Cord blood samples from 504 Polish preterm neonates including 106 extremely/early preterm (born at gestational age <33 weeks, according to the classification of The World Health Organisation, WHO) and 398 moderate/late preterm (gestational age 33-37 weeks) ( Table 1 ) were obtained from the Department of Newborns’ Infectious Diseases (University of Medical Sciences, Poznań, Poland), Department of Neonatology (Medical University of Gdańsk, Poland) and Department of Perinatology (Medical University of Łódź, Poland). 326 babies came from singleton pregnancies, 172 from 97 twin pregnancies (in 22, material from only one sibling was collected) and 6 from 2 triple pregnancies. The study was approved by the corresponding local ethics committees (Bioethics Committee of The Karol Marcinkowski Poznań University of Medical Sciences, Independent Bioethics Committee for Scientific Research at Medical University of Gdańsk, Bioethics Committee of The Medical University of Łódź). Written informed parental consent was obtained. This work conforms to the provisions of the Declaration of Helsinki. Isolated serum (blood taken into tubes with clot activator) was kept at -80°C. DNA (blood taken into tubes with sodium citrate) was isolated using GeneMATRIX Quick Blood Purifaction Kit (EURx Ltd. Gdańsk, Poland), according to the manufacturer’s protocol.

The FCN2 3’UTR region was generally amplified in three separate PCRs: reaction A (amplified region on chromosome 9: 134 887 339-134 888 460), reaction B (amplified region on chromosome 9: 134 888 251-134 890 157) and reaction C (amplified region on chromosome 9: 134 889 426-134 890 521). Alternatively, in the case of problems with sequencing of the product of reaction B, the additional reaction D (amplified region on chromosome 9: 134 887 339-134 890 506) was performed. PCR conditions, as well as sequences of primers and expected PCR products lengths are listed in Table 2 . The PCR mixtures (in a total volume of 20 µl) consisted of 50 ng of genomic DNA, 1× DreamTaq polymerase reaction buffer (includes 20 mM MgCl₂) (Life Technologies, Carlsbad, CA, USA), 1 U DreamTaq polymerase (Life Technologies, USA), 1 mM of each deoxynucleotide, and 0.5 µM of each primer. Reactions were performed using a Veriti™ 96-Well Thermal Cycler (Thermo Fisher Scientific, Waltham, MA, USA). PCR amplicons were analyzed using horizontal 2% agarose gel electrophoresis in a 1 × TAE buffer. The 100 bp Plus DNA size marker (Thermo Fisher Scientific) was used to normalize the size of each PCR product. The PCR products were purified using the EPPiC Fast mixture (A&A Biotechnology, Gdańsk, Poland), according to the manufacturer’s protocol. For sequencing of PCR products, the BrilliantDye™ Terminator (v.3.1) Cycle Sequencing kit (Life Technologies) was used, according to the manufacturer’s protocol. For both A and C reactions – primers A_ F, A_R, C_F and C_R, were used ( Table 2 ), respectively whereas for sequencing of the PCR products of B and D reactions, the forward primer GCCTGACCAGGCTTTTAGAG or reverse primer AGGTGCACACACACACACAC were employed. For the analysis of following SNP: rs4521835, rs73664188 and rs11103564 SNP, the reaction E (amplified region on chromosome 9: 134 887 339-134 888 155) was performed. The sequencing was successful in 97% at rs11103564 and 99% at rs4521835 and rs73664188. PCR products were purified with the use of BigDye XTerminator™ Purification Kit (Thermo Fisher Scientific), according to the manufacturer’s protocol and then sequenced using the 3500xl Genetic Analyzer (Applied Biosystems, San Diego, CA, USA). Sequencing results were analyzed using Chromas software version 2.4.1 (Technelysium, Brisbane, Australia).

Three SNPs were analyzed using TaqMan^® Assays purchased from Thermo Fisher Scientific: rs6537958 (Assay ID: C_29220561_10), rs6537959 (Assay ID: C_29220562_20) and rs11103565 (Assay ID: C_31552275_10). For all three SNPs, the PCR mixture consisted of 10 µl of 2 × TaqMan™ Genotyping Master Mix (Thermo Fisher Scientific), 1 µl of appropriate 20X TaqMan^® Assay, and 10 ng of genomic DNA. The PCR mixture was filled up with a distilled, DNase and RNase free water (Gibco, Waltham, MA, USA) to the final volume of 20 µl. The real-time PCR was performed using the 7900HT Fast Real-Time PCR System (Applied Biosystems) with 96-Well Block Module using standard mode thermal cycling setting under the following conditions: an initial denaturation step at 95°C for 10 min, followed by 40 cycles of denaturation (95°C, 15 s), single annealing and extension step (60°C for 1 minute). Fluorescence signal detection for both dyes was performed after each cycle. For each plate, a no template control (NTC) was included. Allelic discrimination analysis was performed using a SDS 2.3 software (Applied Biosystems).

Promoter polymorphisms at positions -986 (rs3124952, A>G) and -602 (rs3124953, G>A) were investigated according to the procedures described by Metzger et al. (32) while those at positions -64 (rs7865453, A>C) and -4 (rs17514136, A>G) as well as exon 8 SNPs (+6359, rs17549193, C>T; +6424, rs7851696, G>T) were determined as described by Szala et al. (33), with minor modifications.

Ficolin-2 in 395 cord blood serum samples was determined in TRIFMA as described by Świerzko et al. (34). Briefly, anti-ficolin-2 mAb (ABS 005-16, BioPorto Diagnostics, Hellerup, Denmark) were employed for 384 HB Optiplate coating, whereas biotinylated mAb (GN4, Hycult Biotech, Uden, The Netherlands) and Eu³⁺-labelled streptavidin (Perkin Elmer, Boston, MA, USA) were used for protein detection. The inter-assay and intra-assay coefficients of variation (%CV) were 18% and 13% for serum with ficolin-2 concentration of 3500 ng/ml, respectively whereas for serum with ficolin-2 concentration of 4900 ng/ml they equalled 10% and 7%, respectively.

Electrophoretic Mobility Shift Assay was performed using LightShift^™ Chemiluminescent EMSA Kit (Thermo Fisher Scientific), according to the manufacturer’s protocol. A 1325 bp DNA PCR product was obtained using F_F and F_R primers ( Table 2 ). The PCR mixtures (in a total volume of 50 µl) consisted of 20 ng of genomic DNA, 1× DreamTaq polymerase reaction buffer (includes 20 mM MgCl₂) (Life Technologies), 1 U DreamTaq polymerase (Life Technologies), 800 µM of each deoxynucleotide, and 0.4 µM of each primer. PCR product was labelled with biotin at 3’ end of the probe using the Pierce™ Biotin 3’ End DNA Labeling Kit (Thermo Fisher Scientific), according to the manufacturer’s protocol. Nuclear extract was prepared from human hepatocyte HepG2 cell line using a Nuclear Extract kit (Thermo Fisher Scientific). Four fmol of biotinylated PCR product was incubated (20 min, RT) with or without 3 µl of nuclear extract in the binding buffer supplemented with glycerol (2.5%), MgCl₂ (5 mM), Poly (dI·dC) 50 ng/µl, NP-40 (0.05%), 50 mM KCl and 10 mM EDTA. For inhibition experiments, four fmol of biotinylated PCR product was incubated (20 min, RT) with or without 3 µl of nuclear extract, as well as with or without inhibitor (0.4 pmol of unlabelled PCR product). After electrophoresis in native 6% polyacrylamide gel in TBE buffer, DNA was transferred to Amersham Hybond-N+ nylon membrane (Cytiva, Marlborough, MA, USA) and crosslinked at 120 mJ/cm² using a commercial UV-light crosslinking instrument (auto-crosslinking function). The biotin-labelled DNA was detected by chemiluminescence by placing a membrane in a film cassette and exposure membrane to X-ray for 1 minute.

The Hardy-Weinberg equilibrium (HWE) exact test was used for the analysis of the genotype distribution consistency for all polymorphisms. Linkage disequilibrium (LD) and haplotype block analysis were performed by Haploview 4.2 software (http://www.broad.mit.edu/mpg/haploview/). LD analysis was performed for each pair of polymorphisms using D’ and r², indicating the amount of LD between two genetic loci. A value of 1 suggests no random association of alleles at different loci. Haplotype block identification was performed based on Four Gamete Rule. The PHASE software (http://stephenslab.uchicago.edu/phase/download.html, version 2.1.1.) was used for diplotype reconstruction from genotype data. In silico analysis was performed with SNPinfo (FuncPred) (https://snpinfo.niehs.nih.gov) and RegulomeDB (https://www.regulomedb.org) databases. The statistical significance of the genetic impact on serum levels was tested with Mann-Whitney U test, Kruskal-Wallis ANOVA test and Spearman Rank correlation. The frequencies of genotypes were compared by two-sided Fischer’s exact test (or χ² when appropriate). The GraphPad Prism 6 software (https://www.graphpad.com) was used for those calculations. P values <0.05 were considered statistically significant. Odds ratio was calculated using online MedCalc software (https://www.medcalc.org).

To identify the polymorphic sites in FCN2 3’UTR region, in the preliminary experiment the 3183 bp fragment in 75 consecutive DNA samples was analyzed via Sanger sequencing. Initial screening allowed us to identify fifteen polymorphic sites with MAF>0.1 ( Table 3 ).

Next, the frequency of those genetic variants was determined in the remaining samples of preterm neonates ( Table 4 ). Rs4521835, rs73664188 and rs111103564 were investigated using Sanger sequencing, whereas rs11103565 was investigated by using the TaqMan Allelic Discrimination assay. The agreement between sequencing analysis and TaqMan probes for 89 samples was 91%. Disagreement was observed in three cases, whereas in five cases the SDS 2.3 software was not able to determine the genotype. Eleven polymorphic sites located between 134 888 378 and 134 889 620 bp analyzed in the 75 samples mentioned were shown to be in complete LD. Using the SNPinfo algorithm for the Caucasian population two tag SNPs: rs6537958 (correlated with rs6537957, rs6537960, rs7040372 and rs7847431) and rs6537959 (correlated with rs7046516, rs6537962a and rs7860507) were indicated. The frequencies of rs4521835, rs73664188, rs111103564 as well as rs6537958 and rs6537959 (representing eleven polymorphic sites) in preterm neonates are presented in Table 4 . They did not differ from those reported for the Central European (CEU) population. Except for rs4521835, all FCN2 polymorphisms tested adhered to the Hardy-Weinberg expectations (p>0.05) ( Table 4 ).

Based on four FCN2 promoter, two exon 8 and five 3’UTR polymorphisms, linkage disequilibrium analysis was performed in 504 samples ( Figure 1 ). According to the r² and D’ parameters, perfect linkage (r² = 1, D’=1) was detected between two 3’UTR SNPs (rs6537958 and rs6537959). As a result of sequencing analysis of 75 DNA samples, a perfect LD with them was also detected for other nine 3’UTR polymorphisms (rs7040372, rs7046516, rs7847422, rs7847431, rs6537957, rs6537960, rs6537962, rs114662298 and rs7860507) and confirmed in the ensembl database (https://www.ensembl.org/Homo_sapiens/Info/Index) for the CEU population. Very high r² and D’ values (0.669 and 0.98, respectively) were observed also between rs11103565 and rs6537958 as well as rs11103565 and rs6537959. Low r² but high D’ values (>0.8) were observed between 3’UTR rs73664188 and promoter rs3124953 (-602), rs7865453 (-64), rs17514136 (-4), exon 8 rs7851696 (+6424) as well as 3’UTR polymorphisms (rs11103564, rs11103565, rs6537958, rs6537959 and presumably nine others mentioned above).

Since rs4521835 did not comply with Hardy-Weinberg equilibrium, it was not included in the analysis. However, in the subgroup of early preterm neonates, no deviation from HWE was found for this SNP and Haploview analysis revealed a high LD with rs73664188, rs11103564, rs11103565, rs6537958 and rs6537959. The LD plots for separated extremely/early and moderate/preterm neonates are presented in Supplementary Materials ( Figure S1 ).

Using the Four Gamete Rule approach, four haplotype blocks were identified: block 1 and block 2 - each created by two promoter SNPs [rs3124952 (-986) and rs3124953 (-602) or rs7865453 (-64) and rs17514136 (-4), respectively], block 3 [involving three SNPs: exon 8 rs7851696 (+6424), rs73664188 and rs11103564, both 3’UTR] and block 4 [created by 3’UTR rs11103565, rs6537958, rs6537959 and presumably nine others variants] ( Figure 2 ). As mentioned above, rs4521835 was excluded from analysis. Four haplotypes corresponding to block 3 were identified. Among them GTT (the reference genotype) and GTC had frequency >1%. Within the block 4, three haplotypes were found, including GTA (the reference genotype) and ATA with frequency >1%. The D’ between 3 and 4 blocks was 0.85. The frequency of haploblocks identified in extremely/early and moderate/late preterm neonates is presented in Supplementary Materials ( Figure S2 ). In the case of neonates born before 33 week of gestation, data analysis gave five haploblocks, whereas for those having gestational age ≥33 weeks at birth, only three.

PHASE software allowed us to identify 49 diplotypes (D1-D49), corresponding to block 3 and block 4 genetic variants: exon 8 rs7851696 (+6424, G>T) and five 3’UTR SNPs: rs4521835 (T>G), rs73664188 (T>C), rs11103564 (T>C), rs11103565 (G>A) and rs6537959 (T>A). Twenty of them, with frequency ≥1% are listed in the Table 5 , whereas all identified diplotypes are presented in the Supplementary Materials ( Table S1 ). GTTTGT/GTTTGT (D15) represents the reference diplotype. Diplotypes differing only in SNPs with no greater influence on ficolin-2 level, were grouped into five subgroups (I-V). The sixth subgroup was associated with relatively low ficolin-2 concentration ( Table 6 and Figure 3 ). There were no significant differences in gestational age or birthweight between subgroups. Two diplotypes: D28, (GTTTGT/GGTTGA) and D35 (GGTCGT/GGTCGA) with frequency of 1% were not included in any subgroup.

The diplotype association with ficolin-2 concentration in cord serum was found to be markedly influenced by SNPs at rs4521835 and rs73664188 ( Figures 4A, B ). Genetic variant GC (present in D2, D5, D7, D12, D14, D17, D19, D21 and D32 diplotypes, included in subgroup VI; Table 5 ) was associated with low ficolin-2 serum concentration. The lowest median was associated with D12, where both haplotypes carry GC variants. Additionally, D12 represents the T allele at rs7851696 (+6424), also associated with low ficolin-2 level in cord serum (data not shown). The trend towards lower protein concentrations was found for the A variant of rs11103565 as well as TA variants at rs6537958 and rs6537959, being in perfect LD ( Figure 4F ). No association was observed for rs11103564 ( Figure 4D ). Interestingly, T/G heterozygosity at rs4521835 seems to be related to a higher protein concentration ( Figure 4A ).

Besides the relationship of 3’UTR diplotypes with ficolin-2 cord serum concentration described above, the association of genotype with selected complications like early prematurity, low birthweight, susceptibility to perinatal infection, pPROM and RDS was analyzed ( Table 6 ).

It revealed that the probability of carrying one of subgroup II diplotypes, [heterozygous at rs4521835: GTTTGT/GGTCGT (D13) and GTTTGT/GGTCGA (D10)] is significantly higher among neonates with early onset of infection or pneumonia (OR=2.69, p=0.008 and OR=2.81, p=0.026, respectively) than in the group of neonates without such complications ( Table 6 ). Diplotypes representing subgroup V (homozygous for C allele at rs11103564) were less frequent among babies born with birthweight ≤1500 g (OR=0.11 p=0.005) than among neonates born with a higher body mass. In contrast the over-representation of subgroup VI diplotypes with GC variants at rs4521835 and rs73664188 was found among newborns with the same birthweight range (OR=1.95, p=0.042). Moreover, that association seems to be independent of multiple pregnancies, since after excluding twins and triplets, the OR was even higher (2.98, p=0.003). No association of diplotypes with other prematurity-related complications was found.

SNPinfo (FuncPred) software was used for in silico determination of the interactions of genetic variants with microRNA ( Table 7 ). Only rs4521835 was predicted to be a target for microRNA (miR-150, miR-484, miR-618 miR-1226 and miR-1229). The regulomeDB database [designed to assess the probability of particular variant that affects binding of transcription factor (35)] helped to determine the regulatory function of SNPs investigated, with lower score indication of more evidence of functionality (36). Rs11103564 with score of 1f was predicted to be linked to expression of a gene target and to be an eQTL (expression quantitative trait locus) for OLFM1 (olfactomedin 1), whereas rs4521835, rs7847431 and rs11462298 with rank of 3a are less likely to be associated within the functional region. The remaining genetic variants with score 4-5 have minimal evidence of binding. Variants of rs4521835, rs11103565, rs7040372, rs7046516, rs7847431 and rs11462298 may alter significantly transcription factor (TF) binding motifs.

Sanger sequencing and LD analysis revealed a perfect linkage among eleven SNPs (rs7040372, rs7046516, rs7847422, rs7847431, rs6537957, rs6537958, rs6537959, rs6537960, rs6537962, rs11462298 and rs7860507) and a very high LD with rs11103565. In order to investigate the biological significance of non-random perfect linkage of twelve SNPs, their impact on nuclear protein binding was tested. Biotin-labelled 1325 bp PCR products of DNA samples from homozygotes for reference/variant alleles as well as from heterozygotes were incubated with nuclear extract. A significant shift in electrophoretic mobility was observed in all cases ( Figure 5A ). Moreover, the interaction between biotinylated wild-type DNA and nuclear extract proteins was inhibited not only by corresponding unlabelled PCR product but also by DNA of homozygote for variant alleles as well as by unrelated DNA PCR product ( Figure 5B ).