on other survival proteins. Therefore, we asked if long-lived IgE ASC relied on MCL1 for survival or not.

Fig. 6: Sensitivity of ASC to ABT263 inhibition and MCL1 loss.

(A) Surface hCD4 as an indicator of Mcl1 promoter activity in BLTcre.Mcl1^fl/+ IgE and IgG1 ASC was compared 3-weeks post Tam treatment. The flow cytometry plot shows concatenated medLN cells from n=5 BLTcre.Mcl1^fl/+ mice, with individual values depicted in the line graph. Similar differentials were confirmed at 3-weeks post-Tam in a second experiment. IgE and IgG1 PC were gated as CD98⁺TDT⁺IgD^loIgG1_(t)^negIgE_(ic+s)⁺ and CD98⁺TDT⁺IgD^loIgG1_(t)⁺IgE_(ic)^neg respectively. (B) Model schematic for C-F. (C) Total hCD4⁺ ASC, (D) hCD4⁺ IgG1 ASC, (E) hCD4⁺ IgE ASC number and (F) IgE representation among hCD4+ ASC found in mice across medLN, spleens and BM. Data shown (C-F) are combined from two experiments with n=4–7 mice per treatment group in each experiment. * P <0.05 by one-way ANOVA with Tukey’s post-test. The dashed line in E represents the assay background. See also Figure S5.

To this end, we treated BLTcre.Mcl1^fl/+ and BLTcre.Mcl1^fl/fl mice with HDM for an extended period, Tam exposed the mice, and 16 days later treated them with ABT-263⁴⁵, an inhibitor with sub-nanomolar affinities for BCL2, BCL_XL and BCL_W, for five days (Figure 6B). In a previously presented experiment (Figure S3I), we demonstrated that Tam exposure in BLTcre.Mcl1^fl/fl mice did not reduce T follicular helper cell abundance, suggesting that effects of MCL1 deletion in this system were restricted to ASC.

For the survival of the total persisting hCD4⁺ ASC from spleen, medLN and femurs, there were clear effects of ABT-263, of dual-allele Mcl1 deletion, and of both treatments together (Figure 6C). When looking at specific isotypes, hCD4⁺ ASC of both the IgG1 and IgE isotypes showed sensitivity to ABT-263 (Figure 6D and 6E). Dual-allele Mcl1 deletion also affected IgG1 ASC persistence, but not that of IgE ASC (Figure 6E). Consistent with comparative resistance of IgE ASC to MCL1 loss, representation of IgE among the total detected hCD4⁺ ASC was substantially increased in the BLTcre.Mcl1^fl/fl mice relative to BLTcre.Mcl1^fl/+ when treated with vehicle (Figure 6F), but ABT-263 treatment of BLTcre.Mcl1^fl/fl animals significantly diminished IgE ASC representation (Figure 6F). Perhaps surprisingly, total serum IgE amounts were undiminished by ABT-263 (Figure S5F), a result we interpret as the ongoing production of IgE ASC being uninhibited by ABT-263 treatment, due either to an MCL1-dominated program early in formation, or the duration of treatment being insufficient for newly-formed IgE ASC to die. These data confirm MCL1 as key in the survival of most ASC³⁰ and show that persistent IgE ASC are more heavily dependent on ABT-263-sensitive survival proteins (BCL2, BCL_XL, BCL_W) than on MCL1.

DISCUSSION

IgE ASC accrue in femoral BM and on that basis are inferred to be long-lived⁹. Importantly, however, most BM ASC turn over every few weeks^11,12, so BM residence does not necessarily equate to longevity. Ascertaining IgE lifespans with indirect methods such as CD20 treatment to shut down de novo ASC production is also difficult, given that around 25% of germinal center B cells persist after treatment⁴⁶ and may therefore contribute to ongoing IgE production. Transcriptomics of BM-resident IgE ASC indicate that a subset of them may become long-lived¹⁰, but this is a correlate of longevity only. Even with anti-IL-4Rα treatment, it is hard to know that the signaling promoting IgE ASC production is completely inhibited, meaning clinical studies with dupilumab, while inferential, leave open the possibility of ongoing production of short-lived IgE ASC at low rates. Here, we establish with timestamping in an unequivocal manner that a small proportion of IgE ASC are retained for several months after allergen provocation has ceased yet show a diminished capacity to home to BM and a marked retention in SLTs.

Consistent with two earlier reports^9,10, we observed elevated numbers of IgE ASC in BM of HDM-treated mice, however, at extremely low abundance. Further, different from these earlier studies we found BM IgE ASC abundances were not dramatically different after 7 and 15 weeks of treatment in terms of number, accrual and as an important extension here, persistence dynamics. We were surprised by the paucity of IgE ASC in BM samples relative to the other tissues, given the report of the cells becoming more abundant with additional weeks of challenge⁹—even long times after cessation of HDM, the IgE ASC were more abundant in the other tissues than in the femoral BM where they peaked at 10-fold lower densities than observed in medLN or spleen. That is, in the mouse lines we applied in this work (Verigem, C57BL/6, BLTcre), IgE ASC did not abound in BM. We also note that only tens of cells are found in femoral BM by 15 weeks of HDM challenge in work by another group¹⁰, consistent with our findings. It is worth noting that femurs contain less than 20% of the total marrow of a mouse⁴⁷, meaning additional IgE ASC likely disperse in BM around the body, but practically speaking the extended weeks of HDM treatment were not sufficient to generate a reliable and large IgE ASC population in the femoral BM in a variety of mouse lines. Ongoing challenge via the airway also had no demonstrable impact on established IgE PC in the medLN, suggesting the cells in that tissue were incognizant of allergens upon airway re-exposure. This does not mean that long-lived IgE ASC were resistant to BCR-mediated depletion. Rather, the availability and configuration of antigen during an ongoing immune response may have prevented effects on ASC. For example, it seems likely that the abundant presence of antigen-specific antibody derived from ASC reduced antigen accessibility ⁴⁸.

It is intriguing that a subset of the IgE ASC persisted, so presumably contributed to long-term IgE production. The existence of persistent IgE ASC runs against some of the assumed biology of IgE-expressing cells, as IgE-expressing ASC are immature and limited in survival in culture relative to IgG1 ASC²⁵, and largely in an immature state in blood and airway mucosa^26,27, all implying they should be short-lived. Further, IgE-expressing cells have limited BM tropism as reflected by CXCL12 chemotaxis ^15,20. Mechanistically, while it would be logical that the low CXCR4 amounts on long-lived IgE PC observed are the reason for this deficiency and their low capacity to seed in BM, ascribing the defect entirely to CXCR4 is probably an oversimplification of the situation; thus far unknown mediators of ASC chemotaxis and adhesion in niches may also be impacted on IgE ASC, reducing their BM engraftment. However, these findings all suggest that IgE ASC are less likely to end up in sites that support long-term ASC survival than IgG ASC and therefore biased toward a short-lived-dominant population. Small numbers of IgE ASC did persist here, but predominantly in the SLT. The unique programming of IgE ASC means their deficient ability to accrue in BM may be an evolutionary mechanism that limits long-lived IgE responses. While the majority of IgE ASC are short-lived, the small long-lived fraction identified pairs well with the finding that a minority of IgE ASC are long-lived as reported by measures of serum IgE titers against specific allergens lasting for years after clinical IL-4Rα blockade begins^6,8. We think there is now strong evidence for the development and retention of rare long-lived IgE ASC, although they exhibit distinct tissue dispersion from ASC of other, conventionally long-lived isotypes that reside in BM in abundance. Given the evolutionary split of IgE from IgY many millions of years ago⁴⁹, we think that probably the biology observed is conserved into humans, so likely the findings are applicable to human IgE-mediated diseases such as allergies.

Important for understanding allergy etiology, we found that IgE ASC are continually produced in the period after cessation of allergen exposure which may parallel the maintenance of IgE titers in periods of allergen avoidance as seen in humans^2,7. What is harder to elucidate, and we could not resolve with the modelling, is whether the revealed long-lived IgE ASC decay slowly or are effectively immortal. While in mice the difference is unmeaningful, as a cellular half-life of 88 days would mean some cells are likely to outlive the animal, for humans, a slow decline would be important relative to cellular immortality. Dupilumab studies suggest that humans typically retain some IgE ASC for several months in the presumed absence of de novo IgE ASC differentiation, and even over a three-year timeframe on treatment IgE titers can remain above background^6,8. Yet, in other clinical work, even without dupilumab treatment IgE titers wax and wane over pollen seasons⁵⁰, suggesting a predominant short-lived IgE ASC population. Overall, we suggest that IgE ASC can persist but are unlikely to have the typical lengthy lifespans of ASC of other isotypes and recognizing certain viruses⁵¹. The presence of a population of IgE ASC that survive only for intermediate periods would help explain why allergen immunotherapy regimens typically take 3–5 years to demonstrate efficacy. This timeframe would allow for turnover of the majority of IgE ASC, while the immune system is geared toward tolerance with limited further IgE ASC production in the period of treatment once successful. The potential for long-lived IgE ASC underscores the importance of finding ways to actively target the IgE ASC that fall at the latter part of the survival distribution. Targeting the survival proteins expressed by IgE ASC, as revealed here, may offer a means to eliminate these cells in disease.

We note that we have a clearer understanding of the PC survival structure since our last work on this topic¹². There, the double-exponential and lognormal steady state models both produce satisfying and interpretable fits to the data. In that data set, although the double-exponential fits are slightly better by AIC, they do not reach the conventional threshold of 10 for significance and the steady-state log-normal offers the simplest explanation that fits with the biology, such as evidence for continuous maturation of surface phenotype over time. Here, we have generated more granular data across four compartments (IgG1 and IgE ASC in SLT and BM). The double-exponential fits were convincing and consistent across compartments, whereas the lognormal steady-state fits appeared ‘unphysiological’ in the sense that they were highly skewed with standard deviation far in excess of mean and poorly-constrained, particularly in the bone marrow compartments. Thus, we concluded that of these two models, the double-exponential was superior for fitting this data set. Likely, however, the biology is more complicated than this—our data does not preclude models with more complex compartment structures, such as those fine-tuning the half-lives within two broader compartments (short- and long-lived) depending on cell state, history or other internal or external factors such as antigen specificity^40,52. Here the modelling clearly demonstrated longevity of IgE and IgG1 ASC and ascribed confines to what constituted long-lived. However, the bigger question of the relationship between the apparent lifespan of a PC and its phenotype remains the subject of ongoing work.

An important finding in our work was the Navitoclax (ABT-263)-sensitive survival program of IgE ASC. The long-lived IgE ASC had low Mcl1 promoter activity and were accordingly resistant to dual allele Mcl1 deletion, but highly susceptible to ABT-263. That is, for IgE ASC, BCL2, BCL_XL, or BCL_W were more important than MCL1, whereas for other PC MCL1 dominated over BCL2, BCL_XL, and BCL_W. Notably, a subset of IgE ASC exhibit high expression of Bcl2¹⁰, so we speculate that this is the primary ABT263-sensitive protein that supports persistent IgE ASC. Consistent with this, BCL2 over-expression makes minimal difference to lymphoid IgG1 PC survival, but increases lymphoid IgE PC ≈30-fold; the same also increases IgE PC density in BM from 0% to 1.5%, meaning BCL2 alone is sufficient to rescue survival of IgE PC that are otherwise destined for death, fitting with a dominant short-lived IgE PC population even in the BM²³, akin to outcomes of our modelling. While the IgE ASC were less reliant on MCL1 in our study here, ABT263-sensitivity was also evidenced for the long-lived IgG1 ASC to an unexpected degree, suggesting in those cells MCL1 is one of multiple proteins that keep long-lived ASC alive non-redundantly, although MCL1 remains dominant for non-IgE PC. For the IgE ASC, the low Mcl1 promoter activity may have been intrinsic as an outcome driven by IgE BCR expression, or, it might have reflected an inability to deposit as effectively in ‘survival niches’ and receive signals that drive Mcl1 expression to prohibit apoptosis, such as BCMA signaling by APRIL and BAFF binding, or stromal anchoring^30,53. Perhaps this was related to the low expression of CXCR4 observed from established cells of the isotype, at least in spleen, where survival niches also exist³⁹. Given that ABT-263 was able to eliminate the majority of long-lived IgE ASC, such treatment may thus be expected to kill IgE ASC in clinical allergy and related IgE-mediated diseases. Titrating the dose of ABT-263 down may reveal a window in which IgE ASC can be killed without deleteriously affecting the populations that have MCL1 reliance, and thus the ‘protective’ long-lived cells recognizing pathogens. This may be helpful for individuals suffering from allergy and hard-to-control asthma, although the use of BCL2 family inhibitors has side-effects that would need to be balanced against any benefit of removing IgE ASC.

A final point we highlight is that while IgE ASC can persist for weeks, a substantial proportion were produced in the period after cessation of HDM treatment. This was exemplified in the low fraction of hCD4⁺ cells among total IgE ASC and the continuing expression of Ki-67 among hCD4^neg IgE ASC after Tam exposure, in the absence of continuing HDM treatment. This means that targeting both IgE ASC survival and genesis pathways will probably be required to abrogate IgE production. Certainly, our inability to modulate IgE amounts acutely with ABT-263 suggests ongoing IgE ASC production is important. A recent advance shows that targeting IgE PC with a bispecific antibody against the PC survival-protein capture receptor BCMA combined with IL-4Rα-blocking antibody treatment diminished IgE reactivity, although had the limitation of killing other ASC as well⁵⁴. Given that most IgE ASC are short-lived cells, this approach of targeting genesis in combination with an IgE ASC-depleting therapy is an attractive means of treating severe IgE-mediated diseases. Based on our results here, we speculate that directing BH3 mimetics to the IgE BCR may be a way to deplete long-lived IgE ASC without disrupting protective antibody-mediated immunity.

Limitations:

This is a mouse study of IgE ASC longevity, in which we found most IgE ASC were continually formed. We acknowledge that the relative abundances of continually forming, short-lived IgE ASC and their long-lived counterparts may differ by individual in allergic diseases, and this may cause variation in the benefit of dupilumab. However, our proposal that both continuing production and retention of longer-lived IgE ASC must be targeted for disease attenuation remains relevant. Second, IgE is not the sole mediator of allergic and atopic diseases, so targeting the IgE component may not resolve symptoms of disease in all cases. Given the lower MCL1 expression in the IgE PC, our ability to resolve timestamped IgE PC may be reduced. While this does not change the major conclusions, it means we may underestimate the size of the long-lived IgE PC compartment. The expression of BLIMP1, even at low amounts, may have led to tamoxifen induced effects on tissue-resident effector T cell function or abundance following deletion of Mcl1, which may impact the formation of IgE ASC in tissues indirectly. We do not consider this to be a substantial risk because only a small proportion of IgE ASC were found in the lung tissue; the majority were in the secondary lymphoid tissues. Last, confirmation of the survival program difference in IgE ASC needs to be verified in human cells for that pathway to be a realistic target for removing IgE ASC in disease.

Resource availability:

Lead Contact:

Please direct requests for further information or resources and reagents to Marcus J. Robinson (marcus.robinson@monash.edu).

Materials Availability:

BLTcre mice are available on request subject to a Monash University material transfer agreement or use license.

Data and code availability:

All data reported in this paper are contained in the data supplement file Raw_data.xlsx or will be shared by the lead contact upon request. All original code has been deposited at GitHub and is publicly available at DOI: 10.5281/zenodo.17284584 as of the date of publication.

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR METHODS

Experimental Model and Subject Details

C57BL/6, BLTcre.Mcl1^{+/+ 34}, BLTcre.Mcl1^fl/+ (BLTcre hCD4 reporter) and BLTcre.Mcl1^fl/fl mice³⁵ were bred and maintained in specific pathogen-free facilities of the Monash University Animal Research Platform, Melbourne, Australia. Verigem (Igh-7^tm1.2Cdca)²³ mice were bred to Jh null mice (Igh-J^tm1Dhu)⁵⁵ to ensure that VDJ recombination was on the same chromosome as the IgE reporter and were maintained on a C57BL/6 background at the University of California, San Francisco (UCSF). All experiments were approved by the Alfred Research Alliance Animal Ethics Committee A, Melbourne, Australia or the UCSF Institutional Animal Care and Use Committee. Mice were eight weeks or older (i.e. adult) at the onset of the experiments, and were sex- and age-matched across timepoints and both female and male mice were used. The studies were underpowered to test if gender impacted IgE ASC persistence so such tests were not performed. The conclusions of the study are unimpacted by the absence of such tests. Unwell and runted animals were excluded from experiments and animals received no interventions other than those involved in the experiments. Researchers were not blinded to treatments or genotypes. The experimental unit throughout was the individual mouse tissue or the summed total ASC count across spleen, medLN and femur pair as indicated in Figures.

Method Details

Intranasal HDM administration: Mice were administered intranasally three times per week for up to 15 weeks as specified with 50 μg HDM (Stallergenes Greer, Lenoir, NC, USA) diluted in 20 μL PBS. Mice treated with PBS, or of the BLTcre.Mcl1^+/+ genotype were used as controls.

Tamoxifen administration: Mice were gavaged at 200 mg/kg with a 60 mg/mL solution of Tamoxifen (Sigma-Aldrich, Saint Louis, MO, USA) in 90% v/v peanut oil, 10% ethanol. ABT-263 treatment: Mice were treated daily for five days perorally as depicted (Figure 6B) with 100 mg/kg ABT-263 dissolved by sonication in 60% Phosphal-50 PG (MedChem Express, Monmouth Junction, NJ, USA), 30% PEG-400 (MedChem Express), 10% ethanol (Sigma-Aldrich) according to the manufacturer’s instructions, or with vehicle only.

Tissue harvesting and cell isolation: For C57BL/6 work, lungs were dissected out after removing bronchoalveolar lavage fluid with two times of 0.5 mL PBS, then sliced and diced into fine pieces with a drop of Hank’s Balanced Salt Solution (HBSS) (Thermo Fisher Scientific, Waltham, MA, USA). Tissue pieces were then digested in 5 mL digestion buffer (0.1 mg/mL Liberase [Roche, Basel, Switzerland] and 40 μg/mL DNase I [Roche] in HBSS) at 37°C on an incubator shaker for 30 min. Digested tissues were collected into C tubes (Miltenyi Biotec, Bergisch Gladbach, Germany) and dissociated on a GentleMACS tissue dissociator (Miltenyi Biotec) for 37 s at 2079 rpm at room temperature. Digested lung samples were topped up with 10 mL flow buffer (0.5% bovine serum albumin and 2 mM EDTA [Sigma-Aldrich] in PBS) and filtered through 70 μm cell strainers. After adding another 30 mL flow buffer, samples were pelleted and treated with red blood cell lysing buffer (156 mM NH₄Cl, 11.9 mM NaHCO₃, 97 μM EDTA, pH = 7.3) before washing in flow buffer. Lung samples for Verigem mouse studies were as previous⁵⁶, briefly, excised lungs were digested with 25 μg DNAse I (Sigma) and 0.28 Wunitz Units of Liberase TM (Roche) in 1 mL before being sliced and diced, incubated for 30 to 45 minutes at 37 °C with shaking at 700 rpm. The reaction was stopped by topping up to 10 mM EDTA and 5% FBS. The resulting cell suspension was dispersed and passed through a 70 μm strainer for preparation for staining. NALT, which is located on the posterior side of the palate, was isolated similar to a described method⁵⁷, namely by first cutting off the tip of the nose and the lower jaw, then peeling away the palate, then trimming the palate, and then placing the tissue in PBS. NALT was dissociated gently from the palate in a drop of PBS using frosted glass slides. LNs were isolated by blunt dissection and pressed through 70 μm cell strainers to obtain single cell suspensions in flow buffer. Spleens were excised by blunt dissection then pressed through a 70 μm cell strainer followed by red blood cell lysis. Femurs were crushed in a mortar and pestle in 3–4 mL per femur pair and BM cells were collected in flow buffer with the addition of 0.09% NaN₃, followed by red blood cell lysis. All cell suspensions were pelleted, resuspended in flow buffer, and filtered through a 50 μm nylon mesh before counting and staining.

Flow cytometry: Reliable identification of IgE-expressing ASC is difficult as the cells are rare and populations such as naïve B cells and basophils capture IgE and can be mistaken for IgE-expressing cells¹³. To minimized these issues, IgE ASC staining was performed with IgE ASC gating based on ic IgE stained in two separate fluorescent colors in the 15-week model, and IgE_BV421 in the 7-week model^23,58. To identify IgE ASC, we blocked IgE_(s) bound to IgE receptors (FcεRI, FcεRII, FcγRIV) using an unlabeled anti-IgE monoclonal antibody (mAb), and then detected ic IgE with a fluorophore-tagged version of the same mAb. Because ic IgE is abundant in IgE ASC but lower in populations internalizing IgE through endocytosis, this facilitates strong enrichment for IgE ASC²³. To increase further the stringency of IgE ASC identification, we excluded IgE_(s) negative cells, as these generally expressed high amounts of other isotypes (Figure S2D) or stained for surface Igκ while lacking IgE(s) (Figure S2E), incongruent with IgE ASC identity. The same strategy was applied to identify IgE ASC in the BM (Figure S2H–L). For the staining procedure, cells were transferred to 5 mL polystyrene tubes (Techno Plas, Saint Marys, SA, Australia) and incubated with purified anti-FcγRII/III (clone 2.4G2) antibody at 1.25 μg/mL and stained with purified anti-IgE along with combinations of fluorophore-labelled antibodies and molecules outlined in Table S2. Surface R35–118-bt anti-IgE staining was performed before purified anti-IgE clone R35–72 used for surface blocking. Cells were subsequently fixed and permeabilized for cytoplasmic antibody staining with the Cytofix/Cytoperm (BD Biosciences, Franklin Lakes, NJ, USA) reagent set according to the manufacturer’s instructions. For Verigem mouse analysis, cells were surface stained using PE-labeled RME-1 anti-IgE. Live cells were identified by eBioscience Fixable Viability dye efluor780 (Life Technologies, Carlsbad, CA, USA) stained during surface antibody staining. Where no events were found in a gate (typical for hCD4⁺ IgE events in BLTcre.Mcl1^+/+ and ungavaged BLTcre.Mcl1^fl/+ tissues), quantitation of the background was set as 1/live event count multiplied by the tissue cell count. The data were analyzed using FlowJo software (Treestar, Inc., San Carlos, CA, USA).

Enzyme linked immunosorbent assay (ELISA): Mice were euthanized and blood samples collected by vena cava bleed using 25–27G syringes. Blood samples were centrifuged at 13000 g for 5 min and sera were removed and stored at −30°C. Total IgG1 and IgE ELISA were then performed. For this, 96 well EIA/RIA plates (Corning, Glendale, AZ, USA) were coated with 2 μg/mL 23G2 rat anti-mouse IgE (Southern Biotech, Birmingham, AL, USA) or rat anti-mouse IgG1 (Southern Biotech) in PBS and incubated overnight at 4 °C. The next day, wells were washed, and coated with a blocking solution containing 1% FBS, and 0.6% w/v skim milk powder. Thereafter, the blocking solution was removed and samples at 1/2, 1/10, 1/50, 1/250 and 1/1250 dilutions, or standards (mouse IgE anti-DNP clone SPE-7, starting concentration 1 μg/mL or mouse IgG1 clone MOPC31c starting at 200 ng/mL) were added to the wells. After 2 hours of incubation, plates were washed and anti-IgE-HRP (Southern Biotech) or anti-IgG1-HRP (Southern Biotech) was added to the wells for 1 hour at room temperature. Detection used the one-step ULTRA TMB-ELISA Buffer reagent (Thermo Fisher Scientific), prewarmed to room temperature and the reaction was quenched with 2 M HCl. Absorbance was measured at 450 nm and optical densities compared to the standard curve to determine concentrations. For HDM IgG1 ELISA, plates were instead coated with 10 μg/mL HDM.

Basophil activation test (BAT): 2×10⁶ splenocytes were added to 5 mL tubes and incubated for 10 minutes at 37 °C and then HDM or PBS added to the tubes to stimulate the cells. Samples were incubated at 37 °C for 20 minutes and then inactivated by incubating on ice for 5 minutes. Cells were then washed and handled at 4 °C to prevent further activation during staining and acquisition. BAT activation was read out by CD63 expression on the basophils.

Quantification and Statistical Analysis

Statistical comparisons were performed in GraphPad (San Diego, CA, USA) Prism software. Tests used are stated in figure legends and outcomes with significant P-values are shown in the attached raw data file. Statistical significance was determined as P ≤ 0.05. Group size was based on estimates from prior works that tissue ASC distributions are log-normal with the geometric SD factor being ≈50% of the geometric mean with power designed at 1-β=0.8 and α=0.05 to reveal an approximate 40% difference (effect size) between groups or timepoints in single group comparisons.

Exclusion Criteria: Seven data points were excluded in the study. Two mice (both Mcl1^fl/+) in Figure 6 were excluded where IgG1⁺IgD^negCD98^loTdT^lo comprised <0.5% of the total live non-autofluorescent medLN events, indicating aberrant responses in the mice. Three medLN samples (one Mcl1^fl/+ vehicle, one Mcl1^fl/+ ABT-263, one Mcl1^fl/fl ABT-263) from Figure 6 were also excluded as <2.5% of the total cell sample was able to be acquired, indicative of a processing error, and one medLN (Mcl1^fl/+ vehicle) and one spleen sample (Mcl1^fl/fl ABT-263) from mice in Figure 6 where the tubes cracked leading to sample loss during centrifugation precluding accurate cell count determination. When a tissue was lost to analysis, mice were excluded from total mouse ASC comparisons. All other data points were included in analyses.

Mathematical modelling: Three mathematical models were used to fit to the hCD4⁺ ASC data. Here $y\left(t\right)$ denotes the absolute number of cells retaining the hCD4 label at time, $t$, and $y_0$ is the starting population at time 0 (taken as the first measured time point at 4 days after Tam gavage, to allow establishment of the hCD4 label).

1. Single-exponential model

   $$y\left(t\right)=y_0{e}^{-\mathrm{l}\mathrm{n}\left(2\right)\cdot t/t_1},$$

   where $t_1$ is the half-life, and $ln\left(\right)$ is the natural logarithm.
2. Double-exponential model

   $$y\left(t\right)=y_0(\left(1-f_{LL}\right)e^{-\ln \left(2\right).\frac{t}{t_1}}+f_{LL}{e}^{-\ln \left(2\right).\frac{t}{t_2}}),$$

   where $t_1$ and $t_2$ are the half-lives of the short- and long-lived subpopulations, respectively, $t_1\le t_2$, and $f_{LL}$ is the fraction long-lived.
3. Lognormal steady-state model

   $$y(t)=\frac{y_0}{2}\cdot \left(1-e^{-\left(m+\frac{s^2}{2}\right)}\cdot t\cdot \left(1-erf\left(\frac{\mathrm{l}\mathrm{o}\mathrm{g}(t)-m}{\sqrt{2}s}\right)\right)-erf\left(\frac{\mathrm{l}\mathrm{o}\mathrm{g}(t)-m-s^2}{\sqrt{2}s}\right)\right),$$

   where $m$ and $s$ are the mean and standard deviation of the underlying normal distribution in log space, and $erf()$ is the error function⁵⁹. The $m$ and $s$ parameters were transformed to the mean, $\mu$, and standard deviation, $\sigma$, in linear space using the formulae: $\mu =e^{m+s^2/2}$ and $\sigma =\sqrt{e^{2\left(m+s^2\right)}-e^{2m+s^2}}$

Model fitting was performed in Matlab using the optimisation toolbox, including the lsqcurvefit() function with the Levenberg-Marqardt algorithm and the sum of square errors (s.s.e.) as the objective function. The MultiStart function was used to obtain optimal solutions from 100 randomly chosen starting points, and the best selected as the global optimum, to reduce the chance of returning a local optimum. A bootstrapping technique was used to estimate confidence intervals on the parameter values. The empirical mean and standard deviation of the experimental data were calculated for each data point. For each bootstrap iteration, new data points were calculated by resampling from a gaussian distribution with the corresponding mean and standard error in the mean at each time point. The model was then refit to this resampled data, to generate empirical distributions of the fitted model parameters, which were then used to define empirical 95% confidence intervals (C.I.). Final iterations of fitting were performed to all four compartments (IgG1 and IgE, SLT and BM), with parameters allowed to vary independently or fixed to be equal between compartments, as indicated in Table S1. From the combined fits with fixed half-lives across compartments but variable fraction long-lived in the double-exponential model we derived the following analytical formula for the transition time:

$$t_{\mathit{trans}}=\frac{\mathrm{l}\mathrm{n}\left(\frac{(1-f)k_1}{f{k}_2}\right)}{k_1-k_2}$$

where the decay rates, $k_1$ and $k_2$, are related to the two half-lives, $t_1$ and $t_2$, by the formula $k=\mathrm{l}\mathrm{n}(2)⁄t$.

The Akaike Information Criteria (AIC) was used to compare goodness-of-fit between the models, considering the trade-off between number of parameters and sum-of-square errors (s.s.e.). By convention, a difference in AIC of greater than 10 (ΔAIC>10) is considered evidence to favor one model over another.

Supplementary Material

Supplemental Information: Document S1. Figures S1–S5, Table S1 and Table S2. Raw data file Raw_data.xlsx.

Supplemental tables not included in the main supplemental PDF

Supplemental Raw data File Raw_data.xlsx, related to Figure 1–6, S1, S3 and S5.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
R35-72 anti-IgE	BD Biosciences	Cat#553413; RRID: AB_394846
R35-72 anti-IgE-BV421	BD Biosciences	Cat#564207; RRID: AB_2738668
R35-118 anti-IgE biotin	BD Biosciences	Cat#553419; RRID: AB_394850
X56 Anti-IgG1-BV605	BD Biosciences	Cat#742477; RRID: AB_2740811
A85-1 anti-IgG1-FITC	BD Biosciences	Cat#553443; RRID: AB_394862
A161A1 anti-human CD4	Biolegend	Cat#357418; RRID: AB_2616933
H202-14 anti-CD98-BV711	BD Biosciences	Cat#745466; RRID: AB_2743009
Biological samples
Dermatophagoides pteronyssinus crushed house dust mite lyophilized extract	Stallergenes-Greer	Cat#XPB91D3A2.5
Chemicals, peptides, and recombinant proteins
Tamoxifen	Sigma	Cat#T5648-1G
Fixable viability dye eFluor780	ThermoFisher Scientific	Cat#65-0865-18
Critical commercial assays
Cytofix/Cytoperm Soln Set	BD Biosciences	Cat#554714; RRID: AB_2869008
Deposited data
Modelling code	This paper	DOI: 10.5281/zenodo.17284584
Experimental models: Organisms/strains
C57BL/6	Monash Animal Research Platform	N/A
Verigem	Yang et al.23	N/A
BLTcre	Di Pietro et al.34	N/A
Mcl1fl/+	Vikstrom et al. 44	N/A
Mcl1fl/fl	Vikstrom et al. 44	N/A
Software and algorithms
Graphpad Prism 10	Graphpad	RRID: SCR_002798
Flowjo	BD Biosciences	RRID: SCR_008520
Excel	Microsoft	RRID: SCR_016137
Illustrator	Adobe	RRID: SCR_010279

Highlights:

• IgE plasma cells can become long-lived but most are short-lived

• Production of IgE plasma cells continues long after the last known exposure

• Low CXCR4 expression is associated with deficient accrual of IgE PC in bone marrow

• IgE plasma cells have a BCL2, BCL_XL, or BCL_W dominated survival program