Differentiation of genetic abnormalities in early pregnancy loss

Abstract

Objective

To characterize the types of genetic abnormalities and their prevalence in early pregnancy loss at different developmental stages. We hypothesized that the prevalence of genetic abnormalities in pregnancy loss would differ across developmental stages.

Methods

Women with a pregnancy loss at < 20 weeks’ gestation (n = 86) were enrolled at the time of diagnosis. Maternal tissue without a fetal component was found in 13 samples. Chromosomal microarray analysis (CMA) was performed on 74 samples (including two samples from a twin pregnancy); 15 were pre-embryonic (no visible embryo on ultrasound examination), 31 were embryonic (embryo; 6 + 0 to 9 + 6 weeks’ gestation) and 28 were fetal (fetus; 10 + 0 to 19 + 6 weeks’ gestation) losses. The twin pregnancy was found to be monochorionic diamniotic and was subsequently treated as a single sample in our analysis. Nine samples that underwent CMA were excluded from analysis because of 100% maternal-cell contamination.

Results

The overall prevalence of genetic abnormalities differed across developmental stages (9.1% pre-embryonic, 69.2% embryonic and 33.3% fetal; P < 0.01). This difference persisted when comparing pre-embryonic with embryonic samples (P < 0.01) and embryonic with fetal samples (P = 0.02) but not pre-embryonic with fetal samples (P = 0.12). Additionally, the prevalence of aneuploidy differed significantly across developmental stages (0.0% in pre-embryonic samples vs 65.4% in embryonic samples vs 25.9% in fetal samples, P < 0.001). Abnormalities were most common in embryonic cases, followed by fetal and then pre-embryonic. Maternal cell contamination (MCC) was noted in 47.4% of 46,XX cases assessed.

Conclusions

Genetic abnormalities detected by CMA are more likely to occur in the embryonic period than in pre-embryonic or fetal stages. MCC is common in early pregnancy loss and should be excluded when results demonstrate a 46,XX karyotype.

INTRODUCTION

The most common cause of sporadic pregnancy loss is cytogenetic abnormalities, which are found in 50–70% of early pregnancy losses^1,2. Of these, the most common are autosomal trisomies (60%), monosomy X (20%) and polyploidy (20%)³. In an overwhelming majority, the parental karyotype is normal, indicating a de novo aneuploidy in the conceptus.

Cytogenetic analysis of pregnancy-loss specimens has been performed traditionally by karyotyping. Because this technique requires metaphase fetal cells, the specimens must be grown in culture; however, this poses some disadvantages: cells may fail to grow because of marginal viability or maternal cells may overgrow. Either can lead to failure to obtain an accurate karyotype, which is very common in cases of pregnancy loss⁴.

Chromosomal microarray analysis (CMA) has become widely available⁵. CMA allows for analysis of the entire genome without the need to culture cells and with a greater resolution than traditional karyotyping^3,5–9. In addition to detecting the larger abnormalities visible by karyotyping, CMA can identify smaller genetic gains and losses that are not detected by karyotyping. Such microduplications or microdeletions are generally referred to as ‘copy number variants’ (CNVs)⁹. CNVs are likely to be pathogenic when they are large, involve human genes and are usually not found in the parents or in databases of structural variation in healthy control samples^10–12. Moreover, although much of the genome has been characterized, there remain CNVs of unknown clinical relevance; these have been labeled ‘variants of unknown significance’ (VOUS).

The overall proportion of pregnancy loss at varying gestational ages that is caused by genetic abnormalities is unknown. Previous studies have not distinguished between the developmental stages of early loss and have used karyotyping rather than CMA^2,13. We hypothesized that the prevalence of genetic abnormalities across developmental stages in early pregnancy loss would differ, with the highest prevalence of genetic abnormalities being in the pre-embryonic group. Furthermore, we hypothesized that by using CMA we would detect a greater proportion of genetic abnormalities than that previously reported from karyotyping. This information can be used to guide future research in early pregnancy loss and to assist providers in counseling patients.

METHODS

This was a prospective cohort study conducted at the University of Utah and Intermountain Healthcare (UT, USA) between May 2012 and June 2013. All women with a pregnancy loss prior to 20 weeks’ gestation and seeking medical care at one of the participating institutions were eligible for participation. Another eligibility criterion was the availability of products of conception for DNA extraction, which were either brought in by the patient from home or were collected at the time of a uterine evacuation procedure. Although efforts were made to offer participation to all patients meeting the eligibility criteria, the study utilized convenience sampling as consecutive patients were not enrolled. Maternal medical records were reviewed and data were extracted by one investigator (S.T.R.). Gestational age was established based on ultrasound findings, and patients were grouped into the following categories: those with an empty gestational sac or a yolk sac only were defined as having a pre-embryonic loss; those with a detectable embryo, measuring less than the corresponding size for 10 weeks’ gestation, were defined as having an embryonic loss; and those with a crown-rump length > 30 mm or other biometry (e.g. biparietal-diameter measurement), consistent with that at ≥ 10 weeks’ gestation, were defined as having a fetal loss¹⁴.

At the time of enrollment, we also sought to obtain maternal DNA in every case, from either a saliva sample (Oragene, DNA OG-500; DNA Genotek Inc., Murrieta, CA, USA) or a blood sample drawn into an EDTA tube. Extracted maternal DNA was stored for analysis if we needed to exclude potential maternal cell contamination in cases of a normal female result without macroscopically identifiable fetal parts.

A portion of the products of conception, either from samples brought in by the patient after spontaneous passage at home or obtained at the time of uterine surgical evacuation, was collected in a 10-mL container. Samples were subsequently washed with normal saline and were examined under a dissecting microscope to identify chorionic villi or obvious embryonic or fetal parts for further processing.

In addition, we recruited eligible women from the patients enrolled in the Effects of Aspirin in Gestation and Reproduction (EAGeR) trial into the cohort¹⁵. These included those enrolled at the Utah site with products of conception available from a pregnancy loss occurring at < 20 weeks’ gestation and with precise documentation of the gestational age at the time of pregnancy loss. Products of conception were stored at −80 °C, as previously described¹⁵.

The study was approved by the Institutional Review Boards at the University of Utah Health Sciences Center and Intermountain Medical Center. All participants, including those enrolled in the EAGeR trial, gave written informed consent for the current study.

DNA was extracted from the tissue samples using standard techniques, digested, ligated and amplified, and then purified using a bead-based method in which DNA was incubated with magnetic carboxyl-coated beads. After the binding of DNA to the beads, the beads were immobilized using a magnetic plate, washed with 75% ethanol and purified. The sample was next fragmented and labeled with biotin. Labeled DNA fragments were denatured and hybridized to CytoScan SNP arrays (Affymetrix, Santa Clara, CA, USA), according to the manufacturer’s instructions, for 16–18 h. Arrays were then washed by saline–sodium–phosphate–EDTA-based wash buffers and stained by stacking layers of streptavidin–phycoerythrin and biotinylated anti-streptavidin IgG¹⁶.

Arrays were scanned and then analyzed with Chromosome Analysis Suite software (ChAS v2.1, Affymetrix). CNVs were interpreted using data from publicly available databases in order to characterize known genes that have been mapped to the regions of interest: the Database of Genomic Variants¹⁰, the Online Mendelian Inheritance in Man (OMIM) database¹², the International Standards for Cytogenomic Arrays (ISCA) database¹¹ as well as internal databases and the peer-reviewed medical literature. CNVs within intronic regions and CNVs containing no genes were considered to be benign. CNVs documented in phenotypically normal individuals and not associated with an abnormal phenotype were considered to be normal variants and benign.

For the purpose of our analysis, we distinguished CMA results as pathogenic or potentially pathogenic, including VOUS⁹. CNVs were categorized as VOUS if there was a potentially adverse phenotypic consequence in liveborn individuals. The prevalence of genetic abnormalities was summarized according to gestational-age category. Fisher’s exact test was used to evaluate prevalence across categories and within subcategories of genetic abnormalities. All analyses were conducted in Stata version 12 (StataCorp LP, College Station, TX, USA)¹⁷. Given the previously published prevalence of aneuploidy in early pregnancy loss using karyotyping and the improved sensitivity of microarray, we hypothesized that there would be a difference of 40% in the prevalence of genetic abnormalities between groups, using pre-embryonic loss as the reference point. In order to achieve 80% power, with an alpha of 0.05, to detect a 40% difference in the prevalence of genetic abnormalities between groups, a total sample size of 59 patients was required

RESULTS

Seventy-eight patients were enrolled prospectively into the cohort, one of whom had a twin gestation, yielding a total of 79 samples. In addition, eight participants of the EAGeR trial were enrolled, providing a total of 87 individual products of conception from pregnancy loss at < 20 weeks’ gestation.

In 13 (14.9%) of the 87 products-of-conception samples obtained, we could not identify placental villi or fetal tissue with certainty and these were therefore excluded. CMA was performed on the remaining 74 samples. These included 15 pre-embryonic losses, 31 embryonic losses and 28 fetal losses. Demographic characteristics according to gestational-age category are shown in Table 1. There were no significant differences in maternal age at enrollment, race or ethnicity, or history of recurrent pregnancy loss among groups.

Table 1.

Characteristics of 73 women who had an early pregnancy loss who had a sample that underwent chromosomal microarray analysis, according to developmental stage at time of pregnancy loss

	Developmental stage
Parameter	Pre-embryonic (n = 15)	Embryonic (n = 31)	Fetal (n = 27)	P
Age (years)	31 (23–45)	33 (22–44)	29 (18–39)	0.18
Race/ethnicity				0.56
Non-Hispanic white	13 (86.7)	26 (83.9)	22 (81.5)
Hispanic/Latino	2 (13.3)	2 (6.5)	3 (11.1)
Black	0	0	2 (7.4)
Asian	0	2 (6.5)	0
Other	0	1 (3.2)	0
Total pregnancy losses*	2 (1–4)	2 (1–5)	2 (1–4)	0.20
Recurrent pregnancy loss (≥ 2 losses)*	5 (33.3)	10 (32.3)	5 (18.5)	0.38

Data are shown as mean (range) or n (%).

^*Including this study.

The twin gestation was found to be monochorionic diamniotic and therefore this pregnancy loss was subsequently treated as a single sample in our analysis. After CMA was performed, nine samples (four pre-embryonic and five embryonic) were found to have 100% maternal cell contamination (MCC) and were subsequently excluded from the analysis. Therefore, we analyzed a total of 64 products of conception.

Table 2 presents the prevalence of genetic abnormalities detected by CMA. The overall prevalence of aneuploidy and pathogenic CNV or VOUS was 28 (43.8%) of 64 cases. This rate differed across the developmental stages; these genetic abnormalities were most common in cases of embryonic loss followed by fetal and then pre-embryonic loss, with a prevalence of 18 (69.2%) of 26, nine (33.3%) of 27 and one (9.1%) of 11, respectively (overall comparison, P < 0.01). This difference persisted when comparing pre-embryonic with embryonic losses (P < 0.01) and embryonic with fetal losses (P = 0.02) but not pre-embryonic with fetal losses (P = 0.12).

Table 2.

Genetic abnormalities in 64 specimens of pregnancy loss determined by chromosomal microarray analysis, according to developmental stage at time of pregnancy loss

	Developmental stage
Abnormality	Pre-embryonic (n = 11)	Embryonic (n = 26)	Fetal (n = 27)
Overall prevalence of abnormalities*	1 (9.1)	18 (69.2)	9 (33.3)
Any aneuploidy	0	17 (65.4)	7 (25.9)
Complex rearrangement chr7	1 (9.1)	0	0
Autosomal trisomy	0	10 (38.4)	4 (14.8)
Trisomy 8	0	2 (7.7)	0
Trisomy 9	0	1 (3.8)	0
Trisomy 13	0	0	2 (7.4)
Trisomy 15	0	2 (7.7)	0
Trisomy 16	0	1 (3.8)†	0
Trisomy 18	0	0	2 (7.4)
Trisomy 21	0	1 (3.8)	0
Trisomy 22	0	3 (11.5)	0
Sex chromosome abnormality	0	4 (15.4)	1 (3.7)
45,X	0	4 (15.4)	0
47,XXX	0	0	1 (3.7)
Triploidy	0	3 (11.5)	2 (7.4)
69,XXY/70,XXY,+18	0	1 (3.8)†	0
69,XXX	0	2 (7.7)	2 (7.4)

Data are given as n (%).

^*Includes variants of unknown significance.

^†Includes one with mosaicism. chr, chromosome.

A separate analysis was performed, excluding the two cases with VOUS. The resulting prevalence of abnormalities remained different across gestational-age categories: one of 11 (9.1%) pre-embryonic samples, 18 of 26 (69.2%) embryonic samples and seven of 25 (28.0%) fetal samples (P < 0.01). An additional analysis was performed excluding the 10 cases with a 46,XX result and no available maternal DNA for MCC testing. The resulting prevalence of abnormalities was, again, different across gestational-age categories: one of eight (12.5%) pre-embryonic losses, 18 of 25 (72.0%) embryonic losses and seven of 21 (33.3%) fetal losses (P < 0.01).

Aneuploidy was the most common abnormality, occurring in 24 of 64 (37.5%) cases. The prevalence of aneuploidy found using CMA also differed across developmental stages: 0 of 11 (0.0%) were pre-embryonic losses, 17 of 26 (65.4%) were embryonic losses and seven of 27 (25.9%) were fetal losses (P < 0.001). Most aneuploidy cases were autosomal trisomies, followed by triploidy (Table 2). Other common abnormalities were observed; however, statistical comparison between gestational-age groups was not performed because of low numbers of each specific abnormality.

We found one case with pathogenic CNVs^10–12; a pre-embryonic sample that had a complex rearrangement of chromosome 7 (a 245.7-kilobase pair (kbp) duplication of the terminal portion of 7p and a 62.3-megabase deletion of the terminal portion of 7q). After karyotype analysis of the parents, this was found to be a de novo rearrangement.

The prevalence of VOUS in our study population was two of 64 (3.1%). Both samples were categorized as fetal losses. One sample had a 34.1-kbp deletion on the long arm of chromosome 16, which includes ZNF778, a candidate gene for autism and variable cognitive impairment in liveborn individuals with the 16q24.3 microdeletion syndrome. The other sample with VOUS had a 567-kbp duplication on the short arm of chromosome 2 that has been previously categorized as a VOUS¹⁸. There is no known association with intrauterine demise for either VOUS.

Three samples (4.8%), one from each of the three gestational age categories, failed to yield a result by CMA because of artifact obviating interpretation. These technical failures were probably caused by poor quality of the extracted DNA samples.

DISCUSSION

Our study has shown that the proportion of pregnancy losses with a genetic abnormality identified by CMA is significantly different at different developmental stages. Surprisingly, the prevalence of genetic abnormalities was quite low in pre-embryonic losses (9.1%). In contrast, over 50% of embryonic losses and almost a third of early fetal losses (10–20 weeks’ gestation) had genetic abnormalities.

These data do not support the hypothesis that the earlier in gestation the pregnancy loss, the more likely it is that the fetus will have aneuploidy¹. This has been an attractive and biologically plausible hypothesis because it is intuitive that the more abnormal the conceptus, the quicker the pregnancy is likely to fail. Indeed, some genetic abnormalities are considered so severe (trisomy 16) that they only appear in pregnancy losses that occur relatively early in gestation. In contrast, less severe conditions, such as trisomies 13,18 and 21, are found in live births as well as pregnancy losses³.

Many studies evaluating pregnancy-loss specimens by microarray do not distinguish between developmental stages of the pregnancy loss^2,13,19. A recently published large study by Levy et al. found high rates of MCC and similar rates of genetic abnormalities to those in our study but did not include data on gestational age at loss¹⁹. Reporting gestational age alone is inadequate because there may be a considerable lag between pregnancy failure and the clinical symptoms or signs of pregnancy loss. This is one of the few studies to distinguish among pre-embryonic, embryonic and fetal losses. In addition, the use of karyotyping in older studies was a potential source of bias. Cells from very early losses may have been less likely to yield results as a result of culture failure, causing a majority of cases in older studies to be embryonic or fetal losses. The use of CMA allowed us to obtain results in cases of very early loss, despite the small number of fetal (and potentially non-viable) cells. Numerous studies have shown increased yields with CMA compared with karyotyping for assessing genetic abnormalities in cases of pregnancy loss across gestational ages^7,9,20.

Two recent studies have attempted to characterize genetic abnormalities in pregnancy loss according to developmental stage. Angliolucci and colleagues assessed ‘morphologic types’ in 156 cases of early pregnancy loss²¹. They noted varying prevalence of aneuploidy with each of six morphologic types based on the appearance of the sac, yolk sac and embryo. It was noteworthy that 13 (72.2%) of 18 pre-embryonic losses had abnormal karyotypes²¹. Similarly, Robberecht and coworkers performed hystero-embryoscopy and cytogenetic analysis on 51 pregnancy losses²². They noted abnormalities in eight (88.9%) of nine pre-embryonic losses based on hysteroscopy. The relationship between pre-embryonic losses based on sonogram vs hysteroscopy has not been evaluated. One possible reason for the lower prevalence of abnormalities noted in pre-embryonic losses in our study was the inclusion of some women with recurrent pregnancy loss. A Japanese cohort reported an increased chance of chromosomally normal pre-embryonic and embryonic losses in women with recurrent pregnancy loss compared with those with a sporadic loss²³.

We also noted different abnormalities at various developmental stages. As anticipated, the abnormalities identified in fetal deaths were those that have been reported in live births, such as trisomies 13 and 18. In contrast, embryonic losses included many aneuploidies not typically seen in live births, such as trisomies 8, 9, 15 and 16.

One limitation of the study was an insufficient sample size to perform meaningful subanalyses (e.g. stratification by maternal age or recurrent loss). A further limitation was that we only ascertained losses in women seeking medical care, which was a potential source of bias. These factors may have led to inclusion in this study of a higher proportion of women with later losses and recurrent loss in pregnancy than found in the general population. The strengths of the study include a considerable number of cases across gestational ages, relative to other studies, and an adequate sample size to test our primary hypothesis. All cases had excellent characterization of developmental stage based on obstetric sonogram. Furthermore, maternal cell contamination was excluded from CMA results using microsatellite genotyping by polymerase chain reaction.

For a clinician, the ability to identify whether or not a pregnancy loss was caused by a genetic abnormality is important. First, determination of the reasons for the loss facilitates grieving and closure for families, and assists with providing guidance for subsequent pregnancies. Second, couples with recurrent pregnancy loss often embark on expensive diagnostic evaluation and potentially morbid treatments for unexplained pregnancy loss. This could be avoided if the etiology of the losses was known to be genetic. These data may help to counsel women regarding the probability of genetic abnormalities. This is an important issue in the USA as karyotyping and/or CMA are often not covered by medical insurance. The recent committee opinion from the American Congress of Obstetricians and Gynecologists affirmed that the data regarding the use of CMA in pregnancy loss < 20 weeks’ gestation were lacking²⁴. If these data are confirmed in larger cohorts, it may be possible to screen for most cases of genetic abnormalities after 10 weeks’ gestation using less expensive methods that assess a limited number of abnormalities. In couples with recurrent pregnancy loss, losses tend to recur at the same gestational age/developmental stage²⁵; therefore, it will be of interest to assess whether chromosomal abnormalities are recurrent at specific developmental stages as well.

Our high prevalence of MCC is noteworthy. Indeed, nine (47.4%) of 19 cases of apparently normal female results were caused by 100% MCC. These included several embryonic as well as pre-embryonic losses. A similarly high prevalence of MCC was noted in a cohort of over 1000 samples²⁶. Our findings are even more concerning as all samples were collected by trained individuals who attempted to isolate fetal tissue. Clinicians should be suspicious of MCC in all cases of normal female karyotype or microarray in early pregnancy loss.

We noted varying rates of genetic abnormalities in pregnancy losses at different developmental stages. Abnormalities were most common in embryonic losses and were uncommon in pre-embryonic losses. Future studies should carefully characterize gestational age and developmental stage at pregnancy loss, so that we may continue to refine our understanding of the genetic causes of pregnancy failure.