Microchimerism is strongly correlated with tolerance to noninherited maternal antigens in mice

Immunosuppressive drugs administered to prolong graft survival increase the risk of systemic infections¹  and may encourage tumor growth.^2,3  Taking advantage of natural tolerance induced by noninherited maternal antigens (NIMAs) is one of the more promising but still relatively unexplored approaches for reducing the immunosuppressive burden in organ and stem cell transplant recipients. The clinical benefits of developmentally acquired tolerance to NIMA were first noted by Owen et al⁴  more than 50 years ago. Since then, tolerogenic effects of NIMA have been documented at both T- and B-cell levels in a variety of clinical settings.^5-7 

The basis of the NIMA benefit to allograft survival is not clear. One possible explanation is that many normal babies go on to accept, as adults, a tiny transplant of cells from their mothers acquired during ontogeny and thus are already predisposed to accept a larger NIMA⁺ organ transplant. Although fetal and maternal circulations are completely separated, fetal tissue is bathed with maternal blood in animals with a hemochorial placenta (eg, mouse and human),^8,9  creating opportunities for bidrectional transfer of mature cells as well as hematopoetic and pluripotent progenitors.^10-15  Moreover, rare maternal cells in liver can be acquired through ingestion of colostrum after birth.¹⁵  The low frequency of maternal cells present in adult offspring (< 0.1%) is called “microchimerism” (Mc), a term also applied to rare donor cells that emigrate from graft-to-host tissue after organ transplantation. It has been suggested that Mc, while providing a miniscule antigen “load” to the host, may nonetheless have major immunobiologic significance.²⁰  Others have argued that the presence of rare foreign antigen-bearing cells in host tissues is either “ignored” by the host immune system or exert no additional impact on tolerance or immunity to self- or alloantigens expressed by solid tissues.²¹  However, recent experiments have shown that not only the quantity, but also the quality (multilineage vs unilineage) of chimerism is important determining full versus “split” tolerance.^22,23  In addition, the discovery of the “semi-direct” pathway of alloantigen recognition, alloantigen acquisition by host dendritic cells, has provided an amplification mechanism whereby allogeneic cells sequestered in tissues may exert a strong antigenic impact on the host.²⁴ 

Yet, if rare maternal cells and antigens are present in professional antigen-presenting cells (APCs), sensitization to NIMA and elimination of maternal cells might be expected to occur in all immunocompetent offspring. A recent study¹⁶  showed why this does not happen in the human fetus. Instead of eliciting a dominant T effector (T_E) cell response, maternal alloantigens promoted T regulatory (T_R) cell proliferation in the fetal lymph node (LN) by a TGF-β–dependent mechanism, sparing the maternal cells.¹⁶  However, this study left unresolved the issue of whether the fetal T_R cells, once induced, are long-lived cells that persist in the adult, regardless of maternal microchimerism (MMc) level or distribution, or whether they are short-lived cells that require continuous tolerogenic antigen input to survive.¹⁶ 

To get at this question, we tested the relationship between MMc and NIMA-specific T_R cells in individual offspring using the mouse F₁ backcross breeding model (B6 × BDF₁) originally described by Zhang and Miller.²⁵  We and others have previously reported strong tolerogenic effects of NIMA^d exposure on fully allogeneic heterotopic heart transplantation survival²⁶  and graft-versus-host disease,²⁷  effects mediated in part by NIMA^d-specific CD4⁺ T_R cells.^28,29  Because antigen-specific T_R cells are likely the key to NIMA-induced tolerance, we wished to test the hypothesis that their level in a given individual would depend upon the level and quality of persisting maternal cells. We show here the interdependence of MMc and NIMA-specific T_R cells in adult mice. Furthermore, we show that despite a high level of MMc in multiple organs at birth, maintenance of maternal cell engraftment, and favorable T_R > T_E balance into adulthood was dependent on nursing of the neonate by the semiallogeneic mother.

C57 BL/6 (B6; H-2^b/b), DBA/2 (H-2^d/d), B6D2F₁ (BDF₁;H-2^b/d), and B6C3F₁ (H2^b/k) were purchased from Harlan Sprague Dawley. B6-GFP^+/− mice (C57BL/6-Tg (ACTB-EGFP)1Osb/J;H-2^b/bGFP^+/−) were purchased from The Jackson Laboratory. F1 backcross breeding was performed to obtain offspring developmentally exposed or nonexposed to NIMAs²⁶  (supplemental Figure 1, available on the Blood website; see the Supplemental Materials link at the top of the online article). All offspring were weaned after 21 days of birth and typed using anti-H2K^d antibody (BD Biosciences) and for green fluorescent protein (GFP) by flow cytometry using a FACS Caliber (BD Biosciences). Homozygous H2^b/b offspring (negative for GFP and/or H2K^d) aged 6 to 8 weeks were used for all experiments. All experiments were performed in accordance with National Institutes of Health (NIH) and United States Department of Agriculture (USDA) guidelines, after approval from the University of Wisconsin Institutional Animal Care and Use Committee.

Heart, lungs, liver, brain, blood, pooled LNs (inguinal, popliteal, axillary, brachial, and cervical), bone marrow, thymus, and spleen were collected and teased apart. DNA extraction and amplification were performed as described in supplemental Methods. H2D^d forward primer (CCTTCCCAGAGCCTCCTTCA), H2D^d reverse primer (AGAACTCAGACCCTGCCCTTTAA), and probe (6FAM-TCCACCAAGACTAACACAGTAATCATTGCTGTTCC-BHQ) were purchased from Biotechnology Center, University of Wisconsin–Madison. GFP forward primer (CCACATGAAGCAGCAGGACTT), GFP reverse primer (GGTGCGCTCCTGGACGTA), and GFP probe (6FAM-TTCAAGTCCGCCATGCCCGAA-BHQ), designed by The Jackson Laboratory, were used to detect GFP⁺ maternal cells in NIMA^D NIMA^GFP offspring by quantitative polymerase chain reaction (qPCR). Target DNA (1 μg, equivalent to ∼ 10⁶ cells) for all samples was quantified using the standard curve and expressed as estimated gene equivalents (GEq) per 10⁵ offspring cells. All samples were run in triplicate, and the 3 GEq values for each organ were averaged to get a final GEq per 10⁵ values. More details are available in supplemental Methods.

Spleen cells from NIMA^d-exposed and NIPA^d control mice were tested in adoptive transfer delayed type hypersensitivity (DTH; subcutaneously in B6 footpad) and in vivo mixed lymphocyte reaction (MLR; intravenously in BDF₁ host). Measurement of swelling, lymphoproliferation, and TGF-β expression on CD4⁺ T cells were performed as described previously.²⁸ 

Live (propidium iodide–negative) cells dimly expressing maternal class I and II MHC antigens were quantified in the weaned F1 backcross offspring using PE-labeled anti-H2K^d antibody and FITC-labeled I-E–specific antibody (BD Biosciences) in thymus, spleen, LNs, bone marrow, and blood. B6 and BDF₁ cells were used as negative and positive controls, respectively. In some experiments, cells were also stained for CD4, CD8, CD11b, and CD11c (BD Biosciences), cell surface TGF-β1, and Foxp3 (eBioscience). The data were analyzed using either CellQuest or FlowJo (TreeStar) software.

Spleen and bone marrow cells stained with magnetic bead–conjugated antibodies against CD11c, CD11b, CD4, and CD8 (Miltenyi Biotec) were sorted according to the manufacturer's protocol. Single-cell suspensions of heart tissue were obtained using dispase II and collagenase (Roche) according to the manufacturer's directions. The cardiac cells were stained with magnetic bead–conjugated lineage cocktail antibodies (CD5, CD45R [B220], CD11b, Gr-1 [Ly-6G/C], 7-4, and Ter-119; Miltenyi Biotec).

Data were analyzed using GraphPad Prism 5 software (GraphPad Software). The nonparametric Mann-Whitney test was used to analyze data. Linear regression was used to find correlation coefficients and P values comparing different parameters.

We used 2 different breeding strategies to analyze the tissue distribution and phenotypes of maternal cells in NIMA^d-exposed mice. The first was a standard (BDF₁ × B6) backcross²⁶  (supplemental Figure 1A) from which 50% of the offspring were H2^b/d and 50% were H2^b/b; the latter were developmentally exposed to maternal H2^d (BDF₁ mother, B6 father; NIMA^d). Control breedings (B6 mother, BDF₁ father; NIPA^d control) would have a similar percentage of H2^b/b and H2^b/d mice, but no maternal sources of Mc. Second, we used a breeding strategy outlined in supplemental Figure 1B wherein mothers were heterozygous for both H-2^d and GFP, resulting in 25% of offspring that were H2^b/b and GFP^−/− (NIMA^d NIMA^GFP), which have been exposed to H2^b/d and GFP^+/− maternal cells. The NIPA control offspring (NIPA^d NIPA^GFP) in the latter model were also H2^b/b and GFP^−/− homozygotes with no in utero or oral exposure to maternal H2^d and/or GFP.

To detect rare H2^b/d (maternal) cells in NIMA^d-exposed offspring, we developed a novel qPCR technique to amplify maternal H2D^d DNA (Figure 1A). In the second breeding model, we used another qPCR to detect rare GFP DNA (supplemental Figure 2). As shown, the titration of BDF₁ or BDF₁-GFP → B6 DNA yielded a linear standard curve sensitive to as little as 0.1 gene equivalents per 100 000 cells (10⁻⁶). Analysis of NIPA^d control mice showed that the transfer of GFP⁺ or H2^d+ cells from siblings or father, either directly or indirectly via mother, to H2 ^b/b offspring is relatively inefficient—82% (13 of 16) of NIPA controls had no detectable Mc in any of the organs tested (Figure 1B). Mc was detected exclusively in heart samples in 3 of 16 NIPA control mice, at the level of 1 Geq per 10⁵ cells (Figure 1B-C). However, this observation was not reproduced in NIPA^d NIPA^GFP control mice using the GFP qPCR method (0 of 6; Figure 1C). In contrast, Mc was frequently detected in multiple organs of NIMA-exposed mice (Figure 1B) at levels ranging from 0.1 to 50.0 Geq per 10⁵ cells (Figure 1C). Specifically, we found that 61% (25 of 41) of the NIMA^d-exposed mice had detectable levels of Mc in at least 1 organ, with most having Mc in 2 or more organs (Figure 1B). When an average Mc level in 8 or 9 organs was calculated, Mc was significantly higher in NIMA versus NIPA offspring (1.318 ± 2.49 GEq/10⁵ vs 0.042 ± 0.07 GEq/10⁵; P = .001 for H2D^d; 0.327 ± 0.511 GEq/10⁵ vs 0.0 ± 0.0 GEq/10⁵; P = .04 for GFP). Thus, one may conclude that the majority of the Mc detected in the NIMA offspring was due to maternal, not sibling, sources. We will therefore refer to the presence of H2D^d and GFP DNA in NIMA^d-exposed offspring as MMc.

The distribution of MMc in offspring was somewhat surprising. A total of 3 nonlymphoid organs (heart, liver, and lung) exhibited the highest levels (Figure 1C) and incidences (Figure 1D) of MMc, with heart showing the presence of MMc most frequently, followed by liver and lungs. Brain tissue was rarely MMc⁺. Among the lymphoid compartments, peripheral blood frequently (5 of 12; 43%) exhibited MMc. However, MMc was rarely detected in total DNA isolated from LNs, bone marrow, thymus, or spleen (Figure 1D). Widespread distribution of MMc, with a similar tissue bias, was confirmed using the GFP qPCR assay (supplemental Figure 2) in NIMA^d NIMA^GFP-exposed mice (Figure 1C-D).

LN and spleen have recently been found to be key staging grounds for development of NIMA-specific T_R cells in the human fetus and were also sites with significant MMc.¹⁶  We were therefore surprised to find that we could rarely detect MMc in unfractionated LN, bone marrow, spleen, and thymus cells (Figure 1D). To determine the incidence of MMc in isolated leukocyte subsets, we fractionated cells from spleen and bone marrow of NIMA^d-exposed mice (supplemental Figure 3) and tested them for MMc. We found MMc in splenic CD11c⁺ dendritic cells (DCs), CD4⁺ (but not CD8⁺) lymphocytes, and CD11b⁺ bone marrow cell subsets (Figure 2A). These results demonstrate the greater sensitivity of subset analysis, revealing a 50% (3 of 6) MMc incidence compared with a 0 of 6 incidence of MMc in unsorted splenocytes and bone marrow cells. Importantly, these data suggest a type of MMc that includes professional class II⁺ APC.

The heart was the organ that exhibited the highest frequency of MMc (Figure 1D). To determine whether the MMc in heart was hematopoietic or parenchymal origin, we sorted cell populations obtained from 8 pooled heart tissues of NIMA^d-exposed mice. We found a very strong H2D^d signal in DNA from hematopoietic lineage cells (lin⁺) in the heart compared with a 100-fold weaker signal in the nonhematopoietic cells (lin⁻), which would include cardiac myocytes, endothelium, and smooth muscle cells (Figure 2B).

Because heart and liver were most consistently MMc⁺, we investigated whether the maternal cells were detectable in these organs by immunohistochemistry. We detected GFP⁺ maternal cells in livers and hearts of NIMA^d NIMA^GFP-exposed offspring (supplemental Figure 4A-B), but not in NIPA^d NIPA^GFP control mice (supplemental Figure 4C-D) using an anti-GFP antibody. Consistent with the identification of lin⁺ cells in the heart as the predominant MMc type, the GFP⁺ cells were scattered within the heart tissue, as has been described for cardiac tissue macrophages.³⁰  The thymus of a NIMA^dNIMA^GFP-exposed mouse, an organ consistently low or negative for MMc by qPCR (Figure 1), was found to contain rare GFP⁺ cells. The thymic GFP⁺ cells costained with anticytokeratin antibodies, indicating maternally derived thymic epithelial cells (supplemental Figure 4E).

To detect NIMA-specific T_R cells, we used DTH bystander suppression^31-34  and “in vivo MLR” assays.²⁸  In the first, we measured the ability of spleen cells from tetanus toxoid (TT)–immunized mice to mediate bystander suppression of a recall response to TT in the presence of noninherited H-2^d (maternal) antigens. Figure 3A shows an example of a DTH assay in which splenocytes from 2 NIMA^d-exposed mice and 1 NIPA^d control mouse were tested. Splenocytes from all 3 responded strongly to TT, and weakly to BDF₁ antigen. The response of NIMA^d no. 7 induced by TT was suppressed by 50% in the presence of BDF₁ (NIMA) antigen, indicating NIMA-specific T_R in this mouse.²⁸  In contrast, splenocytes from NIMA^d no. 8 showed no evidence of bystander suppression, similar to the NIPA^d control mouse. The variability in the degree of DTH bystander suppression in NIMA^d-exposed mice (n = 36) is shown in Figure 3B. Approximately half of the NIMA^d-exposed mice showed some degree of bystander suppression (20%-75%) in the presence of BDF₁ antigen. In contrast, none of the NIPA^d control mice had more than 20% bystander suppression (P = .001 vs NIMA^d mice), confirming previously published data.²⁸  Suppression seen in the NIMA^d-exposed mice was NIMA-specific, as 4 of 6 NIMA^d-exposed offspring showed some degree of bystander suppression in the presence of BDF₁ antigen but not in the presence of third party B6C3F₁ (H2^b/k) antigen (Figure 3A-B).

To determine if MMc was correlated with NIMA-specific DTH regulation, we compared the percentage of tissues containing MMc to the percentage of DTH suppression values in individual mice and found a strong linear correlation between these 2 variables (P = < .001, r = .92; Figure 4A). These data suggest that the widespread distribution of MMc is important for inducing or maintaining regulation to NIMA.

Another way to evaluate MMc is to take an average of the estimated Geq per 10⁵ cells in all tissues tested, which takes into account both the calculated levels as well as the number of positive organs. It should be pointed out that a very high GEq per 10⁵ cells value in 1 organ could skew the average, even if the distribution of MMc was very narrow. We compared this parameter with the extent of bystander suppression as both a discontinuous and a continuous variable. The average H2D^d Geq per 10⁵ cells detected in a given mouse was strongly correlated with the presence of a greater than 50% bystander suppression response to maternal alloantigen (P < .001; Figure 4B), a threshold that we have previously found correlates with tolerance in human and mouse transplant recipients.^33,34  There was also a significant correlation with the percentage of suppression of DTH as a continuous variable (P = < .001), although the linear relationship seen between DTH suppression and the number of MMc⁺ organs (r = .92; Figure 4A) was not as strong when the average Geq per 10⁵ cells was used as the measure of MMc (r = 0.82; Figure 4C).

Bystander suppression of a recall DTH reaction in presence of maternal antigens can be mediated by antigen-specific T_R cells that produce IL-10 and express surface TGF-β/LAP complexes.²⁸  Foxp3⁺ T_R cells are also known to inhibit T_E cell responses by both TGF-β–dependent and –independent mechanisms.^35,36  We therefore investigated whether those offspring with high levels of MMc are more likely to exhibit decreased lymphoproliferative responses to maternal antigen, and to have CD4⁺ T cells with increased levels of Foxp3 or surface TGF-β in presence of maternal antigens. To this end, we performed in vivo MLR assays as described previously.²⁸  Inguinal LN (ILN) and spleen cells were harvested from BDF₁ mice, which had received CFSE-labeled splenocytes by intravenous injection 3 days earlier from either NIMA^d-exposed or NIPA^d control mice (H2^b/b). Transferred (H2K^d-negative) lymphocytes present in the host ILNs were gated and analyzed for proliferation by CFSE dilution as shown in Figure 5A. Total lymphocytes derived from the MMc^neg NIMA^d-exposed mouse no. 10 and NIPA^d control mouse no. 2 proliferated more vigorously in the semiallogeneic BDF₁ host (maternal type) than that from MMc⁺ NIMA^d-exposed mouse no. 9. When lymphoproliferation data from all NIMA^d-exposed mice tested were compared with the average level of MMc, we found a strong inverse correlation in both BDF₁ ILNs (P = < .001, r = −.90; n = 22; Figure 5B) and spleen (P = < .001, r = −.89; n = 22; data not shown).

To determine if we could identify T_R cells among those recovered from the spleen or ILN after the 3-day in vivo MLR, CD4⁺ T cells were further analyzed for TGF-β and Foxp3 expressions. As shown in Figure 5C, we gated on the CD4⁺ T cells within the H2K^d-negative CFSE⁺ cells for determining surface TGF-β or nuclear Foxp3 expression on the donor CD4⁺ T cells. As shown in Figure 5D, NIMA^d no. 9 (MMc⁺) had approximately 7-fold higher percentage of surface TGF-β⁺ cells within the CD4⁺ subset recovered from ILNs compared with MMc⁻ NIMA^d no. 10 or the NIPA^d offspring. Strikingly, the TGF-β⁺ CD4⁺ T_R cells were present almost exclusively in a CFSE^bright, nonproliferating subpopulation. No significant difference was found in Foxp3 expression by transferred CD4⁺ T cells, either CFSE^bright or CFSE^dim (Figure 5E). When data from all NIMA mice tested were compared, there were significant correlations between the average level of MMc and the percentage of spleen-derived CD4⁺ T_R cells expressing surface TGF-β in both ILNs (P = < .001, r = .92; n = 15; Figure 5F) and spleen (P = < .001, r = .88; n = 15; data not shown).

The strong correlation of CD4⁺ T_R cells with MMc in adult offspring was impressive, but it did not explain how so few maternal cells could provide enough antigen for maintenance of tolerance to NIMA. One possibility is signal amplification via antigen acquisition or trogocytosis, a process of surface membrane exchange between cells that can be readily demonstrated both in vitro (supplemental Figure 5A) and in vivo.²⁴  We found that H2K^d+ dimly stained cells (H2K^d-dim) were indeed detectable by flow cytometry in peripheral blood of NIMA^d-exposed mice at levels above B6 “background” but were absent in NIPA^d control mice (Figure 6A). The median channel fluorescence of H2K^d staining of the cells in NIMA^d-exposed mice was intermediate between that of positive (BDF₁) and negative (B6) parental strain controls. The H2K^d-dim cells might be maternal cells that have reduced their MHC class I expression. If this were so, then isolating the dimly positive cells should result in a high degree of enrichment for maternal DNA. To test this idea, we sorted H2K^d-dim splenocytes from 3 NIMA-exposed offspring and analyzed them for MMc. We found that H2K^d-dim cells from only 1 of 3 mice had detectable maternal DNA, and the level was low (∼ 1:50 000 cells). This suggests that a majority of the H2K^d-dim cells were offspring-derived and not maternal cells (supplemental Figure 5B). Interestingly, most of the offspring cells that expressed low levels of maternal H2K^d (MHC class I) also expressed IE^d (MHC class II; supplemental Figure 5C). Together, these results suggest a process of antigen acquisition by migrating leukocytes of the offspring that encounter rare maternal cells, including professional APC in tissues. Double-staining with anti-H2K^d mAb plus CD4-, CD8-, CD11b-, or CD11c-specific antibodies revealed that primarily CD11b⁺ and CD11c⁺ APCs, and not T cells, expressed low levels of maternal antigen in vivo (supplemental Figure 6). Overall, we found an elevated percentage of H2-K^d-dim cells in all lymphoid tissues tested except thymus. The difference between NIMA^d-exposed versus NIPA^d control mice was highly significant in spleen (P < .001) and blood (P < .001; Figure 6B).

Given the strong correlation between the levels of NIMA-specific T_R cells and MMc, and the fact that MMc has been reported to vary over time in mice and humans,^37,38  we wished to determine whether the levels of MMc correlated with maternal H2K^d antigen expression. Blood was sampled from F₁ backcross offspring weekly for 4 weeks after weaning, and all mice were killed 1 week later for MMc analysis. We found variability in the level of circulating maternal antigen-dim cells over time (Figure 6C). The 2 mice that expressed cell-surface maternal antigens, no. 11 and no. 12, were MMc⁺, while NIMA no. 13, negative for H2K^d antigen expression, was also MMc^neg. Overall, MMc⁺ NIMA^d-exposed mice had higher peak percentages of H2K^d-dim cells than MMc^neg NIMA^d-exposed and NIPA^d control offspring (both P = .016; n = 13 mice tested; Figure 6D).

We have previously reported that when NIMA^d-exposed mice were foster-nursed by B6 females, they failed to become tolerant to a DBA/2 heart allograft.²⁶  We wished to determine whether (1) mice at birth have a high or low level of MMc, (2) the impact of foster-nursing by a B6 female on H-2^d MMc, and (3) if depriving the offspring of oral NIMA^d-exposure upsets the balance between NIMA-specific T_R and T_E cells, as measured by DTH analysis. When we measured MMc in newborn offspring (< 1 day old), we found 9 of 9 NIMA^d-exposed offspring had MMc in at least 1 of 3 organs tested (Figure 7A). The level of MMc was variable in heart, liver and lungs, with the highest readings recorded in heart tissue. No H2D^d Mc was detected in the organs from 4 newborn NIPA^d control offspring. The average level of MMc in heart, liver, and lungs of neonatal NIMA^d-exposed offspring was approximately 14 times higher than that in the adult offspring (6-8 weeks old). A majority (61%; 25 of 41) of adult NIMA^d mice nursed by their own mothers had detectable MMc in 1 or more organs (Figure 7B). However, when the NIMA^d-exposed offspring were foster-nursed by B6 female mice, there was detectable MMc in only 1 of 10 adult mice (Figure 7B), and the 1 exception was a mouse that had MMc only in the heart (data not shown). As a control, some NIMA^d-exposed offspring were foster-nursed by another BDF₁ mother; 2 of 3 of these mice had a level of MMc similar to the average level found in naturally nursed NIMA^d-exposed offspring (Figure 7B).

The loss of MMc in mice exposed in utero only to NIMA^d antigens might be due to a passive process, or to active elimination of maternal cells by host T_E cells now unopposed by T_R cells. If the latter mechanism is correct, then there ought to be a shift away from regulation and toward sensitization in mice deprived of oral exposure to antigen. To determine the level of sensitization to maternal antigen, we used the DTH assay. We found that none of the naturally bred and nursed NIMA^d mice (n = 36) responded to BDF₁ antigen (0-5 × 10⁻⁴ inches net swelling), whereas 4 of 12 mice foster-nursed by a B6 mouse displayed a marginally elevated response (10-20 × 10⁻⁴ inches net swelling; Figure 7C). The mean footpad swelling was less than that observed in NIPA^d controls that were sensitized as adults by BDF₁ splenocytes before assay but significantly higher than that of standard bred and nursed NIMA^d mice (P = .006). This finding suggests that, in the absence of oral exposure immediately after birth, in utero exposure to maternal antigens can result in NIMA^d-specific sensitization, along with loss of MMc.

Microchimerism has previously been shown to contribute to tolerance by induction of anergy or deletion of donor-specific cytotoxic T lymphocytes (CTLs).^18,19  Although the presence of maternal thymic epithelial cells may indicate deletion of high-affinity maternal antigen-specific offspring T cells, the current study strongly supports a peripheral tolerance mechanism for the NIMA effect, a mechanism suggested by the recent discovery of coexisting MMc and NIMA-specific Foxp3⁺ T_R cells in the human fetal LN.¹⁶  Given the well-recognized importance of T_R cells in inducing and maintaining tolerance to allografts,³⁹  the current study linking MMc to establishment of NIMA-specific T_R cells, and the previous demonstrations of heart transplantation tolerance in approximately half of all NIMA^d-exposed mice,^26,28  the link between Mc and tolerance suggested previously²⁶  has now gained additional support. However, when Mc is not the only source of alloantigen, as in the case where Mc is eliminated or reduced by antibody depletion but a transplanted organ remains,⁴⁰  the importance of Mc in maintaining tolerance may be diminished. Moreover, not all forms of microchimerism were equally effective at inducing and maintaining T_R cell–based tolerance. For example, isolated Mc in the heart, observed in some NIPA controls (Figure 1) and in 1 foster-nursed (by B6) NIMA^d-exposed offspring (data not shown), may be instead a sign of split tolerance resulting from elimination of professional class II⁺ APCs.^22,23 

The immediate postnatal period appears to be critical for establishing a favorable T_R > T_E cell balance. When NIMA^d-exposed mice were foster-nursed by B6 mothers, they lost not only their tolerance to a NIMA-expressing heart allograft,²⁶  but also MMc (Figure 7). The loss of MMc was not due to the trauma of removal from the birth mother, because MMc was observed in 2 of 3 mice foster-nursed by a BDF₁ female. This suggests a critical role of oral exposure to NIMA antigens in maintaining widespread MMc. When NIPA^d control mice were nursed by BDF₁ mothers, 2 of 6 offspring exhibited H2D^d Mc exclusively in the liver (P.D., unpublished observation, March 2009), consistent with a previous report.¹⁵  Aoyama et al,⁴¹  using the NIMA^d breeding model, recently showed that nursing alone could induce limited protection against graft-versus-host disease (GVHD), whereas full protection required both in utero and oral exposure.

Because oral tolerance is known to generate TGF-β–producing T_R cells,^42,43  oral exposure to maternal MHC antigens present in breastmilk⁴⁴  may generate additional NIMA-specific T_R cells. The latter may be necessary to suppress NIMA-specific T_E cells and thereby maintain MMc. Consistent with this interpretation, one-third (4 of 12) of the NIMA^d-exposed mice foster-nursed by B6 mothers (in utero exposure only) responded to BDF₁ antigen without any priming, while none (0 of 36) of the NIMA^d-exposed offspring nursed by their own (BDF₁) mothers did so. The poor outcome of maternal renal allografts in the precyclosporine era in recipients who had not been breastfed, compared with breastfed recipients,⁴⁵  is consistent with sensitization in individuals lacking oral exposure to NIMA as infants.

T_E cells, once they have expanded from naive precursors in response to antigen, often eliminate their antigen source; yet, because the memory T_E express high levels of IL7 receptor, this compartment is still maintained in presence of IL7.⁴⁶  In contrast, T_R cells are generally IL7R^low,⁴⁷  and therefore may require antigen stimulation for their maintenance. The finding that CD11b⁺ and CD11c⁺ subsets of spleen and bone marrow contain rare maternal cells, the high MMc levels in hematopoeitic lin⁺ cells of the heart, and the evidence that professional APCs of offspring origin may acquire maternal MHC antigen, suggest that NIMA-specific T_R cells could interact with a large number of APC-expressing NIMA antigens and allopeptides. T_R-modified suppressive APCs^48-50  of both host and maternal donor origin may well explain the phenomenon of MMc-linked CTL inhibition in a transplant recipient tolerant of a maternal kidney allograft.¹⁹  The finding of MMc in DCs and macrophage subsets is particularly important, because these rare maternal cells would express both NIMA class I and class II antigens, as well as the shared (inherited) class I and II antigens from the mother. This creates an additional avenue for tolerance: the induction of dual/allospecific CD4⁺ T_R cells capable of interacting with both donor and host APCs. Such T_R cells have recently been shown to be the extremely potent in inducing tolerance to allografts in mice,⁵¹  and may help explain the acceptance, after short-term immunosuppression, of renal allografts that are MHC class I–mismatched, but MHC class II–matched.^52-54 

In conclusion, we have found a striking correlation between MMc and NIMA-specific T_R cells capable of suppressing both DTH and lymphoproliferative responses of T_E cells in adult mice. The data suggest that maintenance of MMc is critically intertwined with NIMA-induced regulation and T_R cell generation, which depends on oral exposure to maternal antigens via nursing, and may explain why some but not all offspring become microchimeric and tolerant to a maternal antigen-expressing allograft. Our data run counter to the idea proposed by Mold et al¹⁶  that the T_R cells induced by fetal exposure to NIMA are long-lived memory cells that normally will persist into adulthood. If this were so, the level of persisting maternal antigen should not matter.

Although we have shown that MMc and NIMA-induced tolerance are correlated, we did not rule out the possibility that tolerance to NIMA can lead to persistence of maternal cells in the offspring. The issue of cause and effect of Mc²¹  can only be definitively resolved by targeted elimination of Mc in the adult mouse, to see if T_R cells are rapidly lost or if their predominance over T_E cells is maintained. Meanwhile, the precise role of microchimerism in acquired tolerance to solid organ or bone marrow transplants remains unresolved. Although trans vivo DTH and in vivo MLR assays are convincing surrogates for conventional allograft transplantations, correlating levels of MMc with subsequent survival of allografts is desirable. In light of our findings, and the recent identification of NIMA-minor H-specific CD8 T_R cells in healthy individuals,⁵⁵  pretransplantation screening to evaluate both Mc and T_E/T_R balance may now be warranted to determine whether a given individual is likely to accept allografts expressing NIMAs.

We would like to thank Steve Schumacher for technical assistance in typing mice, DNA extraction, and qPCR; M. Suresh, A. Shaaban, and A. Stevens for critical reading of the manuscript; and L. Haynes and E. Jankowska-Gan for their careful editing of the manuscript.

This work was supported by grant no. 5R01AI066219-04 from NIH (to W.J.B.).

National Institutes of Health

Contribution: P.D. designed and performed all experiments (except DTH assay), analyzed data, and prepared the manuscript; M.M.-D. performed DTH assay and subset trogocytosis experiment (supplemental Figure 6), and edited the manuscript; J.L.B. tested primers; D.A.R. and J.R.T. helped with immunohistochemistry; Z.Y. designed primers; and W.J.B. helped in experiment design and manuscript editing.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: William J. Burlingham, H4/749 Clinical Science Center, 600 Highland Ave, Madison, WI 53792; e-mail: burlingham@surgery.wisc.edu.