Natural killer (NK) cells play a crucial role in the innate immune system, acting as a frontline defense against viral infections and tumor growth and contributing to tissue homeostasis. In the endometrium, uterine NK (uNK) cells are the predominant leukocytes, accounting for 30% to 40% of total leukocytes during the proliferative phase and up to 70% during the secretory phase [1]. These cyclical fluctuations suggest that steroid hormones may influence the migration of peripheral NK (pNK) cells into tissues [2].

In the peripheral circulation, uNK cells represent only about 1% of lymphocytes; they predominantly have the CD56^bright (CD56^bright CD16−) phenotype, whereas pNK cells predominantly exhibit the CD56^dim (CD56^dimCD16+) phenotype [3]. Despite a decrease in uNK cytotoxicity during the secretory phase, these cells create a favorable environment for embryo implantation and placental development. They contribute to the regulation of trophoblast invasion and enhance vascular remodeling by extravillous trophoblasts [4].

Therefore, aberrant increases in the number of pNK or uNK cells, or heightened cytotoxicity, could lead to reproductive failure such as recurrent pregnancy loss (RPL) and/or repeated implantation failure (RIF) [5-9]. Because measuring the number or percentage of uNK cells in clinical practice is difficult, the number or percentage of pNK cells is usually assessed. However, it remains unclear whether pNK cell status directly reflects uNK cell status. In women undergoing in vitro fertilization (IVF) and embryo transfer (ET), the percentage of pNK cells on the day of ET was significantly higher in women with failed cycles than in those with successful implantation [10]. Women with RIF also had a significantly increased number of pNK cells compared with fertile controls [11].

According to the recent European Society of Human Reproduction and Embryology (ESHRE) Guideline on RPL, there is insufficient evidence to support testing for uNK/pNK cells in patients with RPL [12]. In addition, the 2022 ESHRE guidelines and good practice recommendations under development do not recommend testing for uNK/pNK cells in patients with RIF [13]. Despite these guideline recommendations, in actual clinical settings pNK cell measurements are commonly performed in patients with RPL and RIF, and intravenous immunoglobulin G (IVIG) is frequently administered when pNK cell percentages are elevated. The Korean Society for Reproductive Immunology recommends IVIG for women with RPL and/or RIF who show an elevated percentage of pNK cells or increased NK cell cytotoxicity [14].

Reported cut-off values for the percentage of pNK cells in reproductive failure differ across studies. A pNK percentage >12% (among circulating mononuclear cells) has been used as a high-level cut-off associated with poor reproductive outcomes [15]. A detailed analysis of cellular immune marker cut-offs in Korean women with RPL defined a pNK percentage >16.1% as abnormal [16]. A study performed in Australia considered a pNK percentage >18% abnormal [17].

The proportion of uNK cells is higher in the mid-luteal phase—around the time of implantation—than in the proliferative phase; however, several studies have shown no variation in the pNK percentage between the follicular and luteal phases [18-20]. As a result, the need to limit pNK testing exclusively to the mid-luteal phase has been reduced [21].

An earlier study reported a lower pNK percentage in obese women [22]. Thus, when interpreting the percentage of NK cells, it is necessary to consider a woman’s body mass index (BMI). It has been reported that the number of uNK cells is not correlated with anticardiolipin or anti-peroxidase antibody levels in women with reproductive failure [23]. Nonetheless, research on whether women with an elevated pNK percentage exhibit abnormal levels of other autoimmune antibodies remains limited.

It has also been reported that the pNK percentage is negatively associated with serum anti-Müllerian hormone (AMH) levels, especially in women with pNK ≥18%, suggesting that higher pNK percentages may harm ovarian reserve or function [24]. However, in that study, such an association was not observed in women with pNK <18% [24].

In the present work, we investigated possible associations between the pNK percentage and age, BMI, serum AMH, and various autoantibodies or thrombophilia-related markers in women undergoing IVF.

We retrospectively identified 109 infertile Korean women in whom the pNK percentage was measured between May 2017 and November 2022 at Seoul National University Bundang Hospital (SNUBH). To reduce potential confounding effects on ovarian reserve, 16 women with a history of ovarian surgery and two women who had received chemotherapy for breast cancer were excluded. As a result, 91 women were included in the final analysis. None of the included participants had a prior diagnosis of autoimmune disease. Among the 91 women, 22 underwent pNK testing before undergoing IVF, and 69 had the test after experiencing at least one failed ET. The Institutional Review Board of SNUBH approved this retrospective study (IRB No. B-2310-857-103). Written informed consent by the patients was waived due to a retrospective nature of our study.

The mean±standard deviation age of the women was 37.7±4.1 years (range, 26 to 48). The following information was collected from medical records: number of previous failed ET cycles, age, BMI, serum AMH, antiphospholipid antibodies (lupus anticoagulant, anticardiolipin antibody, and anti-β2-glycoprotein 1 antibody), thyroid-stimulating hormone (TSH), anti-thyroglobulin (TG) antibodies, anti-microsome antibodies, fluorescent anti-nuclear antibody (FANA), antithrombin III, protein C, and protein S levels.

The pNK percentage was quantified using flow cytometry. First, mononuclear cells were separated from other blood cells by specific gravity, and then stained with fluorochrome-labeled monoclonal antibodies that bind leukocyte surface antigens. The stained cells were analyzed using a flow cytometer (BD FACSCanto II Flow Cytometer; BD Biosciences) with BD FACSCanto Clinical Software. The number of pNK cells was determined by gating on CD9+CD3−CD56^LowCD16^High lymphocytes, and results were reported as the percentage of total mononuclear cells. pNK cell testing was performed in women with a history of RIF, RPL, or in other situations in which it was clinically deemed necessary.

In this study, elevated pNK levels were evaluated using both ≥12% and ≥18% cut-offs, reflecting prior evidence of associations with adverse reproductive outcomes [15] and current clinical practice patterns in Korea.

Serum AMH levels were quantified using automated immunoassays—Access (Beckman Coulter) or Elecsys (Roche Diagnostics). Serum AMH was measured within 1 year of the pNK measurement. To ensure consistency, Access-derived AMH values were converted to Elecsys-equivalent values using the correction formula: Elecsys=(Access+0.05)/1.10 [25].

Plasma concentration of lupus anticoagulant was measured using a clot-based method, and anticardiolipin antibody (immunoglobulin G [IgG]/immunoglobulin M [IgM]) and anti-β2-glycoprotein 1 antibody (IgG/IgM) were measured by chemiluminescence immunoassay. The lupus anticoagulant was assessed using both a screening test and a confirmatory test. If the titer from the initial screening test was greater than 1.10, a confirmatory test was conducted. A titer of 1.2 or higher on the confirmatory test was considered positive for lupus anticoagulant. For anticardiolipin antibody and anti-β2-glycoprotein 1, a titer of ≥20 phospholipid units was considered positive.

Serum TSH was measured using an immune radiometric assay. Anti-TG and anti-microsome antibodies were measured by radio-immune assay. An indirect immunofluorescence technique was used to detect FANA. The normal reference range for TSH was 0.3 to 4.0 mIU/L.

Chromogenic and/or clot-based methods were used to measure antithrombin III, protein C, and protein S. The normal reference values were as follows: antithrombin III, 80–120; protein C, 70–140; and protein S, 59–120.

Statistical analyses were performed using SPSS ver. 29.0 (IBM Corp.). The Mann-Whitney U test was used to compare medians between two groups, and the Kruskal-Wallis test was used to compare medians among three groups. The chi-square test or Fisher exact test was used to compare proportions. Univariate analysis was performed to identify parameters associated with the pNK percentage, and variables with statistical significance were then entered into the multivariable analysis. The correlation structure was assessed using Spearman’s rank correlation. A p<0.05 was considered statistically significant.

Table 1 shows baseline clinical and laboratory characteristics of three patient groups categorized by the number of previous failed ET cycles: (1) women before entering IVF (n=22); (2) women with one or two failed ETs (n=28); and (3) women with three or more failed ETs (i.e., RIF group; n=41). The median percentage of pNK cells and the proportion of women with pNK percentages ≥12% or ≥18% were all similar among the three groups.

As shown in Table 2, none of the participants tested positive for lupus anticoagulant, nor did any show decreased levels of antithrombin III or protein C. Values for these markers were within normal ranges across all subjects. The remaining eight autoantibodies or thrombophilia-related markers did not differ among the three groups defined by the number of previous failed ET cycles.

Among the evaluated autoantibodies or thrombophilia-related markers, only anti-β2-glycoprotein 1 IgG antibody positivity was significantly associated with pNK levels: the median pNK percentage was significantly lower in women who were positive for anti-β2-glycoprotein 1 IgG compared with those who were negative (9.50% vs. 17.00%, p=0.022) (Table 3).

Table 4 shows the results of univariate analysis for five continuous variables (age, BMI, serum AMH, serum TSH, and the number of previous failed ETs). The percentage of pNK cells was negatively associated with BMI (r=–0.213, p=0.043) and serum AMH levels (r=–0.213, p=0.049), while age (r=0.122, p=0.268), serum TSH (r=0.009, p=0.930), and number of previous failed ETs (r=–0.079, p=0.458) were not correlated with the percentage of pNK cells. As shown in Table 5, multivariate analysis for three variables (BMI, serum AMH, and β2-glycoprotein 1 IgG antibody positivity) revealed that only serum AMH levels (B=–0.249, p=0.028) showed a significant negative association with the percentage of pNK cells.

A significant negative association between serum AMH levels and the pNK percentage is depicted in Figure 1.

In this study, we observed a significant negative association between serum AMH levels and the pNK percentage. These findings suggest that ovarian reserve may influence immune cell profiles. In a previous study by Hur et al. [24], a similar negative association was also found, but it was observed only in women with pNK ≥18%. They further reported that, in women with pNK ≥18%, the number of retrieved oocytes was negatively associated with pNK levels. The exact mechanism by which pNK cells may impact ovarian function remains unclear. Several studies have reported decreased basal serum AMH in women with autoimmunity. Vega et al. [26] suggested that the observation linking antiphospholipid antibodies to lower AMH supports the idea that non-specific autoimmunity may negatively affect ovarian reserve.

One possible explanation for the harmful effect of high pNK levels on ovarian function is the release of granzyme B, an exogenous serine protease, from cytoplasmic granules of NK cells and cytotoxic lymphocytes, which is expressed and activated in granulosa cells in the ovary. This action allows apoptotic signals to circumvent mitochondria-mediated apoptosis [27], potentially leading to excessive immune-cell-mediated apoptosis and damage to ovarian tissue.

In the present study, the median percentage of pNK cells and the proportion of women with pNK cell percentage ≥12% were similar among the three groups of previous failed ET cycles. Especially, the proportion of women with pNK cell percentage ≥12% in RIF group (78.0%) was similar to that in women before entering IVF (86.4%). Likewise, when using a stricter threshold of pNK ≥18%, there was no significant difference among the groups, further suggesting that elevated pNK levels may not be directly associated with implantation failure. This indicates that elevation of the pNK percentage is common in the non-RIF group and suggests that an elevated pNK percentage may not be involved in the etiopathophysiology of RIF. We also found similar levels of eight autoantibodies or thrombophilia-related markers among the three groups. This suggests that these markers are not elevated more in the RIF group than in the non-RIF group. Nevertheless, a few case-control studies with limited sample sizes have reported a higher prevalence of congenital and acquired thrombophilias in women experiencing implantation failure [28-30].

We observed a significant negative association between BMI and the pNK percentage. This inverse relationship aligns with previous studies; for example, Lynch et al. [22] reported decreased pNK levels in obese women. In our study, women with BMI <25.0 kg/m² had higher mean pNK percentages (17.57%±8.05%) than those with BMI ≥25.0 kg/m² (14.90%±6.33%). Thus, when interpreting the percentage of pNK cells, it is necessary to consider a woman’s BMI.

In this study, the median pNK percentage was significantly lower in women who were positive for anti-β2-glycoprotein 1 IgG compared with those who were negative. It is unclear why the pNK percentage is not associated with the presence or absence of anticardiolipin antibody IgG/IgM or anti-β2-glycoprotein 1 IgM, and why the pNK percentage is lower in anti-β2-glycoprotein 1 IgG-positive individuals. In fact, the positivity rates of these four antibodies were very low, so the clinical significance may be limited.

Anti-D1 antibody has been known as the primary binding site for anti-β2-glycoprotein antibody [31]. Manukyan et al. [32] demonstrated that, among antiphospholipid-positive patients, the anti-D1-negative group shows an increased pNK percentage and cytotoxicity. Another study demonstrated an association between the presence of anti-D1 antibody and decreased pNK percentage [33].

It is important to note several limitations. This was a retrospective study conducted among women undergoing, or planning to undergo, assisted reproductive technology, which may not represent the general population. Furthermore, the higher-than-expected proportion of elevated pNK cells in our cohort may reflect selection bias, as many participants underwent pNK testing due to a history of RPL or previous ET failures. Additionally, serum AMH was measured within 1 year of pNK testing; although this was the shortest feasible interval within routine clinical workflows, it remains a limitation when interpreting associations. Lastly, the relatively small sample size limits the generalizability of our findings, and larger, prospective cohort studies are warranted.

Nevertheless, our study adds to the expanding body of research on the role of pNK cells in reproductive health and highlights the need for further investigation. Although our findings demonstrate a strong association between pNK cell levels and serum AMH in women undergoing IVF, further research is required to determine the underlying mechanisms and causality.