Male origin microchimerism and brain cancer: a case–cohort study

Despite considerable research effort, causes of brain cancer are largely unknown. Male brain cancer predominance and reduced brain cancer risk with increasing parity among women, however, support a favourable role of pregnancy. We set out to determine whether fetal-origin microchimerism, namely the presence and long-term persistence of fetal cells, likely obtained via natural trafficking across the placenta during pregnancy, associates with reduced risk of brain cancer in women. Using a case–cohort design, we sampled 505 middle-aged women randomly at baseline in the Diet, Cancer and Health cohort (controls), and 73 women with incident brain cancer diagnosed during follow-up in the Danish Cancer Registry (cases). Male origin microchimerism was determined by presence of Y chromosome sequences in female blood samples. Data were analysed using weighted proportional Hazards regression, yielding Hazard Ratios with 95% confidence intervals. Compared with male origin microchimerism negative women, positive women had half the risk of developing brain cancer (Hazard Ratio = 0.50 [0.33–0.77]). Sensitivity analyses support that our findings are unlikely due to bias or chance. Here, for the first time, we demonstrate half the risk of brain cancer in male origin microchimerism positive compared with negative women. Our findings resemble those of previous studies of male origin microchimerism and other female cancers.


Introduction
Despite considerable research effort, established risk factors for brain cancer (BC) are few, and its aetiology remains poorly established (Ostrom et al. 2020). However, male BC predominance (Ostrom et al. 2019) and reduced BC risk with increasing parity among women (Chiu et al. 2012) support a favourable role of pregnancy. Cells originating from males are commonly present in peripheral blood of women (Kamper-Jørgensen et al. 2012a). This phenomenon-which is known as male origin microchimerism (MOM)-is thought to derive from naturally occurring traffic of cells from a male foetus to the pregnant woman over the placenta during pregnancy. In several reports, we and others have shown that MOM indeed associates with cancer in women (Cirello and Fugazzola 2016). A recent report documents that male origin cells pass the blood brain barrier, and that MOM resides in autopsied female human brain tissue samples (Chan et al. 2012). Also, substantial variation in MOM prevalence is observed between subtypes of female brain cancer (Broestl et al. 2018). Sparked by these observations, we set out for the first time to determine whether fetal-origin microchimerism, namely the presence and long-term persistence of fetal cells, likely obtained via natural trafficking across the placenta during pregnancy, associates with reduced risk of BC in middle-aged women in the Danish Diet Cancer and Health cohort and 20 + year follow-up for BC in the Danish national Cancer Registry. Male origin microchimerism, as a marker of fetal Mc, was determined by presence of Y chromosome sequences in female blood samples.

Material
Our study was done using the Danish population-based prospective Diet, Cancer and Health (DCH) cohort, comprising 57,053 cancer free men and women aged 50-64 at enrolment during 1993-1997 (Tjønneland et al. 2007). Only women were included in our study, because we used detection of Y chromosome sequences as a marker of MOM. At baseline, participants completed a questionnaire on lifestyle, demographic and reproductive factors, and visited a study centre for blood draw and anthropometric measures. Using the unique Danish identification number, participants were followed until 19 September 2017 for incident BC registered in the Danish Cancer Registry (DCR) (Gjerstorff 2011). Also, we identified hospital admissions with injury and poisoning in the Danish National Patient Register (DNPR) (Lynge et al. 2011) until 19 September 2017. Migration anddeath data, also until 19 September 2017 were obtained from the Danish Civil Registration System (Pedersen 2011).

Design
To optimize limited resources, we used the case-cohort sampling technique to identify all 85 women in DCH who developed BC (cases), and a randomly sampled sub-cohort of 600 women (controls). The case-cohort design is a highly effective design which mirrors results obtained in a standard cohort study, which is chosen over other epidemiological designs e.g. when exposure assessment is not viable for all cohort members due to cost considerations, and when there is an interest in several outcomes. The case-cohort sampling technique implies identification of controls at baseline among all cohort members, i.e. before cases fell ill. Thus, in our study, sampling allowed for controls who were free from BC at baseline to later develop BC, and thereby be included in the case as well as the control group. Also, due to random sampling at baseline, controls could be followed-up prospectively for any outcome over time using a standard cohort approach. Aside from the primary case-cohort analysis of association between MOM and BC, we followed-up the 505 control women for injury and poisoning in a standard cohort analysis. We anticipated injury and poisoning, respectively, to have no association with MOM, and thus these analyses were carried out to test for spurious associations.

Diagnoses
Brain cancer diagnosis was identified in DCR using ICD10 codes C70-72, including malignant neoplasm of the meninges (C70), malignant neoplasm of the brain (C71), and malignant neoplasm of the spinal cord, cranial nerves and other parts of the central nervous system (C72). Also in DCR, ICDO3 morphology code 99903 was used to exclude metastases, codes 94013, 94403, 94423, and 94513 were used to identify high-grade glioma (HGG), codes 93823, 93913, 94003, 94503, and 93913 were used to identify lowgrade glioma (LGG), and codes 95503, 93803, 99903, 99993 were used to identify other malignant neoplasms of the spinal cord, cranial nerves or other parts of the central nervous system. Injury was identified in DNPR using ICD8 codes 800-869, and ICD10 codes S00-S99 and T07-T14. Likewise, poisoning was identified in DNPR by ICD8 codes 960-989 and ICD10 codes T36-T65.

Laboratory analysis
From each woman, we obtained DNA purified from buffy coat equivalent to 180,000 maternal cells. To identify presence of male DNA, we applied a validated qPCR assay targeting Y chromosome gene DYS14 sequences, which should never be present in women. We ran 1.2 μg DNA obtained from buffy coat in 6-plicate with DNA corresponding to 30,000 cells per well. We performed qPCR as previously described by our group (Muller et al. 2015). For each run, we included six wells with no template controls. For a run to be accepted, all six wells should test negative. We dichotomized presence of MOM as positive versus negative. We denoted a sample MOM positive if at least one well tested positive, and MOM negative if all wells proved negative. To avoid male contamination, laboratory analyses were performed by female technicians who were blinded to case/ control status.

Statistical analysis
To characterize the study population, we calculated frequencies, percentages, median and inter quartile ranges (IQR). We evaluated whether MOM was associated with risk of BC by hazard ratios (HR) with 95% confidence intervals (CI), estimated using Cox proportional hazards regression. To take into account the overrepresentation of cases in the case-cohort setup, we used the weighing scheme suggested by Prentice (Prentice 1986) in the Cox proportional hazards regression model. To handle dependency between women included as both cases and controls, we used the robust covariance (sandwich) estimator suggested by Barlow (Barlow 1994) when estimating CI and p-values. In a separate analysis, we evaluated the association between MOM and subsequent risk of injury and poisoning, respectively, using standard Cox proportional hazards regression. We performed analyses in SAS version 9.4 using the PHREG procedure with time on study as the underlying time-scale. Finally, we performed a quantitative bias analysis estimating the E-value (VanderWeele and Ding 2017), to evaluate robustness of the association between MOM and BC to unmeasured or uncontrolled confounding.

Results
Of the 85 identified case women and the 600 control women, we excluded 12 (14.1%) and 95 (15.8%) women, respectively, due to missing blood samples or unsuccessful qPCR. Baseline characteristics for excluded women did not differ from those of included women (Supplementary Table 1). Baseline characteristics of the included women, according to case/control status are shown in Table 1. We included as cases 73 (12.6%) women with BC, and as controls 505 (87.4%) women. Among the 73 cases, 60 (82.2%) were registered with HGG, 10 (13.7%) had LGG, and 3 (4.1%) had other malignant neoplasms of the spinal cord, cranial nerves or other parts of the central nervous system. Two women were included in the case as well as the control group. Median follow-up time for controls was 21.2 years, and median age was 56.5 and 56.0 years, in cases and controls, respectively. Cases as well as controls had given birth to a median of 1 boy. There was no indication of differences between cases and controls (p < 0.05), with respect to the baseline characteristics shown in Table 1. Table 2 shows that MOM was detected in 34 (46.5%) of the 73 cases, and 333 (65.9%) of the 505 controls. This translates to a crude HR of 0.50 (95% CI 0.33-0.77), implying 50% reduced risk of developing BC among MOM-positive women. Results did not change when adjusting for age at enrolment (Table 2). In subgroup analyses comparing the 60 HGG cases to the 505 controls, we found a HR of 0.47 (95% CI 0.29-0.75), and similarly, we found HRs of 0.52 (95% CI 0.15-1.79) and 1.03 (95% CI 0.09-11.31), for the 10 LGG cases and the 3 other BC cases, respectively (Supplementary Table 3). Supplementary Fig. 1 shows the concentration of male cells per 10 6 female cells among participants, who tested positive for MOM. Among cases, the median concentration was 4.79 (IQR 2.01-9.50), and among controls, the median concentration was 3.20 (IQR 1.82-6.71).
Of the 505 control women, 354 (70.1%) were registered with an injury during follow-up, including injuries to head and neck (n = 56), thorax and abdomen (n = 17), shoulder, arm and hand (n = 150), hip, leg and foot (n = 122), and unspecified injuries (n = 9) (Supplementary table 3). The association between MOM and injury is shown in Supplementary Table 4. Injury occurred in 68.0% of MOM negative, and 71.2% of MOMpositive women, translating to a HR of 1.00 (95% CI 0.80-1.25). Of the 505 control women, 32 (6.3%) were registered with poisoning during follow-up, including poisoning by medical (n = 9) and non-medical substances (n = 23) (Supplementary Table 3). Poising was registered in 7.0% of MOM negative, and 6.0% of MOMpositive women, which translates to a HR of 0.86 (95% CI 0.42-1.75) (Supplementary Table 5). Thus, we observed no association between MOM and later injury or poisoning.
Finally, calculation of the E-value revealed that for unmeasured or uncontrolled confounding to cause a spurious HR of 0.50 given the observed data, the confounder or group of confounders should be associated with threefold increased risk of testing MOM positive, as well as threefold decreased risk of BC (data not shown).

Discussion
Our data suggest a halving of the risk of developing BC among middle-aged women who test MOM positive in their peripheral blood. We demonstrate that MOM does not associate with subsequent injury or poisoning, and that our results are very unlikely caused by unmeasured or uncontrolled confounding. In our view, together this lends to the notion that MOM has a biological role in protection against BC rather than being a spurious or occasional finding. Further, our findings are supported by previous findings that BC occurs more often in men than in women (Ostrom et al. 2020), that BC risk declines with increasing parity (Chiu et al. 2012), that asthma and allergy associates with reduced risk of BC (Amirian et al. 2016) (because asthma and allergy occurs more often in women throughout reproductive years (Osman 2003)), and that Varizella Zoster virus infection is inversely associated with BC (because shingles predominates in women (Fleming et al. 2004)). Although the leading source of MOM is pregnancy, other sources have been proposed, including spontaneous abortion, induced abortion, foeto-maternal haemorrhage, caesarean section, twinning, older brothers, transfusion, transplantation, and unprotected sexual intercourse. This is supported in a study by our group eliminating most of these sources, where we documented MOM in 13% of healthy girls aged 10-15 years (Muller et al. 2015). Compared with most previous MOM research, by design our study drastically minimized the risk of reverse causation by using prospective rather than retrospectively collected data (Kamper-Jørgensen et al. 2012a). As previously described by our group, we used a high quality measure of exposure to MOM by applying a validated assay to identify male Y chromosome sequences, and sample handling procedures and sample retest to avoid male *Probability of equal distributions among cases and controls based on the chi-square test contamination (Muller et al. 2015). Thus, we could not identify cells originating from females. To the best of our knowledge, no studies have reported differential effects of BC according to offspring sex, why we have no reason to believe that cells deriving from females would impact risk of BC differently than cells originating from males. Previously, exposure to female sex hormones has been associated with increased risk of malignant neoplasm of the meninges (Wigertz et al. 2006). In our study, the association between MOM and BC overall remained similar in 13 cases and 170 controls exposed to hormones (aggregate measure of ever using hormone replacement therapy, never using oral contraceptives, and nulliparity) (HR = 0.35, 95% CI 0.18-0.68) versus 21 cases and 163 controls unexposed to hormones (never using hormone replacement therapy, never using oral contraceptives, and parity) (HR = 0.56, 95% CI 0.31-0.99) (Supplementary Table 6).
Despite small numbers, we found similar associations with male origin microchimerism in all subgroups of BC. This is in contrast with a recent report by Broestl et al. in storff 2011). The quality of the BC diagnoses in DCR has not been ascertained, but is thought to be high. We adjusted our main analyses for time on study only. Besides, risk factors for BC include e.g. home and work exposures, familial history, infections, and electromagnetic fields (Savage 2018).
We have previously shown that MOM is poorly predicted by reproductive, lifestyle, hospital or clinic visit history, and other variables (Kamper-Jørgensen et al. 2012a), why these variables could not confound the association. We did not have information on individual level exposure to infection or electromagnetic fields, and were thus not able to control for possible associated confounding. We find it unlikely, however, that infections and/or electromagnetic fields induce a threefold increased chance of testing positive for MOM as well as a threefold reduced risk of developing BC, as suggested by the calculated E-value. In the case as well as the control group 1 in seven women were excluded due to missing blood samples or unsuccessful qPCR. Beyond reduced statistical precision, we do not believe that the exclusion substantially impacted our results, because characteristics of excluded women were similar to those of included women. Much remains in understanding the MOM phenomenon and its role in health and disease, why we find it untimely to propose clinical application of our findings. Reports of exposures associated with halving of the risk of disease are few and far between in BC research, however, why we hope the association between MOM and BC will be unfolded further.