DNA methylation profiles and cancer in children conceived after assisted reproductive technology.

First, using a method inspired from our previous systematic review comparing DNA methylation profiles of ART compared to naturally conceived individuals from birth to adulthood [ 7 ], we established a list of epigenetically dysregulated genes in ART individuals. Briefly, epigenome-wide associations studies (EWAS), matching the theme of DNA methylation profiles in ART, were identified by an extensive literature search in PubMed without year restriction until 2024 (Additional File 1 ). Eligible studies included original reports that measured genome-wide DNA methylation in any human embryonic tissue from birth to adulthood (thus excluding placenta and chorionic villus sampling as compared to Barberet et al. 2022), in populations conceived with all types of ART, compared to age-matched natural conceptions. With the latter, we collected information on differentially methylated positions (DMP), regions (DMR), and genes (DMG) associated with ART conception: CpG identifier, coordinates of regions analyzed, gene annotation, methylation difference between groups (Δβ), and hypo- or hyper-methylation. Because datasets were obtained from several sources and sequencing methods, genomic regions of DMP/DMR were re-annotated with the same database using R package GenomicFeatures [ 8 ]. The second step was to screen for dysregulations in ART that overlap with those seen cancer/leukemia profiles, focusing on identification of differentially methylated genes in ART that are also reported as genetically/epigenetically modified in cancer/leukemia. Our first cross-check of differentially methylated positions, regions, and genes in ART was performed with a general list of genes involved in cancers. Two sources were interrogated: the OncoKB™ cancer gene list (updated on 04/06/2024) [ 9 ] and cancer genes from the Cancer Gene Census (CGC) from the COSMIC database-release v100 [ 10 ] (1142 and 748 genes, respectively, corresponding to a total of 1246 individual genes) (Fig.  1 , Additional File 2 ). Fig. 1 Sources interrogated for the selection of cancer- and leukemia-related genes. B: B-cell. DMP: Differentially methylated position. T: T-cell Sources interrogated for the selection of cancer- and leukemia-related genes. B: B-cell. DMP: Differentially methylated position. T: T-cell The second comparison was tailored to leukemia genes, focusing on ALL and pediatric acute myeloid leukemia (AML) setting up a catalog of 532 genes, gathering: 1. a list of germline variants involved in hematopoietic malignancies [ 11 ], 2. a large whole-genome sequencing study with high statistical robustness (2754 children with ALL) [ 12 ], 3. a re-analysis study of somatic mutations in multiple cohorts of pediatric AML patients ( n  = 887) [ 13 ], and 4. genes from the WHO classification of ALL and AML [ 14 , 15 ] provided in three commercial genetic panels (NeoType® ALL Profile, Tapestri ALL panel and Eurofins AML NGS panel), detailed in Fig.  1 and Additional File 2 . The third cross-check examined genes/regions with known hypo- or hyper-DNA methylation in pediatric ALL (including T-ALL and B-ALL subtypes) to directly compare with the DNA methylation profiles of ART conceptions [ 5 , 16 , 17 ] (Fig.  1 , Additional File 2 ). We restricted the selection of ALL methylation studies to those with sample size ≥ 150 to retain the most robust evidence of DMG in ALL which yielded a total of 2642 DMG associated with ALL (B and T). Finally, we performed a comparison against a list of pre-leukemic DNA methylation regions from 41 twin pairs discordant for leukemia [ 6 ] and case–control cohort studies (sample size ≥ 150) corresponding to a list of 381 genes [ 18 , 19 ] (Fig.  1 , Additional File 2 ). All gene aliases from each list recorded in this study were updated to the official symbol. Leukemic and pre-leukemic regions were intersected with DMP/DMR coordinates to check for their co-occurrence in the two conditions compared (ART versus natural conceptions/leukemia versus healthy). We used Fisher’s exact tests to assess the enrichment of DMG in ART in the different comparisons performed.

Eighteen EWAS investigating DNA methylation with ART were considered in this study, representing 2369 methylome profiles from ART samples compared to 5806 natural conception profiles (Table  1 ). Two studies investigated DNA methylation at two different stages of life (birth-childhood/birth-adulthood) [ 20 , 21 ], two cohorts were investigated in different genomic regions in two complementary studies (autosomes and X chromosome analyzed in Håberg et al . 2022 and Romanowska et al . 2023, respectively; imprinted genes plus transposable elements and all other regions in Barberet et al . 2021 and Ducreux et al . 2021, respectively); and one study separated the ART cohort into IVF and intracytoplasmic sperm injection (ICSI) [ 22 ]. The main tissue studied was cord blood [ 16 , 23 – 32 ] followed by dried blood spots and peripheral blood [ 20 , 21 , 33 , 34 ] and buccal cells [ 35 , 36 ]. Methylome analysis was mainly studied at birth (14 cohorts) but also during childhood (2 cohorts), adolescence and adulthood (1 cohort each) (Table  1 ). DMP or DMR with ART were reported in all studies except three of them [ 24 , 28 , 34 ]. For the Chen et al. 2020 study, only genes common to the IVF ( n  = 3129) and ICSI ( n  = 3434) differential analyses compared to controls were retained ( n  = 623). Table 1 Characteristics of studies investigating DNA methylation with ART at genome-wide scale Study Tissue No. of CTL No. of ART Technique No. of DMP No. of DMR No. of DMG (≥ 1 DMR or ≥ 2DMP) ART technique Transfer (day/fresh-ET/FET) Other group comparison Birth Katari et al. 2009 Cord blood 13 10 GoldenGate 358 - 69 IVF fresh-ET and Day 3 (7/10) FET and Day 2 (3/10) – Camprubi et al. 2013 Cord blood 121 73 GoldenGate - 0 0 – – – Melamed et al. 2015 Cord blood 8 10 27 K 733 – 24 IVF – – Castillo-Fernandez et al. 2017 Cord blood 60 46 MeDIP-seq – 1 1 IVF or ICSI fresh-ET or FET – El Hajj et al. 2017 Cord blood 46 48 450 K 4 730 – 704 ICSI – – Gentilini et al. 2018 Cord blood 41 23 450 K 0 – 0 ICSI fresh–ET – Chen et al. 2020 (IVF) Cord blood 34 30 RRBS – 4 361 3 129 IVF fresh-ET or FET fresh-ET-FET: 9711 DMR (IVF) ICSI-IVF: 8495 DMR (fresh) Chen et al. 2020 (ICSI) Cord blood 34 26 RRBS – 4 831 3 434 ICSI fresh-ET or FET fresh-ET-FET: 9711 DMR (IVF) ICSI-IVF: 8495 DMR (fresh) Tobi et al. 2021 Cord blood 70 87 450 K 19 – 1 IVF or ICSI fresh-ET (77/87) or FET (10/87) Same results after FET exclusion IVF-ICSI: No difference Caramaschi et al. 2021 Cord blood 2439 205 450 K 5 – 1 IVF, ICSI, or other – – Håberg et al. 2022 Cord blood 983 962 EPIC 607 – 104 IVF (524/962) or ICSI (366/962) or unspecified/combination (72/962) fresh-ET (764/962) or FET (126/962) fresh-ET-FET: 3 DMP IVF-ICSI: No difference Romanowska et al. 2023 Cord blood 963 982 EPIC (X chr.) 5 (3 for♀ & 2 for♂) 15 (12 for♀ & 3 for♂) 14 IVF or ICSI or unspecified/combination fresh-ET or FET - Estill et al. 2016 Dried blood spots 43 94 450 K –  > 1000 54 ICSI (76/94) or IUI (18/94) fresh-ET (38/76) or FET (38/76) fresh-ET-FET: > 2000 DMR Novakovic et al. 2019 Dried blood spots 58 140 EPIC 2340 18 19 IVF (105/140) or GIFT (35/140) fresh-ET (75/105) or FET (30/105) IVF-GIFT: No difference fresh-ET-FET: No difference Yeung et al. 2021 Dried blood spots 520 158 EPIC 12 9 11 ICSI (100/158), IVF, assisted hatching, FET, GIFT, ZIFT, with or without the use of donor eggs or embryos fresh-ET or FET ICSI-CTL: 4 DMP Childhood Barberet et al. 2021 Buccal cells 12 36 EPIC (IG & TE) 401 33 38 IVF (15/36) or ICSI (21/36) fresh-ET IVF-ICSI: No difference Ducreux et al. 2021 Buccal cells 12 36 EPIC 127 16 15 IVF (15/36) or ICSI (21/36) fresh-ET IVF-ICSI: No difference Yeung et al. 2021 Peripheral blood 95 23 EPIC 1 1 1 ICSI, IVF, assisted hatching, GIFT, ZIFT, with or without the use of donor eggs or embryos fresh-ET or FET - Adolescence Penova-Veselinovic et al. 2021 Peripheral blood 1188 231 EPIC 0 0 0 IVF (160/231), ICSI (57/231), or unknown (14/231) fresh-ET (129/217) or FET (88/217) IVF-ICSI: No difference fresh-ET-FET: No difference Adulthood Novakovic et al. 2019 Peripheral blood 75 158 EPIC 0 4 4 IVF, GIFT, or unknown – – Studies in bold type include more than 100 ART samples. DMP: differentially methylated position. DMR: differentially methylated region. FET: frozen embryo transfer. fresh-ET: fresh embryo transfer. GIFT: gamete intrafallopian transfer. ICSI: intracytoplasmic sperm injection. IUI: intrauterine insemination. IVF: in vitro fertilization. OI: ovulation induction. ZIFT: zygote intrafallopian transfer Characteristics of studies investigating DNA methylation with ART at genome-wide scale fresh-ET and Day 3 (7/10) FET and Day 2 (3/10) fresh-ET-FET: 9711 DMR (IVF) ICSI-IVF: 8495 DMR (fresh) fresh-ET-FET: 9711 DMR (IVF) ICSI-IVF: 8495 DMR (fresh) Same results after FET exclusion IVF-ICSI: No difference fresh-ET-FET: 3 DMP IVF-ICSI: No difference IVF-GIFT: No difference fresh-ET-FET: No difference IVF-ICSI: No difference fresh-ET-FET: No difference Studies in bold type include more than 100 ART samples. DMP: differentially methylated position. DMR: differentially methylated region. FET: frozen embryo transfer. fresh-ET: fresh embryo transfer. GIFT: gamete intrafallopian transfer. ICSI: intracytoplasmic sperm injection. IUI: intrauterine insemination. IVF: in vitro fertilization. OI: ovulation induction. ZIFT: zygote intrafallopian transfer Combining all genes with a reported DNA methylation modification in at least 1 of the 18 ART EWAS (ART versus natural conceptions), 4866 DMG were identified (Fig.  2 ). This list was filtered to retain only genes showing DNA methylation differences in at least two studies, to ensure robustness of results, yielding 93 DMG (Fig.  2 A, Additional Files 3 and 4 ) [ 20 – 25 , 27 , 29 – 37 ]. This 93 DMG list was cross-checked with our set of 1246 cancer genes yielding few common targets (8 genes, Fig.  2 B) that were either oncogenes ( FGFR2, GNAS , IGF1R , and SFRP2 ) or tumor-suppressor genes ( CHD2, MGMT, PTPN14, QKI, and SFRP2 ) (Table  2 , Additional File 4 ; no significant enrichment, p  = 0.17). Also, we found differential methylation, with ART, for only five leukemia-related genes, as per genetic studies [ 12 ] ( ATP10A , CHD2 , FBRSL1, FGFR2, and SORCS1 ) (Fig.  2 B; no significant enrichment, p  = 0.08). There were no gene hotspots along the genome for the different cross-checks of DMG in ART with cancer/leukemia genes and DMG in leukemia (Fig.  2 C). Fig. 2 Quantitative summary of the cross-check of differentially methylated genes (DMG) in ART with cancer- and leukemia-related genes. A Venn diagram for the number of genes overlapping in the different comparisons. B Upset plot of the cross-check of differentially methylated genes (DMG) in ART with cancer- and leukemia-related genes. The list of the 93 DMG with ART is displayed on the right part of the plot (genes found in at least two different ART studies, marked by green dots linked together). First, the list of 93 DMG with ART is cross-checked with genes showing genetic or expression alterations in cancer and leukemia (genes overlapping are given in the right part of the plot, highlighted in dark blue). These general lists of genes involved in cancer and leukemia are mainly obtained from genetic studies and sequencing panels. Secondly, the list of 93 DMG with ART is cross-checked with genes showing DNA methylation alterations in ALL or pre-leukemic alterations (genes overlapping are given in the right part of the plot, highlighted in light blue). These lists of genes differentially methylated in ALL are obtained from the largest epigenetic studies (> 150 cases), and pre-leukemic regions are obtained from a twin study (MZ twins at birth among which one developed ALL later in life) or largest epigenetic studies (> 150 cases). C Genomic location of genes in the different sets of genes compared Table 2 List of DMG found in cancer- and leukemia-related genes ranked by frequency across ART studies No. of ART studies Genetic and expression variations in cancer Genetic and expression variations in leukemia DNA methylation modifications in leukemia DNA methylation modifications in pre-leukemia 2 CHD2, FGFR2, MGMT, PTPN14, QKI, SFRP2 CHD2, FBRSL1, FGFR2, SORCS1 ADARB2, CACNA1A, CELF4, DPP10, GALNT9, IRX2, ISL1-DT, L3MBTL1, LIN28B, LINC01115, MEG3, MGMT, NTM, PANTR1, PEG3, POTEA, QKI, SCAND3, SFRP2, SORCS1, SORCS2, SOX1, SYCP1, TAFA3, TP73, TSPEAR, UNCX, ZNF184, ZNF391 FBRSL1, ISL1-DT, MGMT, PTPRE, UNCX 3 IGF1R ATP10A INS-IGF2 IGF1R 4 PRSS16 5 GNAS GNAS, MEST Quantitative summary of the cross-check of differentially methylated genes (DMG) in ART with cancer- and leukemia-related genes. A Venn diagram for the number of genes overlapping in the different comparisons. B Upset plot of the cross-check of differentially methylated genes (DMG) in ART with cancer- and leukemia-related genes. The list of the 93 DMG with ART is displayed on the right part of the plot (genes found in at least two different ART studies, marked by green dots linked together). First, the list of 93 DMG with ART is cross-checked with genes showing genetic or expression alterations in cancer and leukemia (genes overlapping are given in the right part of the plot, highlighted in dark blue). These general lists of genes involved in cancer and leukemia are mainly obtained from genetic studies and sequencing panels. Secondly, the list of 93 DMG with ART is cross-checked with genes showing DNA methylation alterations in ALL or pre-leukemic alterations (genes overlapping are given in the right part of the plot, highlighted in light blue). These lists of genes differentially methylated in ALL are obtained from the largest epigenetic studies (> 150 cases), and pre-leukemic regions are obtained from a twin study (MZ twins at birth among which one developed ALL later in life) or largest epigenetic studies (> 150 cases). C Genomic location of genes in the different sets of genes compared List of DMG found in cancer- and leukemia-related genes ranked by frequency across ART studies By contrast, when we investigated DNA methylation signatures of ALL compared to ART-associated events, we found 33 genes that showed modifications in both conditions (Fig.  2 B, Table  2 ; significant enrichment, p  = 8.10 –9 ), albeit at different genomic positions/regions (Additional File 6 ). For six of these 33 genes, DNA methylation changes were in the same direction in ART compared to leukemia: PRSS16 (19 hypermethylated DMP in ALL; hypermethylated in more than 3 DMP in 4 ART studies), SYCP1 (3 hypermethylated DMP in ALL; hypermethylated in more than 2 DMP in 2 ART studies), ZNF184 (13 hypermethylated DMP in ALL; hypermethylated in more than 3 DMP in 2 ART studies), NTM , SCAND3 , and TP73 (8, 25, and 4 hypermethylated DMP, respectively, in ALL; ≥ 15 hypermethylated DMP in one ART study) (Additional File 6 ). For these six genes, the magnitude of the hypermethylation retrieved in ALL was between 30 and 50% compared with healthy controls, and between 1 and 30% in ART conceptions compared to those naturally conceived (Additional File 6 ). Finally, we identified DMP and DMR in ART, in the vicinity of previously described DMR which are associated with ‘pre-leukemia,’ (Fig.  2 B; significant enrichment, p  = 6.10 –3 ) [ 6 ]. Hypo/hypermethylated sites associated with ART were identified along the FBRSL1, IGF1R , ISL1-DT , MGMT, PTPRE , and UNCX coding sequence but the genomic location did not overlap with the ’pre-leukemia’ signature (Additional File 6 ). Direction of methylation variation was discordant for all these genes except for ISL1-DT , encoding a long non-coding RNA of unknown function (hypomethylated in pre-leukemia and ART).

The growth of the use of Assisted Reproductive Technology (ART) throughout the world is generating increasingly comprehensive data on the health of children conceived in this way. Evidence from large observational studies suggests that ART conceptions present a small but significant increased risk of neonatal complications, congenital malformations, and possible long-term effects in offspring [ 1 , 2 ]. In line with the literature, in the largest registry data-based study, we recently observed an increased risk for leukemia, mainly acute lymphoblastic leukemia (ALL), in children conceived by in vitro fertilization (IVF) within the cohort of French births ( n  = 8,562,306) [ 3 ]. These findings reinforce the need for further investigations to identify the possible causes. Genetic alterations involved in childhood cancers such as neuroblastoma, medulloblastoma, and leukemia can initiate prenatally [ 4 ]. Also, epigenetic regulation (i.e., DNA methylation) has been identified as an important cooperating mechanism in utero in pediatric cancers including leukemia [ 5 , 6 ]. It is suspected that periconceptional conditions linked to ART which occurs concomitantly to the epigenetic reprogramming undergone by both gametes and embryos, are responsible for DNA methylation modifications that induce epigenetic reprogramming anomalies. Recently, our team has shown that the use of ART is also often associated with modifications in the DNA methylation profiles at birth and during childhood [ 7 ]. In this study, we explored for the first time the hypothesis that increased risk of leukemia in the context of ART could be due to destabilization of DNA methylation profiles at key loci involved in leukemia onset. To investigate this, we first established a list of genes epigenetically dysregulated in ART individuals with a method inspired from our previous work comparing from epigenome-wide associations studies (EWAS), DNA methylation profiles of ART, and age-matched natural conceptions from birth to adulthood [ 7 ]. We next searched for evidence of overlap between differentially methylated loci seen in ART compared to a curated list of target genes for genetic/epigenetic alterations in cancer/leukemia (derived from public cancer databases, original germline and somatic sequencing studies, meta-analyses, and gene lists derived from international references).

Our study shows that ART is associated to DNA methylation changes across a subset of genes involved in various biological processes of importance to normal cellular function. Herein, we have derived a robust candidate gene catalog of 93 DMG in ART by restricting to genes found in at least two studies. Among these ART DMG, more than one-third ( n  = 33) were leukemia, and six were pre-leukemia DMG, all representing a significant enrichment. Furthermore, seven of the enriched genes ( NTM , PRSS16 , SCAND3 , SYCP1 , TP73 , ZNF184 , and ISL1-DT) showed concordant methylation between ART and leukemia/pre-leukemia. A further five of the ART DMG are known targets for somatic/germline alterations in leukemia: ATP10A , CHD2 , FBRSL1 , FGFR2 , and SORCS1 . Of note, the ART DMG were not significantly enriched in cancer genes, supporting the hypothesis that ART may not link broadly to any cancer type but rather to leukemia. The presence, among the differentially methylated targets common to ART and ALL of genes involved in immune ( PRSS16 ) and tumor-suppressor functions ( TP73 , for example) is noteworthy. PRSS16 encodes a thymic protease produced by cortical thymic epithelial cells and which plays a role in the development of CD4 T-cells [ 38 ] which are key for immune-mediated tumor suppression. Interestingly, ART-associated PRSS16 hypermethylation is maintained from birth to adulthood thus indicating that monitoring the methylation status at this region could be useful clinically [ 20 ]. Similarly, hypermethylation in ART, of TP73 which encodes the p73 protein, homologue of p53 is of interest. In its full length form, p73 acts as a tumor suppressor, similar to p53 [ 39 ]. The role of TP73 in ALL has not been explored in detail; however in a lymphoid leukemia cell line model, the inactivation of p73 is related to hypermethylation of TP73 [ 40 ]. We focused our results on genes with concordant methylation patterns, but those where methylation varies in opposite directions in ART and leukemia could still have physiopathological significance. Our finding related to aberrant methylation of imprinted loci, ATP10A , CHD2 , and GNAS, also known to occur in cancer, is also noteworthy [ 41 ]. Targeted approaches in ART have revealed loss of imprinting (LOI) at the KvDMR1 region [ 42 ]. KvDMR1 controls CDKN1C , a critical cell cycle regulator that is a target for inactivation in cancer [ 43 ]. Microarrays, used in most ART studies to date, only partially cover imprinted regions. LOI is thus likely underestimated in ART. Even though we applied a stringent filter to select robust ART-associated DMG, insufficient sample size in ART EWAS may limit assessment of the full effects of ART on DNA methylation. The number of DMGs is variable across studies but generally remains within a range of fewer than 100. Our selection of DMGs found in more than two studies helps avoid the bias associated with studies that identified a large number of DMGs. Ideally, a meta-analysis could highlight the best target genes for ART, but this could not be carried out here due to a lack of available data. The studies often include samples from different ART (fresh or frozen embryo transfer, conventional IVF, or with sperm microinjection or other method, Fig.  2 ), which makes it impossible to identify the effect of each method individually. However, where this was possible, the studies did not show any DNA methylation difference between conception by IVF or ICSI, and studies with large sample size found only minimal modifications associated with frozen embryo transfer (Fig.  2 ). ART alone may not explain all of the epigenetic changes associated with ART as there is also the possibility that underlying causes of parental infertility may have a role. For instance, polycystic ovarian syndrome, endometriosis, or advanced maternal age could generate additional epigenetic interferences consecutively due to oxidative stress, which have profound impact on DNA methylation [ 44 – 46 ]. It is hypothesized that pediatric cancers could derive from embryonic “accidents” in vivo, as the age of onset is often very early, suggesting that pre-leukemic profiles are set prenatally [ 47 , 48 ]. One of the characteristics of stem cells is their high sensitivity to environmental stimuli, and perturbations of developmental process could disturb their differentiation, leading to embryonic cancer cells [ 49 , 50 ]. Although ALL development has been linked to environmental exposures and DNA methylation changes are discussed, causal association remains to be clarified [ 51 , 52 ]. Some manipulations in ART (ovarian stimulation, in vitro culture) have the potential to disturb the set-up of DNA methylation profiles acquired during the epigenetic reprogramming occurring during the 1st days of development [ 53 ]. We can hypothesize that ART may increase the risk of such accidents and leukemia risk by directly targeting leukemia-related genes such as those identified here. In addition, in many cancers, a global DNA hypomethylation is observed as well as under ART where the genome is overall demethylated up to 1% as extrapolated from the transposable elements Alu and LINE-1 [ 54 ]. Oxidative stress experienced by embryonic cells in vitro as a consequence of the application of ART is a mechanism leading to global DNA methylation hypomethylation [ 55 ].

In summary, this study indicates that relatively few genes that are known targets for somatic/germline mutation in cancer undergo DNA methylation changes in individuals conceived through ART. By contrast, DNA methylation disturbances reported in leukemia/pre-leukemia represent a significant proportion of those associated with ART, thus raising the question of their role in ALL risk in ART-conceived individuals. With further validation in the future functional studies, as well as in studies incorporating hematologic surveillance and outcome analyses, this DMG target list could provide leads for future research and potentially serve as biomarkers of pediatric leukemia risk from conception and as targets for epigenetic therapy in the long run.

Additional File 1 . Flowchart of the study selection and the cross-check strategy. CTL: control. Additional file 2 . Additional file 3 . Additional file 4 . Additional file 5 . Additional file 6 . Additional File 1 . Flowchart of the study selection and the cross-check strategy. CTL: control. Additional file 2 . Additional file 3 . Additional file 4 . Additional file 5 . Additional file 6 .