2 Genetics of preterm labour Nicolas M. Orsi* BSc, PhD Senior Research Fellow Nadia Gopichandran BSc, PhD Research Fellow Nigel A.B. Simpson MBChB, MRCOG Senior Lecturer & Honorary Consultant in Obstetrics & Gynaecology Perinatal Research Group, The YCR & Liz Dawn Pathology & Translational Sciences Centre Level 4, Leeds Institute of Molecular Medicine, Wellcome Trust Brenner Building, St James’s University Hospital, Beckett Street, Leeds LS9 7TF, UK The identification of women at risk of preterm labour remains an important challenge. While current prevention programmes rely on overt clinical and environmental parameters, the clustering of preterm labour within families and recurrence in susceptible women presents the case for a complex underlying genetic predisposition. Genetic polymorphisms are useful markers to identify high risk groups, although they provide little information either to their un- derlying functionality or the pathophysiological mechanisms involved; these must be validated through complementary analytical approaches. Data interpretation and inter-study comparisons must be made with caution, taking into account population size, study power, racial differences, inclusion/exclusion criteria and any underlying gene–environment and feto–maternal interac- tions. Large-scale, multicentre genetic studies coupled with high-throughput screening techniques are the most viable approaches to identify multilocus preterm labour susceptibility screening panels. Preventive strategies may then be applied to those women most likely to ben- efit from intervention. Key words: preterm labour; genetic polymorphisms; gene expression profiling; proteomics; cytokines; matrix metalloproteinases. * Corresponding author. Tel.: þ 44 (0)113 3438533; Fax: þ 44 (0)113 3438431. E-mail address : n.m.orsi@leeds.ac.uk (N.M. Orsi). 1521-6934/$ - see front matter ª 2007 Elsevier Ltd. All rights reserved. Best Practice & Research Clinical Obstetrics and Gynaecology Vol. 21, No. 5, pp. 757–772, 2007 doi:10.1016/j.bpobgyn.2007.03.020 available online at http://www.sciencedirect.com THE CHALLENGE OF IDENTIFYING HIGH RISK PATIENTS For the researchers of today and the clinicians of the future there are few challenges of greater importance than that of determining individual susceptibility to disease and response to therapy. This is illustrated most clearly in pregnancy, where the majority of women and babies have an uneventful and successful course, but a significant num- ber will encounter serious adverse outcomes such as pre-eclampsia, growth restric- tion and preterm birth. Preventive strategies for these conditions require accurate predictive methods to identify the women most likely to benefit from intervention and to enable the healthy majority to be left alone. This review presents the evidence for a genetic basis in preterm labour, the laboratory methods that are currently de- ployed in that search, and the potential application of such knowledge in the future. PRETERM LABOUR: A GENETIC PREDISPOSITION? While studies have increasingly focused on the underlying genetic susceptibility to pre- term labour (PTL), establishing the true genetic contribution to this disorder is prob- lematic as it is a complex condition involving non-Mendelian transmission, multiple genes, gene–gene and gene–environment interactions. 1,2 Furthermore, true genetic susceptibility effects can be masked by confounding environmental risk factors, such as socioeconomic status, cigarette smoking and infection. 3 Despite this, there is mount- ing evidence supporting a genetic contribution to the aetiology of PTL, with epidemi- ological observations indicating the clustering of PTL within families. 4 The most consistent risk factor identified across studies is a previous pregnancy re- sulting in PTL. A study by Hoffman and Bakketeig in a cohort of 27 677 Norwegian women reported that the incidence of preterm delivery (PTD) was 17.2% in women with a previous preterm birth; this figure increased to 28.4% in those with two previous PTDs. 5 These findings have been supported by other studies in cohorts of Danish 6 and Scottish 7 women where preterm birth risks of 15–18% and 32% were associated with a history of one and two previous PTDs, respectively. A similar observation was made in a mixed population of 178 896 women in Georgia, USA with first and second single- ton deliveries between 20–44 weeks’ gestation: a second PTD (20–36 weeks’ gestation) occurred in 28.3% of white women and 36.8% of black women. 8 It is less clear whether a woman’s own gestational age at birth increases her risk of a PTD. A multigenerational study in Utah, USA reported a positive correlation between these parameters, with women born under 30 weeks’ gestation having a 2.4-fold higher incidence of PTD. 9 However, other investigations reported no such correlation in a population of 11 092 primiparous Norwegian women 10 , suggesting that these differ- ences may in part be attributable to race. Nevertheless, these observations present a compelling case for a genetic contribution to PTL. GENETIC POLYMORPHISMS AND SUSCEPTIBILITY TO PRETERM LABOUR: THE RATIONALE The most thoroughly investigated genetic markers are point mutations in both coding and non-coding regions of DNA that affect a single nucleotide; the single nucleotide polymorphisms (SNPs). These single base substitutions in the DNA sequence have been the subject of numerous studies for susceptibility screening markers for an array of disorders 11–13 , although their precise biological effects often remain ill-defined. 758 N. M. Orsi et al The fact that some SNPs may occur in non-coding regions of the genome by no means indicates that these have no effect. Where they occur in promoter regions, they may affect the affinity of a particular DNA sequence for transcription factors, thereby influencing the transcriptional efficiency of that particular gene. 14 Other SNPs occur in an intron of the primary transcript which will ultimately be spliced out; while these will have minimal impact on that particular gene expression/protein profile, this does not preclude their potentially having some effect on a nearby gene. They may affect the gene’s transcription or have a more direct effect, given that certain gene sequences are known to overlap. 15 Other SNPs involve non-synonymous substi- tutions which, when occurring in an exon, alter the amino acid sequence of a given pro- tein and, potentially, its secondary/tertiary structure, function, receptor interaction and/or catalytic properties. Occasionally, such mutations can introduce a stop codon instead of coding for an amino acid: this can result in the production of mutant/inac- tive/truncated proteins or, in some cases, the absence of a secretory sequence which prevents the protein from being secreted outwith the cell. 16 All of these effects may result in an up- or down-regulation of a specific pathway which may influence host sus- ceptibility to an abnormal physiological response (e.g. an inflammatory dysfunction). There are, of course, other types of mutations beyond SNPs, involving deletions, insertions and repetitions of one or more nucleotides, which can also be used as screening markers in much the same way. However, where these are frameshift mu- tations occurring in coding regions (i.e. insertions/deletions of a number of nucleo- tides not divisible by three), disruptions in the genetic reading frame result in aberrant translation and the production of truncated, over long and/or dysfunctional amino acid sequence variants of the protein. Consequently, frameshift mutations generally result in severe disorders; where they affect the fetus, they usually result in pregnancy loss. 17 They are not therefore conventional markers of pregnancy complications. The premise underlying many screening systems is the merit of each SNP in iden- tifying a high risk population, either alone or cumulatively. In this respect, it may not matter whether the SNP has any functional effect; indeed, some of these mutations may be silent (i.e. code for the same amino acid in a protein sequence) or have little effect on transcriptional/translational efficiency. The observation that a SNP is associ- ated with an increased incidence of disease could simply be due to its acting as a marker of another co-inherited SNP (forming part of a so-called ‘haplotype’: a set of closely linked alleles inherited as a unit). However, as outlined, many SNPs are functional: the matrix metalloproteinase (MMP)-2 1306 (C/T) SNP T allele reduces promoter activity, while the tumour necrosis factor (TNF)- a 308 (G/A) SNP A allele is associated with higher protein levels. 14,18 This provides a stronger starting point to investigate SNP genetic–mechanistic rationales rather than basing SNP choice simply on the observation that specific polymorphisms have a known association with other disease states. Although some promoter polymorphisms have been linked with functional changes in transcriptional efficiency, it is a common observation that alterations in mRNA ex- pression are not necessarily reflected in protein levels and vice versa. This is often the case with cytokines, the much studied mediators of inflammation-induced preterm birth. 19 In this respect, non-coding regions of mRNA are increasingly understood to play a pivotal role in cell- and tissue-specific responses. More specifically, mRNA 5 0 and 3 0 untranslated regions (UTRs) can influence gene function through inherent mo- tifs which can regulate the nuclear export, cytoplasmic localisation, translational effi- ciency and stability of transcripts. While the synthesis of unstable/inefficiently Genetics of preterm labour 759 translated mRNAs may appear wasteful, this extensive post-transcriptional regulation of gene expression allows further subtle control of tissue-specific responses to various physiological stimuli. Perhaps most importantly, most of this regulation cannot be re- vealed by standard microarray technology 20 (see below). This degree of complexity can, of course, be further compounded by mutations in UTRs themselves 21,22 , which can presumably change base-pairing within their specific motifs. These observations highlight that establishing the true physiological impact of SNPs, even when functional, remains a complex issue. TRIALS AND TRIBULATIONS OF GENETIC STUDIES Most PTL prevention programmes have been based on overtly identifiable param- eters, such as socioeconomic status, maternal age, parity, race, previous personal/ family history, multiple pregnancy, uterine malformations and bacterial vaginosis. 9 However, tackling genetic studies presents a new set of complications, particularly in the case of disorders with a multifactorial aetiology such as PTL. The first con- sideration is the population that is being targeted, and whether the inclusion of multiple racial backgrounds is likely to introduce confounding variables. While some populations are understood to be relatively genetically isolated (e.g. Finnish, Icelandic, French Canadian), other more broadly distributed racial types have been associated with high carriages of specific allelic subtypes, as noted for Korean and African–American populations. 23,24 This will obviously affect the results of smaller studies, where control and study populations should be ethnically and geographi- cally matched as much as possible. At the other end of the spectrum, large, world- wide multicentre studies tackle this issue with the sheer scale of the numbers of patients recruited. In this way, genetic predispositions to disease which are applica- ble at a species level can be distinguished from those relating to specific populations. Study and control population sizes are a major consideration in genetic investiga- tions. While many published studies raise interesting points about genetic susceptibil- ity, not all have been suitably powered to consider the distribution of specific genotypes and the incidence of allelic variants within the population. 25 This can lead to potential false-positive/false-negative associations being made between genetic background and predisposition to disease. Other misinterpretations may stem from the selection criteria for the study population. For example, while PTL is defined as occurring before 37 completed weeks of pregnancy 9 , it may be more pertinent to pro- file the genetic/mechanistic processes governing PTL before 32 weeks’ gestation in- stead, given that this accounts for the majority of neonatal complications. 26,27 In this respect, the two populations may well be aetiologically very distinct. Given that the incidence of premature deliveries before 37 weeks and before 32 weeks is around 10 and 1–2% of all pregnancies, respectively, the number of patients recruited in indi- vidual studies can become a problem in the latter scenario and may impact on the breadth of clinical inclusion/exclusion criteria. In turn, these greatly influence study findings and the validity of inter-study comparisons. For example, not all studies to date account for potentially confounding variables such as maternal diabetes, blood pressure and fertility treatment. Many prematurity-related studies also compare poorly because of their focus on PTD as a whole, with little regard to the iatrogenic subset of deliveries contained within them. Indeed, the genetic and aetiological 760 N. M. Orsi et al background to iatrogenic preterm delivery, e.g. through pre-eclampsia or intrauterine fetal growth restriction, is likely to differ from that of PTL alone. GENETIC SCREENING: PROSPECTS, PROGRESS AND CHOICE OF SINGLE NUCLEOTIDE POLYMORPHISMS Clinical and research-based genetic screening is facilitated by the stability and analytical flexibility of DNA. Unlike plasma, serum, mucus or tissue samples, which may contain labile proteins or RNA requiring prompt purification, DNA-based work benefits from the greater stability of this nucleic acid. DNA suitable for genotyping can be extracted from an extensive array of samples ranging from buccal swabs to dried blood spots. In many cases, processing and preparation are minimal, which greatly facilitates sample collection and has contributed to the popularity of genetic studies. Traditionally, the emphasis in perinatal research has been on hypothesis-driven studies: the biological pathways implicated in labour have been dissected and the genes involved have been screened for potential functional polymorphisms. In many cases, investigators have also taken their cues from SNP panels used in other conditions, such as autoimmunity and cancer. To date, a single gene determinant for PTL inherited in a conventional Mendelian manner has not been identified. While the oft-cited ex- ception is fetal myotonic dystrophy, which is associated with an increased incidence of polyhydramnios and PTL 28 , this only accounts for a small minority of preterm births. Instead, it is generally accepted that susceptibility to PTL is governed by multigene effects, which makes the identification of genetic screening markers a formidable chal- lenge. The numerous genes implicated are thought to regulate inflammatory (e.g. cy- tokines, toll-like receptors), uteroplacental (e.g. coagulation factors, angiotensinogen), endocrine (e.g. steroid hormone receptors, dopamine), uterine contraction/remodel- ling (e.g. MMPs, oxytocin), oxidative stress (e.g. nitric oxide synthase, catalase) and metabolic (e.g. N -acetyltransferase, alcohol dehydrogenase) pathways 1 and, indeed, have been central to a number of studies (Table 1). Since many of these pathways ex- hibit a degree of functional redundancy, no single SNP is likely reliably to predict a pre- disposition to PTL; more accurate prediction will only follow from the cumulative/ interrelated activity of multiple SNPs. This observation highlights the multifactorial na- ture of the pathophysiology underlying PTL and, consequently, the broad array of genes (or multilocus susceptibility screening panels) involved. Given the multiple downstream effector pathways affected by inflammatory dys- functions allied to PTL, the obvious future hypothesis-driven candidate markers are MMPs. This family of proteolytic enzymes orchestrates extracellular matrix degra- dation allied to tissue remodelling and destruction. In PTL, MMP activation is a cyto- kine-mediated inflammatory process, resulting in rupture of the fetal membranes, and cervical dilation/ripening. Conveniently, a broad array of polymorphisms has been de- scribed in these proteases (MMP-1, MMP-2, MMP-3, MMP-7, MMP-8, MMP-9, MMP-12, MMP-13) and in their tissue inhibitors (TIMP-1, TIMP-2, TIMP-3) 29–39 , although these have only received limited attention in obstetrics thus far. 40 While this approach prom- ises to be comprehensive in investigating many of the SNPs which may regulate MMP expression and activity, it – like that of many studies to date – tends to ignore the location of these genes. MMP genes can be co-localised within discrete areas of the genome: the MMP-3 gene, for example, is located adjacent to the telomere side of the MMP-1 gene. 41 As a result, allied SNPs may be in linkage disequilibrium (i.e. the alleles at two distinctive loci occur in gametes more frequently than expected given Genetics of preterm labour 761 Table 1. Selected genetic polymorphisms and susceptibility to preterm birth (maternal genotypes unless otherwise specified). Polymorphism Variant allele Population Outcome Reference IL-1 b þ 3953 (C/T) IL1 b þ 3953 T increases protein production in vitro African e American/ Hispanic Fetal T allele carriage associated with spontaneous PTL Genc et al (2002) 66 Mixed American Maternal T allele not associated with spontaneous PTL and PTD Edwards et al (2006) 13 TNF- a 308 (G/A) A allele increases protein production African e American A allele carriage associated with preterm birth following PROM Roberts et al (1999) 76 TNF- a 863 (C/A) A allele alters protein expression: increased in vitro/decreased in vivo Mixed American A allele carriage not associated with preterm birth ( < 34 weeks) Amory et al (2004) 77 Mixed American AA genotype associated with earlier deliveries Amory et al (2004) 77 IL-6 174 (G/C) C allele decreases transcription/protein levels Mixed American CC genotype protective against spontaneous PTL Simhan et al (2003) 62 European (mainly German) GG genotype associated with PTD Hartel et al (2004) 78 Mixed American SNP not associated with PTL Jamie et al (2005) 79 Haplotypes Australian (European descent) Independently associated with preterm birth ( < 29 weeks) Annells et al (2004) 80 IL-10 1082 (G/A)/ 819 (T/C)/ 592 (C/A) e TNF- a þ 488 (G/A)/ 238 (G/A)/ 308 (G/A) IL-4 509 (C/T) Haplotypes White American GAG haplotype associated with decreased risk/A and C variants associated with increased risk of spontaneous PTD Engel et al (2005) 81 TNF- a 308 (G/A) e Lymphotoxin- a ISI 82 (G/C) Coagulation factor V (FV) þ 1691 (G/A) e Mexican Neither associated with PTD Valdez et al (2004) 63 Coagulation factor II (FII) þ 20210 (G/A) 762 N. M. Orsi et al Methyltetrahydrofolate reductase þ 677 (C/T) e Mexican T allele associated with PTD Valdez et al (2004) 63 b 2 -Adrenergic receptor ( b 2 AR) Arg16Gly Gly16 increases receptor desensitisation in vitro Hispanic Arg16 homozygous genotype associated with decreased spontaneous PTD Landau et al (2002) 82 Vascular endothelial growth factor (VEGF) 634 (G/C) C allele increases VEGF production Greek No association with PTL Papazoglou et al (2004) 83 VEGF þ 936 (C/T) T allele reduces protein production Greek T allele carriage increases susceptibly to spontaneous PTL Dihydrofolate reductase 19 bp deletion in intron 1 Deletion decreases transcription/folate availability African e American/ Hispanic Deletion allele associated with increased risk for PTD Johnson et al (2005) 84 PROM ¼ premature rupture of membranes; PTD ¼ preterm delivery; PTL ¼ preterm labour. Genetics of preterm labour 763 known allele frequencies/recombination fraction between the two loci): this suggests that the two are in close proximity. In this respect, examining specific haplotypes rather than single SNPs is an essential step in gaining a fuller understanding of their interrelations. In instances where specific alleles at two loci are always inherited together, this will also be economical on the analytical time involved. GENOTYPING SUCCESSES AND FAILURES: A CASE OF THE BAD WORKMAN BLAMING HIS TOOLS? Genetic susceptibility studies into PTL to date have focused on single or small panels of SNPs. Consequently, many investigations have relied on rather laborious and low-throughput genotyping approaches, such as polymerase chain reaction–restriction fragment length polymorphism (PCR–RFLP), amplification-refractory mutation system (ARMS)–PCR and, more unusually, heteroduplex PCR. Genotyping using PCR–RFLP is based on the polymorphic SNP base creating or destroying a restriction sequence. When the PCR-amplified products are digested with a suitable endonuclease, the frag- ment sizes generated reveal individual genotypes when resolved by agarose gel elec- trophoresis. ARMS–PCR is based on allele-specific paired PCR reactions performed using a generic reverse primer and two allele-specific forward primers which differ at their 3 0 terminus, thereby conferring the required specificity. A co-amplified pair of control primers is also included to confirm the suitability of cycling conditions. By contrast, the heteroduplex PCR approach bases its genotyping on the formation of heteroduplexes between amplified DNA and an existing stock of homozygote DNA for both alleles investigated. Although generally straightforward, these appro- aches remain relatively labour intensive and low throughput. More recently, the development of new PCR buffers has facilitated multiplex genotyping approaches; these can be based on ARMS–PCR and allow genotyping of several SNPs si- multaneously in paired reactions. This is made possible by designing primers which gener- ate allele-specific amplicons whose size differs for each individual SNP. 42 While this method significantly increases throughput, multiplex PCR still requires detailed optimisation and validation to account for the possible vagaries of allele-specific amplification (this is rou- tinely achieved by DNA sequencing). Although PCR-based approaches are generally robust once validated, genotype distribution provides an additional indication of protocol reliability. Genotypes are routinely checked for deviation from Hardy–Weinberg equilib- rium: this test is commonly used for quality control of large-scale genotyping and is one of the few ways to identify systematic genotyping errors in unrelated individuals. 43–45 Given the improved efficiency of such techniques, why has the quest for SNP markers of susceptibility to PTL not been resolved? The problem lies with the fact that even multiplex PCR genotyping based on random SNP markers alone fails to satisfy the need for high-throughput genetic screening. Indeed, the aim is not only to study large cohorts; the multifactorial nature of complex disorders like PTL also requires the profiling of large numbers of candidate SNPs given that the disorder is likely to com- prise different pathophysiological processes with different aetiologies and, conse- quently, a different underlying genetic predisposition. COMPLEMENTARY METHODS While genetic screens are potentially powerful tools for identifying high risk patient subgroups, they are largely uninformative about the molecular mechanisms underlying 764 N. M. Orsi et al both the aetiology and pathophysiology of PTL. Indeed, few SNP-based studies have fully validated the functional effects of the polymorphisms investigated, leaving the is- sue of whether they have any tangible effects on mRNA and protein levels open to question. However, these mechanistic answers can be sought by using a number of complementary analytical approaches. In this respect, microarray technology tackles the transcriptome (i.e. mRNA profile), allowing the simultaneous quantitative and qualitative genome-wide comparison of gene expression profiles in different cells, tissues and patient groups. As such, temporal and topographic changes in the patterns of gene expression, their co-regulation and in- teraction(s) can be simultaneously determined in large numbers of genes in a relatively short time. Furthermore, novel genes whose function was previously unknown in PTL may also be identified in this way. However, while microarrays are relatively flexible, their widespread implementation is limited by their relatively high cost, quality of avail- able RNA and the bioinformatic challenges allied to their complex data analysis. To a de- gree, some of these problems may be obviated by other, more focused cytometric multiplex mRNA quantification approaches. 46 In addition to the inherently high variabil- ity of microarray data, mRNA profiles may poorly correlate with the concentration/ac- tivity of their protein, as outlined earlier. Furthermore, many microarray-based studies rely on relatively small sample sizes and, while they provide insight into the progress of ongoing obstetric disorders and may highlight areas for therapeutic intervention, they provide limited information on the underlying individual responsiveness to labour-in- ducing triggers and predisposition to PTL. Despite this, microarrays have proved very useful in determining the ethnic and functional diversity in fetal membrane inflam- matory responses 47,48 , as well as alterations in myometrial gene expression profiles in normal, preterm and infection-induced labour. 49–52 Post-transcriptionally, the profiles of proteins and small peptides can also be deter- mined by a number of analytical approaches. However, the more ‘traditional’ methods of protein detection and analysis – liquid chromatography, spectroscopy, Western blots – have proven to be less useful in this particular remit due to issues of throughput, labour-intensiveness and multiplexing capacity. Mass spectrometry is a powerful, flexible technique which requires minimal sample volumes and addresses some of these issues. Thousands of proteins can potentially be identified according to their specific mass- to-charge ratios in ionised sample constituents. The identification of disease-specific biomarkers requires a comparison of PTL and control samples in order to generate so-called ‘molecular signatures’ of disease; the subsequent challenge is then to identify the discordant elements in both populations. Although such detailed analysis can remain laborious 53 , it has already been used successfully to identify panels of intra-amniotic in- flammatory markers. 54 More focused immunoassay-based approaches are available for instances where specific target proteins are to be profiled. In this respect, many stan- dard enzyme-linked immunosorbent assays (ELISAs – antibody-based protein detec- tion) are being superseded by their multiplexed, fluid-phased counterparts, which have the benefit of high throughput and low sample volume requirement. 55 However, not all ‘traditional’ complementary approaches are redundant. Immunohistochemistry, although poorly quantitative, provides unique tissue topographical evidence of protein production and localisation, although it has the disadvantage of potentially being inva- sive. Whatever the merit of the complementary approaches chosen, these techniques are essential in determining the true functional effects of many SNPs in order to attri- bute them a discrete functional role rather than purely considering them as markers. In addition, this sort of analysis could conceivably go full circle where broad-based mRNA and protein screens may also identify genes (and therefore candidate SNPs) of interest. Genetics of preterm labour 765 While genomic and proteomic approaches are therefore useful in profiling changes in gene expression and protein levels which may precede the manifestation of PTL by days, weeks or months 56 , SNP-related effects may also be gleaned from in vitro studies using patient tissue by eliciting specific responses to be studied outwith disease/pregnancy (e.g. lipopolysaccharide challenge to profile inflammatory responses of whole blood, gesta- tional membranes or specific cell subtypes). 57 Although the caveat to such studies is that responses in vitro may be prone to artefacts, these approaches provide more de- tailed insight into responses to disease triggers, onset of symptoms and therapy develop- ment. Furthermore, samples may be used to correlate genotype, mRNA expression and protein profile using the techniques outlined above. As such, this approach is undoubt- edly more useful in discerning the genetic control of PTL than animal models, where knock-out or over-expression models poorly mimic the action of many SNPs. GENE INTERACTIONS AND INTERPRETATION OF RESULTS The age-old debate about nature versus nurture is a significant confounding variable in the determination of genetic susceptibility to PTL. While many ‘traditional’ environmen- tal variables are known to operate independently to increase susceptibility to PTL, some form part of gene–environment interactions, where the result of their combined effects is much greater than the sum of their individual contributions. This has been elegantly highlighted by the interactive increase in predisposition to PTL induced by carriage of the variant A allele of the TNF- a –308 (G/A) SNP and bacterial vaginosis. 58,59 In this in- stance, carriage of the A allele and bacterial vaginosis each increase PTL risk by 1.8- and 1.6-fold, respectively. When present together, however, the risk escalates dramatically to 10.1-fold. 59 Such comprehensive, integrated studies are essential to future develop- ments in obstetric management and screening (both genetic and microbiological) inas- much as the risk of PTL can be substantially reduced by comparatively straightforward prophylactic treatment of – in this case – bacterial vaginosis. Other easily addressed en- vironmental factors, such as smoking 60 , may also demonstrate their interaction with var- ious PTL genes in the fullness of time, as they have in the atherosclerosis arena. 61 Although studies identifying SNP markers of susceptibility to PTL have provided convincing odds ratios, their impact has remained insufficient to warrant the imple- mentation of large-scale genetic screening of women who are pregnant/planning a fam- ily. Part of the problem lies with the inherently different distribution of allelic variants in different populations, which cannot warrant widespread deviations in clinical prac- tice based on a few markers alone. Furthermore, while specific SNP genotypes/allelic variants are considered to confer some protection against PTL (e.g. the interleukin [IL]-6 174 [G/C] SNP) 62 , others increase susceptibility to the disorder to a greater or lesser extent (e.g. the methyltetrahydrofolate reductase þ 677 [C/T] SNP) 63 : a bal- anced view should therefore consider all their relative contributions. However, while simple analyses such as cumulative variant allele carriage 64 have provided a more sen- sitive detection threshold than that of several genotypes considered singly, more com- plex bioinformatics approaches still need to be implemented. In this respect, some studies have started to address the issue of genetic susceptibility panels and multigene interactions. 48,65 Algorithms geared to determine host predisposition to PTL have, at the very least, to consider: (1) the identification of reliable susceptibility markers; (2) a measure of their relative contribution to overall risk; and (3) the potential additional clinical (e.g. history, age, microbiology), environmental (e.g. smoking, socioeconomic status) and genetic interactions involved. 766 N. M. Orsi et al This analytical complexity is compounded by the fact that predisposition to PTL is, like its term counterpart, the product of complex feto–maternal interactions. Although there are many examples to date of maternal genotype determining susceptibility to PTL, particularly with respect to an array of cytokine SNPs (see Table 1), other studies have indicated that fetal genotype (e.g. for the IL-1 b þ 3953 [C/T] SNP) is also associated with PTL. 66 This is further supported by other preliminary studies which have implicated both maternal and fetal IL-8 251 (T/A) SNP A allele carriage in increasing the risk of spontaneous PTL under 32 weeks’ gestation. 67 Nevertheless, determining fetal geno- type poses a pragmatic challenge. Its involvement in PTL can be readily inferred retro- spectively by genotyping DNA extracted from cord blood, but prospective, predictive analyses are not straightforward. On the one hand, while genotype could be verified through preimplantation genetic diagnosis of biopsied embryos from the minority of women undergoing fertility treatment, multiplex-based reactions may prove unreliable due to the potentially high rate of allele dropout. 68,69 While subsequent amniocentesis or chorionic villus sampling may be more accurate, these approaches present a minor risk of pregnancy loss 70 , which would have to be clearly offset by the putative merits of fetal screening. Given that maternal genotype is easily assessed, an obvious alternative strategy consists of evaluating possible fetal genotypes based on the father’s profile. In this respect, while a single study highlighted the potential association between paternal genotype and susceptibility to PTL 71 , most others have concluded that it has little, if any, role to play in this scenario. 6,72,73 Moreover, since paternal discrepancy can have a sur- prisingly high incidence (0.8–30% across studies) 74 , the diagnostic value of paternal ge- notype on that of the fetus should be viewed with diffidence. As such, maternal genotype remains the most realistic means of determining susceptibility to PTL. FUTURE DIRECTIONS IN GENETIC SCREENING FOR PRETERM LABOUR More advanced studies based on high-throughput genotyping and haplotype inheritance are becoming more popular in relation to determining the susceptibility to PTL. 75 Nonetheless, there is little doubt that chip arrays aimed at detecting multiple SNPs with improved resolution will underpin future studies into genome-wide screens; in- deed, a chip for screening 1 million SNPs is shortly due for release. This is pertinent because the human genome is thought to harbour at least 10 million SNPs. There are pragmatic approaches to tackling such colossal numbers: the international HapMap project aims to identify distinct sections of genome, each containing dozens of SNPs. Given that 65–85% of the human genome falls into so-called ‘haplotype blocks’ around 10 kb in length, investigators need only profile select marker (or so-called ‘tag’) SNPs to identify specific blocks and, consequently, each one’s own polymorphisms. As discussed earlier, SNP variants in close proximity within such blocks tend to be inherited together. However, despite their promise to generate a cornucopia of data very rapidly, SNP chip-based analyses remain expensive and currently offer restricted flexibility in cus- tomising small-scale candidate SNP panels. Furthermore, there are limits to the bioin- formatic methods for analysing data from such high-plex approaches. While these issues of cost and data management continue to pose a barrier to small-scale genetic research, it is unlikely to be long before this is finally overcome. We have seen that isolated allelic variants are unlikely to be sufficiently predictive of adverse outcome to merit serious consideration in any screening programme. How- ever, determination of an array of susceptibility SNPs (in combination with identifying Genetics of preterm labour 767 environmental triggers) may well form the basis of an effective targeted preterm pre- vention strategy in the near future. We already know that women who have had a mid-trimester delivery are at an increased risk of a subsequent preterm birth – current preventive strategies include screening and treating abnormal vaginal flora, checking for asymptomatic bacteriuria, improving placentation using heparin and aspi- rin, cervical cerclage, immunomodulation using progesterone or tocolytic treatment using indomethacin. A more detailed understanding of host genetic susceptibility would enable a more targeted approach in a subsequent pregnancy, requiring some or none of the above. However, the group who stand to gain most would be primigravid women, in whom there are currently few data to guide effective preventive strategies. CONCLUSION Large-scale, multicentre genetic studies are the most pragmatic way to approach the identification of multilocus PTL susceptibility screening panels. While genetic screening in perinatal research remains the focus of numerous ongoing studies, there is an undeni- able lag in analytical performance and output in the field compared to others such as autoimmunity, oncology and cardiovascular disorders. Some of these limitations are im- posed by the comparatively small size of our research field, which in turn impacts on available funding, expertise and analytical capacity. Nevertheless, given the substantial overlap between the mechanistic causes of obstetric complications and those of the aforementioned areas, it is only a matter of time before genetic screening fulfils its promise in identifying women at risk of PTL and related obstetric complications. While genetic analysis remains a powerful tool to this end, its full potential is best exploited in conjunction with other analytical approaches which relate the role of genetic polymor- phisms to clear physiological effects. Identifying individual predisposition to PTL only solves part of the problem: only a detailed understanding of the mechanisms involved at a systemic level can underpin the implementation of both existing and novel prophy- lactic therapeutic interventions (e.g. immunomodulators, progesterone analogues) over current remedial and palliative approaches (e.g. tocolytics, steroids). Practice points preventive strategies for preterm labour require accurate predictive methods to identify the women most likely to benefit from intervention a genetic predisposition to preterm labour has been suggested by familial clus- tering and recurrence genetic polymorphisms may have functional effects, such as affecting the tran- scription and translation of specific genes, or alter the protein itself comparisons between studies are problematical due to differences in study population size, underpowering of studies, racial differences and variations in inclusion/exclusion criteria genetic polymorphisms are useful markers but provide little mechanistic infor- mation alone; their effects need to be validated by using complementary ana- lytical approaches individual predisposition can be determined/increased by microbiological features, gene/polymorphism interactions and fetal contribution efficient genetic screening requires high throughput and multilocus profiling 768 N. M. Orsi et al ACKNOWLEDGEMENT The authors are indebted to Drs Richard Birnie and Thomas Hughes (Leeds Institute of Molecular Medicine) for constructive discussions on this topic. REFERENCES *1. Role of gene-environment interactions in preterm birth. In Behrman RE & Butler AS (eds.). Preterm birth: causes, consequences and prevention . Washington D. C.: The National Acaemies Press, 2007. pp. 207–228. *2. Dizon-Townson DS. Preterm labour and delivery: a genetic predisposition. Paediatr Perinat Epidemiol 2001; 2: 57–62. *3. Adams KM & Eschenbach DA. The genetic contribution towards preterm delivery. Semin Fetal Neonatal Med 2004; 9: 445–452. 4. Winkvist A, Mogren I & Hogberg U. Familial patterns in birth charact