Analysis of CDKN1C in fetal growth restriction and pregnancy loss

Background: Cyclin-dependent kinase inhibitor 1C (CDKN1C) is a key negative regulator of cell growth encoded by a paternally imprinted/maternally expressed gene in humans. Loss-of-function variants in CDKN1C are associated with an overgrowth condition (Beckwith-Wiedemann Syndrome) whereas “gain-of-function” variants in CDKN1C that increase protein stability cause growth restriction as part of IMAGe syndrome ( Intrauterine growth restriction, Metaphyseal dysplasia, Adrenal hypoplasia and Genital anomalies). As three families have been reported with CDKN1C mutations who have fetal growth restriction (FGR)/Silver-Russell syndrome (SRS) without adrenal insufficiency, we investigated whether pathogenic variants in CDKN1C could be associated with isolated growth restriction or recurrent loss of pregnancy. Methods: Analysis of published literature was undertaken to review the localisation of variants in CDKN1C associated with IMAGe syndrome or fetal growth restriction. CDKN1C expression in different tissues was analysed in available RNA-Seq data (Human Protein Atlas). Targeted sequencing was used to investigate the critical region of CDKN1C for potential pathogenic variants in SRS (n=66), FGR (n=37), DNA from spontaneous loss of pregnancy (n= 22) and women with recurrent miscarriages (n=78) (total n=203). Results: All published single nucleotide variants associated with IMAGe syndrome are located in a highly-conserved “hot-spot” within the PCNA-binding domain of CDKN1C between codons 272-279. Variants associated with familial growth restriction but normal adrenal function currently affect codons 279 and 281. CDKN1C is highly expressed in the placenta compared to adult tissues, which may contribute to the FGR phenotype and supports a role in pregnancy maintenance. In the patient cohorts studied no pathogenic variants were identified in the PCNA-binding domain of CDKN1C. Conclusion: CDKN1C is a key negative regulator of growth. Variants in a very localised “hot-spot” cause growth restriction, with or without adrenal insufficiency. However, pathogenic variants in this region are not a common cause of isolated fetal growth restriction phenotypes or loss-of-pregnancy/recurrent miscarriages.


Introduction
Cyclin-dependent kinase inhibitor 1C (CDKN1C, also known as P57/kip2) (OMIM 600856) is a key negative regulator of cell proliferation that is encoded by a paternally imprinted (maternally expressed) gene on the short arm of chromosome 11 (11p15.4) in humans (Stampone et al., 2018).
Consistent with its role in growth and development, maternallyinherited loss-of-function variants in CDKN1C are found in approximately 5-10% of individuals with the "overgrowth" condition, Beckwith-Wiedemann Syndrome (BWS) (OMIM 130650) (Eggermann et al., 2014). Clinical features of BWS include macrosomia, hyperinsulinism and adrenal tumors.
In contrast, gain-of-function variants in CDKN1C have been shown to cause growth restriction as part of IMAGe syndrome (OMIM 614732) (Arboleda et al., 2012). IMAGe syndrome is characterised by fetal/ Intrauterine growth restriction, Metaphyseal dysplasia, Adrenal hypoplasia and Genital anomalies (in males, usually relatively mild hypospadias and undescended testes) as well as additional features such as hearing loss and hypercalciuria (Bennett et al., 1993;Vilain et al., 1999).
To date, children with IMAGe syndrome have all been found to harbour pathogenic single nucleotide variants (SNVs) in a very specific region of the PCNA-binding domain of CDKN1C (Arboleda et al., 2012;Cabrera-Salcedo et al., 2017). These changes potentially lead to increased activity through increasing protein stability, thereby preventing cell cycle progression into S phase (Borges et al., 2015;Hamajima et al., 2013).
More recently, SNVs in the PCNA-binding domain of CDKN1C have been reported in families with maternally-inherited fetal growth restriction (FGR) without adrenal insufficiency and in familial Silver-Russell syndrome (SRS) (OMIM 180860) (Brioude et al., 2013;Kerns et al., 2014;Sabir et al., 2019). SRS is characterised by variable clinical features including fetal and post-natal growth restriction, relative macrocephaly, feeding difficulties and characteristic facies. SRS is also described as phenotypically and genotypically opposite to BWS and approximately half of the molecular anomalies are attributed to Chr11p15.5 imprinting clusters, including several individuals with maternal duplication of the locus containing CDKN1C (Bonaldi et al., 2011;Boonen et al., 2016;Schönherr et al., 2007). These findings suggest that the growth restriction phenotype associated with CDKN1C may be more variable and adrenal insufficiency is not always present.
The aim of this study was therefore to review published CDKN1C variants associated with FGR/IUGR phenotypes, to study CDKN1C expression in different tissues, and to analyse the critical region in CDKN1C in a range of growth restriction and adverse pregnancy phenotypes, with a hypothesis that severe restriction of feto-placental growth may, in some situations, result in pregnancy loss or recurrent miscarriage.
Analysis of CDKN1C expression RNA-Seq data for CDKN1C expression was obtained with specific permission from the Human Protein Atlas (Human Protein Atlas available from www.proteinatlas.org) and re-drawn in R (version 3.4.2) (Uhlen et al., 2015).

Study cohorts
The following growth restriction cohorts were included in this study: 1) SRS (n=66) (isolated, non-familial) diagnosed on consensus criteria and where maternal uniparental disomy or H19/IGF2:IG-DMR (also known as ICR1 or IC) hypomethylation had been excluded (Wakeling et al., 2017); 2) FGR (n=37) (isolated, non-familial) defined as birth weight less than the 3rd centile, as part of the Baby Bio Bank cohort (UCL-GOS Institute of Child Health & St Mary's Imperial College London) (Leon et al., 2016). Additional analysis was undertaken in DNA from 3) products of conception (POC) (n=22) where there had been a spontaneous loss of pregnancy and 4) women who had a history of recurrent miscarriages (n=78) (at least three miscarriages) where an underlying cause was not known (Baby Bio Bank). An overview of these cohorts is provided in Table 1.

Consent
Ethical Committee approval for the Baby Bio bank was obtained from the Trent Derby Ethics Committee (09/H0405/30) and Ethical Committee approval for the Silver Russell trios was from GOSH Research Ethics Committee (REC No. 1278). Written informed consent was obtained from participants or parents. DNA was extracted from blood lymphocytes, placental tissue or products of conception, as appropriate.

Amendments from Version 1
In the revised version of this paper, we have: 1) Updated the literature review with more search terms and modified Figure 1/ . In brief, genomic DNA samples (100ng) underwent enzymatic shearing followed by end repair and dA tailing to ligate molecular identifier adapters. DNA was purified using Agencourt AMPure beads (Beckman Coulter Inc., USA) to remove nonligated adapters, and amplified using adapter primers. Libraries were hybridised with the customised probes, washed, and then targeted library DNA sequences were amplified. After washing and quantification, the libraries were sequenced on a MiSeq platform (Illumina Inc.). Fastq files were analysed on a bioinformatics pipeline provided by Nonacus and variant calling undertaken in Platypus (v 0.8.1). Inspection of BAM files was done in the Integrative Genomics Viewer.

Single nucleotide variants in CDKN1C
Review of available data revealed seven publications describing isolated individuals (7) or families (5)    Variants in CDKN1C associated with familial Silver-Russell syndrome or growth restriction but normal adrenal function were found towards the carboxyl-terminal region of this "hot-spot" domain (p.Arg279Leu, p.Arg281Ile) (Figure 1).
Analysis of population data from the gnomAD browser showed a complete absence of variants in the key codons listed. Very rare heterozygous SNVs were found that are predicted to cause p.Ala277Val (11:2905355G>A;1 in 10512 alleles) and p.Ala283Val (11:2905337G>A; rs776541692; 1 in 1158 alleles) changes. Of note, these codons are two of the lesser-conserved amino acids within this "hot-spot" region ( Figure 2).

Expression of CDKN1C in human tissue
RNA-Seq analysis of CDKN1C in a panel of human tissues showed highest expression in the placenta (Figure 3; Human Protein Atlas data, https://www.proteinatlas.org/ ENSG00000129757-CDKN1C/tissue), with strong expression also in adipose tissue, ovary, adrenal, endometrium and kidney. Immunohistochemistry in the Human Protein Atlas repository shows strong staining in the nuclei of both decidual and trophoblastic cells (https://www.proteinatlas.org/ENSG00000129757-CDKN1C/ tissue/placenta).

Analysis of CDKN1C variants in growth restriction
Analysis of the PCNA-binding domain of CDKN1C by Sanger sequencing in a cohort of 58 children with isolated (nonfamilial) Silver-Russell syndrome did not reveal any pathogenic variants. Sequencing data for exon 2 is shown in Supplementary  Data 1 (Achermann, 2018).
A next-generation sequencing (NGS) approach of CDKN1C in children with SRS (n=8), IUGR/FGR (n=37), products of conception (n=22), and mothers with a history of recurrent miscarriage (n=78) also did not reveal pathogenic variants in this PCNA-binding domain region (Table 1).
Details of coverage for each sample/cohort and each nucleotide is shown in Supplementary Data 2-8 (Achermann, 2018).

Analysis of CDKN1C variants outside of the PCNA Binding Domain
Extended analysis of other regions of CDKN1C did not reveal any nonsense, frameshift and canonical splice site variants, which is not surprising as these variants would be loss of function and would be predicted to be associated with overgrowth.
We did identify two heterozygous, non-synonymous missense variants in the cohort of 78 women with recurrent miscarriages. One of these is a 9:g.2906703A>G variant (rs201715947) predicted to result in a p.Leu6Pro change that is present in gnomAD with allelic frequency of 0.0001102 (8/72604). The other variant is a 9:g.2906589C>T change predicted to result in p.Arg44His. This variant is not present in gnomAD but a p.Arg44Leu is present in one individual. We cannot definitively say whether these variants are clinically relevant and pathogenic, but feel it is unlikely.

Discussion
CDKN1C is now well-established as a key regulator of cell cycle and growth through G1 phase cell cycle arrest. Although loss-of-function of CDKN1C is known to cause macrosomia as part of Beckwith-Wiedemann Syndrome, it is only in the past eight years that "gain-of-function" variants in CDKN1C have been shown to cause growth restriction and IMAGe syndrome. These findings demonstrate clearly how opposite effects in protein function can have opposite phenotypes (Arboleda et al., 2012;Eggermann et al., 2014). Sometimes these features affect not just growth but also endocrine systems (e.g. adrenal tumours (BWS)/adrenal hypoplasia (IMAGe); congenital hyperinsulinism (BWS)/diabetes mellitus (one family with growth restriction)).
Review of published literature confirms that CDKN1C SNVs associated with IMAGe syndrome or growth restriction are all located within a "hot-spot" of the PCNA-binding domain of the protein. The exact function of this region is not clear, and the crystal structure of this region of CDKN1C has not yet been solved. Studies to date suggest pathogenic variants may increase protein stability or reduce the rate of degradation, thereby enhancing the negative effects of CDKN1C on cell cycle progression (Borges et al., 2015;Hamajima et al., 2013). The very localised nature of these variants clearly demonstrates a key role for this region in CDKN1C function. Further evidence for a potential role for CDKN1C in fetal growth restriction phenotypes has emerged with reports of CDKN1C variants in familial growth restriction and familial SRS (Brioude et al., 2013;Kerns et al., 2014;Sabir et al., 2019). Several features of IMAGe syndrome and SRS overlap, such as bi-frontal bossing and relative micrognathia. To date these children have not shown evidence of adrenal insufficiency. Whilst a lack of adrenal features could be due to underlying mosaicism or a somatic "rescue" event, as recently reported for variants in SAMD9 in the related condition MIRAGE syndrome (Buonocore et al., 2017;Narumi et al., 2016), the fact that several individuals in a family with the same CDKN1C SNV were affected provides strong evidence that the primary genomic event is influencing the phenotype rather than a rescue mechanism. Of note, review of the two SNVs associated with FGR and normal adrenal function reveals that they affect amino acids at the C-terminal region of the hotspot. In two families a charged arginine at codon 279 is replaced by a non-polar leucine, whereas a proline at this position is found in classic IMAGe syndrome. In another family, the arginine at position 281 is replaced by a non-polar isoleucine. In some situations, variants in amino-acids flanking critical motifs can be associated with milder phenotypes (Kyriakakis et al., 2017).

References
Despite these findings, we did not identify any CDKN1C variants in children with isolated (non-familial) SRS or IUGR/ FGR. These results are similar to data from Brioude et al. who only reported a CDKN1C variant in familial SRS and did not find variants in 68 children with an isolated condition (Brioude et al., 2013). Whilst this does not exclude CDKN1C as a potential cause of isolated or sporadic SRS/FGR, it does suggest that it is not a common cause.
Identification of such strong expression of CDKN1C in the placenta lead us to consider whether gain-of-function/increased stability of CDKN1C could be associated with loss of pregnancy or recurrent miscarriage. This hypothesis was supported further by recent studies linking placental genes to pregnancy loss (Perez-Garcia et al., 2018); a potential role for placental CDKN1C expression in fetal growth and regulation by oestrogen (Chen et al., 2018;Gou et al., 2017;López-Abad et al., 2016;Unek et al., 2014); and the fact that paternal imprinting of the gene means that a pool of deleterious variants could be present in the population and cause pregnancy loss (together with live births) in women who carry this variant and inherit it from their father.
Analysis was therefore undertaken in a cohort of mothers who had recurrent miscarriages (often together with live birth(s)) and also from products of conception. However, no variants were found in the PCNA binding domain of CDKN1C in these cohorts.
This work has several limitations. Although 203 total individuals were studied, each sub group is still relatively small and rare CDKN1C variants might be discovered if the sample size is increased. The causes of recurrent miscarriage and fetal growth restriction are clearly diverse and can be influenced by many factors, and regulatory regions or enhancers (Perez-Garcia et al., 2018;Rai & Regan, 2006;Stalman et al., 2018;Wakeling et al., 2017). Nevertheless, this study does highlight the role of the key region of CDKN1C in human fetal growth restriction phenotypes, and starts to address the potential role of single gene growth restriction phenotypes in more common obstetric and fetal conditions.

Data availability
Data has been uploaded to OSF Access for samples from the Baby Bio Bank is available and can be requested from the steering committee by emailing the Baby Bio Bank Manager (nita.solanky@ucl.ac.uk). More information about data access can be found on the website under the 'Protocol for the management of the Baby Bio Bank', section 15: https://www.ucl.ac.uk/child-health/research/genetics-and-genomicmedicine-programme/baby-biobank. An application form for the use of Baby Bio Bank resources must be completed prior to application (see appendix 2 in the protocol).
monozygotic twins with selective intrauterine growth restriction. Twin Res Hum 1.

Kahyee Hor
University/British Heart Foundation Centre for Cardiovascular Science, The Queen's Medical Research Institute, University of Edinburgh, Endinburgh, UK In this report Suntharalingham and colleagues investigate the extent to which variations in a known negative regulator of cell growth (CDKN1C) may be involved in isolated growth restriction or recurrent pregnancy loss.
They approach this in several ways: Through analysis of published literature Through analysis of CDKN1C expression in multiple tissues using available RNA-seq data Targeted sequencing in a number of patient populations. As a result of the literature search, they conclude that reported variants are located in the PCNA-binding domain of CDKN1C and therefore focus on this for their sequencing analysis. The analysis of available RNA-seq data suggests that the gene is highly expressed in placenta, suggesting it may have a role in pregnancy maintenance. Sequencing analysis revealed no pathogenic variants in the targeted region of CDKN1C suggesting this is an uncommon cause of fetal growth restriction or recurrent pregnancy loss.
The paper is well written and easy to read, with appropriate citation of the literature.

Minor comments:
PubMed search terms appear quite limited. Are the authors certain they have not missed any papers by not including terms such as IUGR, FGR, SGA?
Do the pregnancies/children in the FGR/IUGR group have any other complications?
The term 'usually a history of live births' in table 2 is rather vague. Were any of these live births also affected by FGR/IUGR etc. or by pregnancy complications commonly associated with poor fetal growth e.g. pre-eclampsia.
In table 2 -do the numbers in brackets refer to the number of families with that variant or the number of affected individuals within a family? Given the limited number of publications describing pathogenic mutations in the PCNA-binding domain of CDKN1C in IMAGe syndrome, familial SRS and in a single family with FGR and early adult onset diabetes, are the researchers planning to analyse other regions of the gene/its regulatory regions for other variations which may affect expression? Could they take this on to include analysis of placental gene expression/protein expression/function in assocision with analysis of sequence variation?

Are sufficient details of methods and analysis provided to allow replication by others? Yes
If applicable, is the statistical analysis and its interpretation appropriate?

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Response: Thank you for pointing this out. he number refers to the number of affected individuals within a family. We have added a footnote to clarify this.
Given the limited number of publications describing pathogenic mutations in the PCNA-binding domain of CDKN1C in IMAGe syndrome, familial SRS and in a single family with FGR and early adult onset diabetes, are the researchers planning to analyse other regions of the gene/its regulatory regions for other variations which may affect expression? Could they take this on to include analysis of placental gene expression/protein expression/function in association with analysis of sequence variation?
Response: We have analysed the rest of CDKN1C beyond the PCNA-binding domain wherever possible and did not find any variants that are clearly pathogenic (see response to reviewer 1). We did not build promoter or regulatory regions of into our capture design but it is an CDKN1C interesting point and maybe this can be addressed as more whole genome sequencing data become available. As pointed out, correlating genomic variations with placental gene and protein expression in the same individual would be a very powerful approach, if appropriate samples were available.
No competing interests were disclosed. In this paper Suntharalingham and coworkers investigate the involvement of CDKN1C, a negative regulator of cell proliferation and growth, in fetal growth restriction (FGR)/intrauterine growth restriction (IUGR) human conditions, pregnancy loss or recurrent miscarriage. While many loss-of-function variants have been reported associated with the overgrowth condition Beckwith-Wiedemann syndrome and affecting all the domains of the protein, only few variants predicted to be "gain of function", have been found associated with syndromic growth restriction conditions (IMAGe and SRS). The authors review the latter cases and find that all reported variants, associated with growth restriction, are localized in the same region inside the PCNA-binding domain, and affect evolutionary conserved amino acids.
By studying RNA-seq data available on the Human Protein Atlas website, the authors consider the high expression level of in placenta to hypothesize its involvement in pregnancy loss and CDKN1C miscarriage as a consequence of placenta restriction. Following a DNA targeted sequencing analysis restricted to the "hot spot" PCNA-binding domain in isolated cases of SRS (n=58), IUGR/FGR (n=26), loss of pregnancy (n=21) and women with recurrent miscarriage (n=71), they do not find any pathogenic variant, concluding that pathogenic variants of this analysed region are not a common cause of these pathological conditions. Response: In the revised submission, we have used the Agilent HaloPlex panel, as well as a Nonacus 3T cell target enrichment approach to add data from an additional 27 subjects (POC (n= 1); women with history of recurrent miscarriage (n=7); Silver Russell Syndrome (n=8) and fetal growth restriction (n=11) ). We have modified It would be interesting if the authors could report the analysis of the variants extended to the CDKN1C domains other than the region corresponding to the "hot spot" PCNA-binding domain (at least for cases analysed by NGS).
Response: Although our focus was on the PCNA-binding domain, we agree it is valid to look at regions outside of the hotspot and thank the reviewer (and reviewer 2) for the suggestion.
Analysis of CDKN1C using the gnomAD browser (accessed March 2020) shows an expected number of synonymous changes (o/e =1.06 (0.86-1.31)) but fewer missense (o/e= 0.51 (0.42-0.63)) and pLoF changes (o/e=0 (0-0.53)) than expected. ( ) https://gnomad.broadinstitute.org/gene/ENSG00000129757?dataset=gnomad_r2_1 Based on the fact that random variants within a gene are more likely to cause disruption rather than gain-of-function, the low number of observed missense changes is more likely to represent an intolerance to disruption and loss of function associated with overgrowth and Beckwith-Wiedemann Syndrome, rather than rare gain of function changes associated with FGR/IMAGe syndrome.
We have undertaken an extended analysis of data for the gene. We did not detect any category 1 variants (nonsense, frameshift, canonical splice site variants) at all, which is not surprising, as these would be loss of function and predicted to be associated with overgrowth.
We only identified two category 2 (non-synonymous missense) variants in the cohort, both in women who have had recurrent miscarriages. One of these is a 9:g.2906703A>G variant (rs201715947) predicted to result in a p.Leu6Pro change that is present in gnomAD with allelic frequency of 0.0001102 (8/72604). The other variant is a 9:g.2906589C>T predicted to result in a p. Arg44His. This variant is not present in gnomAD but a p.Arg44Leu is present in one male. Although the SIFT score for this variant suggests it is deleterious, we find that prediction models are not always reliable for gain-of-function effects. On balance, we do not feel that either of these variants are likely to be clinically pathogenic although we cannot say this definitively. We have commented on this in the revised results. More certain is our finding that there is not a large pool of additional CDKN1C variants in our cohort outside of the PCNA binding domain.

Minor comments:
• The authors should clarify why they have analyzed CDKN1C variants only in the isolated SRS cases, excluding the familial cases.
Response: Our cohort was only made up of sporadic Silver Russell Syndrome subjects; we had no familial cases. We have mentioned this in the methods. Table 2 are already present in Figure 1, therefore Table 2  Response: We agree there is some overlap, but we would prefer to keep both in the main body of text as Figure 1 shows the structure of CDKN1C and position of variants, and Table 2 shows nucleotide data, whether mineralocorticoid or glucocorticoid was given, and links to the original references.

• Most of information reported in
No competing interests were disclosed.

Competing Interests:
The benefits of publishing with F1000Research: Your article is published within days, with no editorial bias You can publish traditional articles, null/negative results, case reports, data notes and more The peer review process is transparent and collaborative Your article is indexed in PubMed after passing peer review Dedicated customer support at every stage For pre-submission enquiries, contact research@f1000.com