PIRs mediate innate myeloid cell memory to nonself MHC molecules

See allHide authors and affiliations

Science  05 Jun 2020:
Vol. 368, Issue 6495, pp. 1122-1127
DOI: 10.1126/science.aax4040

Innate immune cells remember

Immunological memory is a phenomenon by which immune cells can quickly recognize an antigen that the host has previously encountered. Certain cells of the innate immune system exhibit memory-like responses know as trained immunity. Rapid, antigen-specific secondary (anamnestic) responses were long thought to be the domain of B and T cells. However, Dai et al. report that monocytes and macrophages can acquire memory specific for particular major histocompatibility complex I antigens using paired A-type immunoglobulin-like receptors (PIR-As) (see the Perspective by Dominguez-Andrés and Netea). This pathway contributes to recognition and rejection of allograft-transplanted tissue from a donor of the same species. Genetic depletion or blockade of PIR-As in mice diminished the rejection of kidney and heart allografts. This work, which expands immunological memory to include myeloid cells, points to targets that may improve organ transplantation outcomes in the future.

Science, this issue p. 1122; see also p. 1052


Immunological memory specific to previously encountered antigens is a cardinal feature of adaptive lymphoid cells. However, it is unknown whether innate myeloid cells retain memory of prior antigenic stimulation and respond to it more vigorously on subsequent encounters. In this work, we show that murine monocytes and macrophages acquire memory specific to major histocompatibility complex I (MHC-I) antigens, and we identify A-type paired immunoglobulin-like receptors (PIR-As) as the MHC-I receptors necessary for the memory response. We demonstrate that deleting PIR-A in the recipient or blocking PIR-A binding to donor MHC-I molecules blocks memory and attenuates kidney and heart allograft rejection. Thus, innate myeloid cells acquire alloantigen-specific memory that can be targeted to improve transplant outcomes.

Immunological memory protects the host against infection but is also a potent barrier to transplant survival (1). Initially thought to be confined to T and B lymphocytes, it is now evident that innate myeloid and lymphoid [natural killer (NK)] cells also acquire features of immunological memory. In lymphoid cells, memory responses are anamnestic (13). By contrast, myeloid cells exhibit memory-like responses, referred to as trained immunity, as secondary stimuli that elicit enhanced reactions need not be related to the primary stimulus (4). Whether myeloid cells acquire memory specific to previously encountered antigens is not known but has become an important question, given the key roles of these cells in host defense (5).

We studied monocyte and macrophage reactions to allogeneic bone marrow plug grafts placed under the kidney capsule. Monocytes and macrophages sense major histocompatibility complex (MHC) and non-MHC determinants on allogeneic tissues and mature to antigen-presenting, inflammatory, or cytotoxic cells (69). To investigate memory, B6-Rag−/−Il2rg−/− mice, which lack B, T, NK, and innate lymphoid cells, were immunized with irradiated allogeneic splenocytes. They were then rechallenged with grafts from same-party (i.e., the same source as the immunizing splenocytes) or third-party allogeneic donors after 7, 28, or 49 days. B6-Rag−/−Il2rg−/− (H-2b) mice immunized with BALB/c (H-2d) splenocytes and rechallenged with same-party BALB/c allografts 7 or 28 days later exhibited significantly greater graft infiltration with recipient monocyte-derived dendritic cells (Mo-DCs) than unimmunized recipients or those immunized with syngeneic (B6) or third-party allogeneic (C3H or H-2k) splenocytes (Fig. 1A). This enhanced response dissipated by 49 days after immunization. B6-Rag−/−Il2rg−/− recipients of C3H grafts mounted a heightened response if previously immunized with C3H but not B6 or BALB/c splenocytes, which further demonstrates allospecificity (Fig. 1B). Enhanced responsiveness could not be attributed to immunogen persistence because neither donor splenocytes nor intact donor MHC-I molecules were detected in the host’s circulation or spleen at the time of rechallenge (fig. S1A). Thus, host monocytes appear to acquire specific memory to previously encountered alloantigens.

Fig. 1 Monocyte and macrophage memory specific to allogeneic MHC molecules.

(A to D) Monocyte responses in immunized mice measured as monocyte-derived dendritic cells (Mo-DCs) in the graft (n = 5 to 12 mice; N = 1 to 2 experiments). d, day. (E) Monocytes from immunized mice transfer memory to naïve recipients (n = 2 to 4; N = 1 to 3). syn, syngeneic; allo, allogeneic. (F) Allospecific killing of carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled cells in immunized mice (n = 6; N = 2). (G and H) Nonself MHC recognition is necessary for eliciting monocyte (n = 6; N = 1) (G) and macrophage (n = 6; N = 2) (H) memory. Killing of target cells lacking nonself MHC (BALB.B) or MHC-I (B2m–/–) is significantly diminished (H). Assays were performed 7 days after immunizing B6-Rag−/−Il2rg−/− mice. (I) Allogeneic MHC-I tetramer (tet) binding to macrophages from BALB/c-immunized B6- Rag−/−Il2rg−/− hosts (representative of three experiments). Max, maximum. Statistical analyses: mean and individual biological replicates [(A) to (H)]; one-way analysis of variance (ANOVA) [(A) to (C), (F), and (H)]; or two-tailed Student’s t test [(D), (E), and (G)]. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; ns, not significant.

To rule out any possible contribution of lymphoid cells in the graft, we repeated the experiments using Rag−/−Il2rg−/− donors. B6-Rag−/−Il2rg−/− recipients responded more vigorously to BALB/c-Rag−/−Il2rg−/− bone marrow plug allografts if previously immunized with BALB/c-Rag−/−Il2rg−/− but not NOD-Rag−/−Il2rg−/− splenocytes (Fig. 1C), whereas NOD-Rag−/−Il2rg−/− immunization enhanced the response to NOD-Rag−/−Il2rg−/− allografts (Fig. 1D). Thus, monocyte memory to alloantigens is not dependent on lymphoid cells from either the graft or the recipient.

The primary innate alloresponse that generates Mo-DCs is mediated by inflammatory (Ly6Chi) monocytes (6, 7). To test whether this monocyte subset acquires memory, we sorted Ly6Chi monocytes from B6-Rag−/−Il2rg−/− mice 7, 21, or 35 days after immunization and transferred them to naïve B6-Rag−/−Il2rg−/− hosts (Fig. 1E). Mice that received monocytes from allostimulated donors mounted a significantly greater response to BALB/c allografts than those that received monocytes from donors stimulated with syngeneic cells. This demonstrates that Ly6Chi monocytes are capable of mediating memory and that they retain memory function for up to several weeks after their first encounter with an alloantigen.

Macrophages acquire the ability to kill allogeneic targets if first challenged with allogeneic cells along with CD40 receptor cross-linking (9). We therefore examined whether macrophages also acquire allospecific memory. B6-Rag−/−Il2rg−/− mice were left unstimulated (naïve) or were immunized with BALB/c splenocytes plus agonistic anti-CD40 antibody, and an in vivo killing assay was performed 14 and 28 days later (Fig. 1F). BALB/c-immunized mice killed BALB/c but not syngeneic (B6) or third-party (C3H) targets, which suggests that macrophages, like monocytes, acquire memory specific to previously encountered alloantigens.

MHC proteins are the principal alloantigens in humans and mice. To test whether monocyte or macrophage memory is dependent on the detection of nonself MHC molecules, we immunized B6-Rag−/−Il2rg−/− (H-2b) recipients with irradiated BALB/c (H-2d) splenocytes and rechallenged them with BALB.B (H-2b) grafts. BALB.B and B6 mice share the same MHC (H-2) but differ at non-MHC loci. Immunization with BALB/c cells did not enhance the response to BALB.B grafts, which indicates that monocyte memory is against nonself MHC molecules (Fig. 1G). The same was true for macrophages (Fig. 1H). Moreover, BALB/c-immunized hosts killed β2-microglobulin knockout targets less efficiently than wild-type cells, which indicates that the response is primarily to nonself MHC-I molecules (Fig. 1H). Furthermore, macrophages from BALB/c-immunized B6 mice bound allogeneic H-2Dd (and, to some extent, allogeneic H-2Kd) but not syngeneic H-2Db and H-2Kb MHC-I tetramers (Fig. 1I). Additionally, host exposure to both MHC and non-MHC polymorphisms at the time of priming was necessary to generate memory (fig. S1B). Polymorphism in the non-MHC signal regulatory protein alpha (Sirpa) gene, which is known to trigger the primary monocyte alloresponse (8), was required (fig. S1C).

A database search across 17 mouse genomes for potential MHC-I–binding molecules on myeloid cells identified several families of polymorphic immunoglobulin (Ig) superfamily receptors. Among these were the paired Ig-like receptors (PIRs), which are orthologs of human leukocyte Ig-like receptors (LILRs) (10). Closely linked genes coding for six PIR-A proteins and one PIR-B protein are located on mouse chromosome 7 (11). PIR-B is nonpolymorphic and binds a wide spectrum of MHC-I molecules (12). Its cytosolic domain transmits inhibitory signals (12, 13), whereas PIR-As are stimulatory (14, 15). PIR-A isoforms are believed to bind distinct MHC-I molecules (12), and PIR-A expression is up-regulated upon myeloid cell differentiation and activation (15). Using anti–PIR-A/B and anti–PIR-B antibodies, we detected PIR-A expression on resting Ly6Chi monocytes (Fig. 2A). Macrophages expressed both PIR-A and PIR-B and up-regulated PIR-A after allostimulation (Fig. 2B). Blocking these receptors with an anti–PIR-A/B antibody inhibited monocyte memory in Pirb-sufficient and Pirb−/− B6-Rag−/−Il2rg−/− hosts (Fig. 2, C and D) and suppressed the killing of allogeneic targets in alloimmunized B6-Rag−/−Il2rg−/− mice (Fig. 2E). We confirmed these results by treating animals with PIR-A3/Fc fusion protein, which preferentially blocks PIR-A3 from binding to its MHC-I ligand, H-2Dd. The preferred ligand of PIR-A3 is H-2Dd (Fig. 2, F to H). H2-Dd tetramers triggered the strongest signal in reporter cells expressing PIR-A3 (Fig. 2F). They bound PIR-A3–transfected 3T3 fibroblasts better than H-2Db tetramers (Fig. 2G), and the binding of PIR-A3/Fc was greater to BALB/c (H-2Dd+) than B6 cells (H-2Db+ but H-2Dd) (Fig. 2H). PIR-A3/Fc inhibited both macrophage (Fig. 2I) and monocyte (Fig. 2J) memory in BALB/c-immunized B6-Rag−/−Il2rg−/− mice. By contrast, PIR-A1/Fc did not inhibit the killing of BALB/c (H-2d) targets (Fig. 2I). Furthermore, PIR-A3/Fc failed to suppress the monocyte memory response to C3H (H-2k) grafts (Fig. 2J), which confirmed the specificity of PIR-A3 to BALB/c H-2Dd molecules. Finally, we generated PIR-A locus knockout (Pira−/−) mice lacking PIR-A1-5 (fig. S2), which were bred onto the B6-Rag−/−Il2rg−/− background. Notably, these mice did not mount a monocyte memory response to BALB/c allografts 7 or 28 days after priming (Fig. 2K). By contrast, the primary response to BALB/c allografts (Fig. 2K) and primary dendritic cell and macrophage functions (antigen presentation and IL-6 production) (fig. S3) were not inhibited in Pira−/− recipients. Thus, PIR-A mediates innate memory.

Fig. 2 Memory to allogeneic MHC-I is mediated by PIR-A molecules on monocytes and macrophages.

(A and B) PIR-A and -B expression on Ly6Chi monocytes (A) and macrophages (B) before [day 0 (d.0)] and after alloimmunization (d.14 and d.28). Colored histograms show biological replicates (n = 3). (C to E) Blocking PIR-A/B inhibits monocyte (n = 12; N = 2) [(C) and (D)] and macrophage (n = 6; N = 2) (E) memory. B6-Rag−/−Il2rg−/− hosts were immunized with BALB/c splenocytes and rechallenged with BALB/c allografts 7 days later. (F to H) H-2Dd MHC-I tetramers bind preferentially to PIR-A3-transfected BWZ.36 (n = 5; N = 1) (F) and 3T3 cells (G). PIR-A3/Fc binds preferentially to H-2Dd+ (BALB/c) cells (H). β-gal; β-galactosidase. (I and J) Specific inhibition of macrophage (n = 5 to 6; N = 2) (I) and monocyte (n = 6; N = 1 to 2) (J) memory to BALB/c allografts by PIR-A3/Fc. Immun., Immunization. (K) Absent memory 7 and 28 days after immunization, but normal primary monocyte alloresponse in Pira−/− hosts (n = 6; N = 1 to 2). Statistical analyses: representative of three experiments [(B), (G), and (H)]; mean and individual biological replicates [(C), (D), (F), and (I) to (K)]; one-way ANOVA [(C) to (E), (I), and(J)]; or two-tailed Student’s t test (K).

We next investigated how monocyte memory is acquired and whether it resembles NK memory (16). B6 mice immunized with BALB/c (H-2d) cells selectively increased the proportion and number of splenic monocytes that bind the H-2Dd tetramer, whereas C3H (H-2k) immunization exclusively increased H-2Dk-binding cells (Fig. 3A and fig. S4A). This effect was preserved in Pirb−/− mice but abolished in Pira−/− mice. Only a minority of monocytes bound more than one type of tetramer (fig. S4B), which suggests that PIR-A expression is variegated. Single-cell RNA sequencing (scRNA-seq) analysis of splenic monocytes performed 1 and 4 weeks after immunization showed that monocytes followed a pseudotime trajectory from a starting state (S), which dominated in mice stimulated with syngeneic cells, to expanding states (E1 to E3), which increased proportionally in allostimulated groups (Fig. 3, B and C). E states were enriched for cell cycle, antigen presentation, and immune pathways such as allograft rejection and graft-versus-host disease (data S1 and S2) and exhibited increased Pira and reduced Pirb expression (Fig. 4D). Cell cycle pathway enrichment is consistent with prior evidence of monocyte proliferation after allostimulation (8). Finally, monocytes that expressed mRNA of two different Pira alleles constituted a small minority before and after allostimulation (Fig. 4E), which provides further evidence for variegated PIR-A expression. Thus, a possible mechanism of monocyte memory is the clonal expansion of cells specific to a particular nonself MHC molecule.

Fig. 3 Mechanisms of monocyte memory.

(A) Tetramer-positive splenic monocytes 7 days after immunizing B6-Rag−/−Il2rg−/− mice (mean and individual biological replicates; n = 3; N = 1 to 3). wt, wild-type. (B to E) Splenic monocyte scRNA-seq analysis 1 and 4 weeks after immunization (n = 3; N = 1) illustrated with: pseudotime plots (S indicates starting state, E1 to E3 indicate expanding states) (B); heatmaps and graphs depicting the proportion of S and E states in each immunization group (C); differential expression of Pir genes in E states versus the S state (D); and proportion of monocytes expressing Pira2, Pira3, or both in each state (E). One-way ANOVA (A).

Fig. 4 Genetic deletion or blocking PIR-A attenuates kidney and heart allograft rejection.

(A to D) Survival (A), serum creatinine (B), histology (C), and chronic rejection scores (D) of BALB/c kidneys transplanted to wild-type (wt), Pira–/–, or Pirb–/– B6 recipients. Arrows in (C) point to infiltrates and fibrosis. H&E, hematoxylin and eosin. (E and F) Survival (E) and histology (F) of BALB/c hearts transplanted to B6 mice treated with PIR-A3/Fc (± CTLA4Ig) or control mouse IgG1 (Ctrl IgG). Arrows in (F) point to chronic rejection vasculopathy. VVG, Verhoeff-Van Gieson. Statistical analyses: mean and individual biological replicates [(B) and (D)]; log-rank [(A) and (E)]; and one-way ANOVA [(B) and (D)].

Mo-DCs and macrophages contribute to allograft rejection in mice and humans (7, 1719). However, it is unknown whether monocyte or macrophage memory mediated by PIR-A plays a role in this rejection. Kidney allografts survived long-term (>125 days) and maintained normal serum creatinine in lymphoid cell–sufficient wild-type and Pira−/− recipients, but they were rapidly rejected by Pirb−/− mice (Fig. 4, A and B), consistent with the inhibitory functions of PIR-B (12, 13). PIR-A3/Fc prevented acute allograft rejection in Pirb−/− mice (Fig. 4, A and B), which suggests that PIR-A accelerates rejection in the absence of PIR-B. Chronic rejection and cellular infiltrates in wild-type mice were significantly attenuated in Pira−/− mice and in Pirb−/− recipients treated with PIR-A3/Fc (Fig. 4, C and D, and fig. S5A). As expected, chronic rejection did not occur in grafts that were rejected acutely [Pirb−/− and Pirb−/− plus control immunoglobulin G (Ctrl IgG) groups] (Fig. 4, C and D). We also observed a pathogenic role for PIR-A in heart transplantation. CTLA4-Ig or PIR-A3/Fc alone caused significant but modest prolongation of allograft survival, whereas their combined administration markedly extended survival and prevented pathology associated with acute or chronic rejection (Fig. 4, E and F, and fig. S5, B and C). Neither PIR-A deficiency nor blockade suppressed alloantibody production (fig. S5, D and E). Thus, PIR-A, which is necessary for monocyte and macrophage memory, promotes allograft rejection, whereas PIR-B tempers it.

The findings described in this work identify a pathway of MHC-I allorecognition that is mediated by monocytes and macrophages, generates memory specific to previously encountered MHC-I alloantigens, and contributes to allograft rejection. This pathway extends the domain of classical immunological memory to innate myeloid cells and can be potentially targeted to improve the survival of life-saving organ transplants.

Supplementary Materials

Materials and Methods

Figs. S1 to S5

Tables S1 to S4

References (2028)

Data S1 and S2

References and Notes

Acknowledgments: We would like to thank C. Bi and Z. Kou of the Transgenic and Gene Targeting Core (Department of Immunology, University of Pittsburgh School of Medicine) for microinjection of zygotes and production of Pira- and Pirb-mutant mice. Funding: This work was supported by NIH grants AI145881 (M.H.O.), AI080779 (X.C.L.), and AI099465 (F.G.L.) and by funds from the Frank and Athena Sarris Chair in Transplantation Biology at the University of Pittsburgh (F.G.L.). J.S.D. was supported by the JDRF, the Canadian Institutes of Health Research, and the SickKids Foundation. Author contributions: H.D., P.L., D.Z., W.L., Wen.C., A.J.F., T.L., L.R.M., and H.R.T. performed experiments and analyzed data. K.A.-D., T.S., and J.C. performed bioinformatic analysis. Wei C. supervised bioinformatic analysis. A.L.W. and S.M.-T. performed mouse breeding and genotyping. J.S.D. supervised generation and genotyping of NOD congenic mice and edited the manuscript. C.W. and P.N. provided data and insights into human correlates of mouse findings. M.L.N. performed comparative mouse genomics and edited the manuscript. S.G. designed and constructed gene knockout mice and contributed to the manuscript writing. E.T. and H.K. designed, constructed, and provided expression vectors and reagents and edited the manuscript. M.J.S. supervised design and construction of Pirb−/− mice, provided the mice, and edited the manuscript. M.H.O. participated in experimental design, coordinated experiments, trained and supervised personnel, and wrote the manuscript. M.H.O., X.C.L., and F.G.L. conceived the idea, designed experiments, analyzed data, supervised overall work, and wrote the manuscript. Competing interests: The Hospital for Sick Children receives royalties from Trillium Therapeutics, Inc. for SIRPa-Fc, of which J.S.D. receives a fraction. J.S.D. is a coinventor on a patent related to this work, “Role of SIRP alpha polymorphisms in human acute myeloid leukemia.” PCT applications are pending in the United States, Canada, and China and have been awarded in Australia (2015) and Japan (2018). Data and materials availability: Raw scRNA-seq data have been deposited into the Gene Expression Omnibus (GEO) database (accession no. GSE147596). All other data needed to evaluate the conclusions of the paper are available in the manuscript or the supplementary materials.

Stay Connected to Science

Navigate This Article