Combination of uterine natural killer cell immunoglobulin receptor haplotype and trophoblastic HLA-C ligand influences the risk of pregnancy loss: a retrospective cohort analysis of direct embryo genotyping data from euploid transfers

The risk of pregnancy loss after 668 euploid single-embryo transfers was significantly influenced by the combination of embryonic HLA-C ligands and maternal uterine natural killer cell immunoglobulin receptor haplotypes.

Volume 107, Issue 3, Pages 677–683


Scott J. Morin, M.D., Nathan R. Treff, Ph.D., Xin Tao, M.S., Richard T. Scott III, B.S., Jason M. Franasiak, M.D., Caroline R. Juneau, M.D., Marcy Maguire, M.D., Richard T. Scott, M.D., H.C.L.D.



To compare maternal uterine natural killer cell immunoglobulin receptor (KIR) genotype and haplotype frequencies between patients whose euploid single-embryo transfer resulted in pregnancy loss and those that resulted in delivery and to determine if the risk of pregnancy loss was affected by the HLA-C genotype content in the embryo.


Retrospective cohort.


Academic research center.


Autologous fresh IVF cycles resulting in positive serum β-hCG during 2009–2014.



Main outcome measure(s)

1) Relative risk of pregnancy loss according to maternal KIR genotypes and haplotypes. 2) Comparison of pregnancy loss rates within each KIR haplotype according to HLA-C ligand present in trophectoderm biopsy samples.


A total of 668 euploid single-embryo transfers with stored maternal DNA and available preamplification DNA from prior trophectoderm biopsy samples were studied. KIR2DS1, KIR3DS1, and KIR2DS5 were more common in patients who experienced pregnancy loss. Carriers of KIR A haplotype exhibited a decreased risk of pregnancy loss compared with KIR B haplotype carriers. However, among KIR A haplotype carriers, the risk of loss was significantly influenced by whether the transferred embryo carried a C1 allele versus no C1 alleles.


KIR A haplotype carriers experienced fewer pregnancy losses than KIR B haplotype carriers after euploid single-embryo transfer. However, this risk was modified by HLA-C alleles present in the embryo. High-risk combinations (KIR A/homozygous C2 and KIR B/homozygous C1) resulted in a 51% increased risk of loss over all other combinations.

Read the full text here.

Please sign in or register for FREE

Your Fertility and Sterility Dialog login information is not the same as your ASRM or EES credentials. Users must create a separate account to comment or interact on the Dialog.

Go to the profile of Darren Ritsick
almost 6 years ago
We thank the authors for this work which we read with great interest and contributes to our understanding of the role that the HLA-C/KIR allorecognition system plays in regulating pregnancy outcome. We have been following work in this area for some time, particularly studies coming from Dr. Ashley Moffett’s group. Some of the author’s findings are in apparent agreement with those of Dr. Moffett’s group, particularly stratification of the risk conferred by a maternal KIR AA haplotype according to the HLA-C1/C2 content of the embryo (although not directly comparable as your study did not consider the embryonic C1/C2 content relative to the maternal C1/C2 content). However, we were surprised to read in your study that a maternal KIR B haplotype overall conferred a higher risk for pregnancy loss than a KIR A haplotype. This result is apparently incongruent with what is currently known regarding HLA-C/KIR signaling in the context of uNK cell function in spiral artery remodeling and placentation – namely that uNK cell activation (leading to secretion of cytokines including IFNy) is required for efficient spiral artery remodeling and that this is promoted by the interaction of HLA-C with activating KIRs found in the B haplotype. One satisfying outcome of the Moffett group’s epidemiologic findings (although this does not render it infallible) is that they are highly compatible with what is known about the basic biology of this system. In considering the apparent discrepancies between the present results and those of some previous studies, the author’s offer some potential explanations. Among these is that the present study ruled out any confounding effects of embryo aneuploidy (for the sake of argument we are ignoring for now debate around determination of embryo ploidy by PGD). Although the authors did not discuss this specifically in relation to the Moffett group’s findings, we don’t believe that this can account for the discrepant results. By studying effects of the HLA-C/KIR system on the risk of developing later pregnancy complications (preeclampsia and IUGR) any potential confounding effects of aneuploidy were minimized. Further, when a population of recurrent pregnancy loss (RPL) patients was considered, the significance of KIR gene frequencies was actually accentuated relative to preeclampsia and IUGR. We expect that if aneuploidy was a confounding factor, then the significance of KIR genes on outcome would be decreased when studying populations with complications manifesting earlier in pregnancy. Instead, the degree of risk associated with KIR gene frequencies positively correlated with decreasing time of clinical manifestation. This is consistent with a currently widely-accepted model in which IUGR, preeclampsia, and RPL represent a continuum of clinical manifestations of increasingly shallow implantation/deficient placentation. The present finding of temporal differences in losses associated with KIR B-C1/C1 (biased toward biochemical losses) and KIR A-C2/C2 (biased toward clinical losses) combinations offers another possible explanation. As the authors state, this may reflect differences in the timing of HLA-C1 and HLA-C2 allele expression by the embryo. Although at first consideration we did not find this hypothesis particularly appealing, we note that allelic differences in HLA-C expression in response to certain stimuli have been previously noted (Hundhausen, 2012). However, this explanation would require that differential expression of HLA-C alleles (presumably due to SNPs and/or differences in patterns of promoter elements) would need to be split relatively neatly between alleles belonging to the C1 and C2 allotypes. This explanation also implies that timing of trophoblast/HLA-C and uNK/KIR interactions, and therefore timing of uNK cell activation and downstream effects on spiral artery remodeling, would differ amongst pregnancies involving different HLA-C/KIR combinations. Although this is possible, we currently find a theory of differential temporal expression of C1 and C2 alleles unlikely to explain the discrepant results. We wish to offer two major comments that we feel are important when interpreting the results of the present work and may, in part, help to explain the discrepancies with results obtained by Dr. Moffett’s group. First, the present study considered KIR A and B haplotypes only as whole entities and disregarded the further description of these haplotypes based on the presence/absence of their centromeric and telomeric regions “for simplicity’s sake” and for reasons of nomenclature. In fact however, it is only these regions within these haplotypes which are in strong linkage disequilibrium (LD) and the presence/absence of these regions has significant functional consequences. In particular, the only known functional activating KIR that binds to HLA-C is KIR2DS1 which is part of the telomeric region of the KIR B haplotype. It therefore becomes critical to define a KIR B haplotype in an individual according to the presence/absence of 2DS1. In fact, Moffett’s group found significant differences in risk for defective placentation based on the presence or absence of the KIR B telomeric region with protection strongly associating with the presence of this region (could not be entirely attributed to 2DS1 due to the strong LD within this region although the biology suggests this is likely the case). Therefore, KIR B haplotypes possessing or lacking the 2DS1 gene are functionally distinct in relation to signaling by HLA-C and should be considered separately to avoid confounding results. Second, the present work considers the HLA-C allotype of the embryo only and not parental HLA-C allotypes. NK cells, including uNK cells, undergo a process referred to as “licensing” or “education” in which their activation is calibrated according to the C1/C2 content of the host. Therefore, the level to which maternal uNK cells will become activated in response to HLA-C-bearing cells (ie, extravillous trophoblasts) will depend not only on the embryonic HLA-C1/C2 allotype but also on the maternal C1/C2 allotype. In fact, Moffett’s group found that it is not the C1/C2 content of the embryo per se, but rather the embryonic C1/C2 content relative to the maternal C1/C2 content, that impacts on the risk for deficient placentation. Although we don’t consider temporally differential expression of HLA-C alleles to be a likely explanation for the result, we are nevertheless intrigued by the observation of temporal difference in losses associated with the KIR A and KIR B haplotypes. Instead of representing differential timing of HLA-C allele expression by the embryo, we instead surmise that this may represent a separate mechanism for HLA-C/KIR-mediated pregnancy failure separate from suboptimal placentation. With the caveats we mention above regarding possible confounding factors (consideration of KIR B centromeric/telomeric regions and parental HLA-C allotypes), this result may imply a mechanism for KIR B-associated losses separate from suboptimal activation of uNK cells that results in faster pregnancy loss. While it’s tempting to suggest that this could represent a mechanism by which overactivation of uNK cells results in rapid failure of embryo implantation (and by extension that a relatively narrow range of uNK cell activation is optimal with too little causing defective placentation resulting on clinical miscarriage, IUGR, and PE and too much causing biochemical losses) one would expect that is this were the case then C2/C2 embryos and not C1/C1 embryos would have the highest rate of loss in women with KIR B haplotypes. Finally, although many IVF clinicians may be tempted based on the present results to perform PGS on individual embryos for C1/C2 content, we argue instead for initial parental HLA-C genotyping that can be done in coordination with maternal KIR haplotyping. The reasons for this are two-fold: 1) As discussed above, it is important to not only consider the C1/C2 content of the embryo, but also the maternal or paternal source of the C1/C2 alleles. Therefore, HLA-C genotyping of the embryo alone is insufficient to adequately determine the risk for defective implantation associated with any particular embryonic C1/C2 allotype. 2) One-time tests for parental HLA-C genotypes will in many cases obviate the need for individual HLA-C genotyping of individual embryos. This will be the case whenever the paternal and maternal C1/C2 allotypes are each homoallotypic for either C1 or C2, thereby allowing prediction of the C1/C2 allotype of each embryo with certainty (in addition to being able to determine the source of the C1 and C2 allele of each embryo in cases where the maternal and paternal HLA-C alleles are homoallotypic for opposite C1/C2 allotypes). Of course, in cases where the maternal and/or paternal C1/C2 allotypes is/are heteroallotypic, then direct typing of individual embryos will also be required in addition to parental typing. We believe however, based on the Moffett group’s results that this is only indicated when a maternal KIR A haplotype or KIR B minus 2DS1 haplotype is present. We again thank the authors for this work and look forward to any response they may have to these comments.