Cervical Cancer-Associated Promoter Polymorphism Affects Akna Expression Levels

ORIGINAL ARTICLE Cervical cancer-associated promoter polymorphism affects akna expression levels

GA Martínez-Nava1, K Torres-Poveda1, A Lagunas-Martínez1, M Bahena-Román1, MA Zurita-Díaz1, E Ortíz-Flores1, A García-Carrancá2, V Madrid-Marina1,3 and AI Burguete-García1,3

Cervical cancer (CC) is responsible for 4260 000 deaths worldwide each year. Efforts are being focused on identifying genetic susceptibility factors, especially in genes related to the immune response. Akna has been proposed to be one of them, but data regarding its functional role in the disease is scarce. Supporting the notion of akna as a CC susceptibility gene, we found two polymorphisms associated with squamous intraepithelial lesion (SIL) and CC; moreover, we identified an association between high akna expression levels and CC and SIL, but its direction differs in each disease stage. To show the potential existence of a cis-acting polymorphism, we assessed akna allelic expression imbalance for the alleles of the − 1372C4A polymorphism. We found that, regardless of the study group, the number of transcripts derived from the A allele was significantly higher than those from the C allele. Our results support the hypothesis that akna is a CC susceptibility genetic factor and suggest that akna transcriptional regulation has a role in the disease. We anticipate our study to be a starting point for in vitro evaluation of akna transcriptional regulation and for the identification of transcription factors and cis-elements regulating AKNA function that are involved in carcinogenesis.

Genes and Immunity (2015) 16, 43–53; doi:10.1038/gene.2014.60; published online 6 November 2014

INTRODUCTION and inflammation located in that particular region. Among them is Cervical cancer (CC) is the fourth major mortality rate cancer in akna, a gene that encodes a transcription factor present in the females worldwide, with an estimated 264 000 deaths in 2012 and germinal center of secondary lymphoid organs and immune 1 system cells, such as B and T cells, natural killer and dendritic accounting for 7.5% of all female cancer deaths. Around 11% of 8 these deaths occurred in Latin America and the Caribbean region, cells. making this disease a serious public health problem for all the The human akna gene is 61-kb long, contains 24 exons and countries of the area, including Mexico.1 encodes 9 different transcripts as the result of alternative It is well established that high-risk human papillomavirus (HPV) promoter usage, splicing and two polyadenylation sites; the F1 isoform was the first one to be described, and one of the few infection is a necessary factor for the development of CC. 8,9 Nevertheless, this factor alone is not sufficient as a significant functionally tested. It contains a N- and a C-terminus AT-hook domain that enables the protein to bind to AT-rich DNA regions percentage of infected women will succeed in HPV infection such as the High Mobility Group (HMG) proteins family, where this clearance, but a minor percentage (0.1–14.2%) with certain kind of domains was first described.10 In vitro experiments have environmental, lifestyle and genetic factors will not and by 2–4 shown the ability of the N-terminus AT-hook domain of AKNA F1 consequence will develop persistent infection and later on CC. isoform to bind to the AT-rich promoter regions in both CD40 and All such factors that predispose women to CC have not been fi CD40 ligand (CD40L), activating their expression. This, and the completely identi ed yet, and given the heritability of this disease notion that B lymphocytes within germinal centers are destined to (27%), which is higher than the one seen in other types of cancer die unless they are positively selected by antigens and signals (like colorectal and lung cancer), efforts are mostly focused in 5 initiated by co-stimulatory molecules interactions, such as identifying CC susceptibility genetic factors. The majority of the CD40–CD40L, suggests the important role of AKNA in achieving polymorphisms associated with CC, beside the known tumor- an efficient immune response.8 In addition, the generation of suppressor genes, oncogenic genes and those involved in cell an akna knockout animal model showed that, in its absence, cycle regulation, are located in genes related to the immune the expression of inflammatory cytokines (interferon-γ, and response.6 This is not surprising given the fact that the immune interleukin-1β), neutrophil-specific chemoattractants (NGP, CRAMP response triggered against the HPV infection by the host is and S100A9) and a neutrophil collagenase (matrix metallo- determinant in CC development. proteinase-9) is enhanced, which proves the implication of this It has been reported that the 9q32 region contains a gene in the negative regulation of inflammatory processes.11,12 susceptibility locus for CC.7 There are five candidate genes for Even though akna is located in a common fragile site linked to CC susceptibility related to antigen-dependent immune response loss-of-function mutations (FRA9E) associated with neoplastic

1Área de Infecciones Crónicas y Cáncer, Centro de Investigación sobre Enfermedades Infecciosas, Instituto Nacional de Salud Pública, Cuernavaca, Mexico and 2Unidad de Investigación Biomédica en Cáncer, Instituto de Investigaciones Biomédicas, Universidad Nacional Autónoma de México and Instituto Nacional de Cancerología, Secretaría de Salud, Distrito Federal, Mexico. Correspondence: Dr V Madrid-Marina or Dr AI Burguete-García, Area de Infecciones Crónicas y Cancer, Centro de Investigación sobre Enfermedades Infecciosas, Instituto Nacional de Salud Publica, Av. Universidad 655, Santa María Ahuacatitlán, Cuernavaca, Morelos 62508, Mexico. E-mail: [email protected] or [email protected] 3These authors contributed equally to this work. Received 27 June 2014; revised 2 September 2014; accepted 24 September 2014; published online 6 November 2014 Akna in cervical cancer GA Martínez-Nava et al 44 diseases,9,13–16 the only evidence of its direct association with CC Four groups of SNPs were automatically formed by LD, as SNPs is the one reported by our group in 2010.17 In that work, we found within each group presented LD only with other SNPs within that a significant, high magnitude association (odds ratio (OR) = 3.66, group. We selected one SNP of each LD group, following the 95% confidence interval (CI) = 1.35–9.94) of a single-nucleotide selection criteria mentioned in Materials and methods section, and polymorphism (SNP) within the N-terminus AT-hook motif the two SNPs that were not in LD with any other candidate SNP. (rs3748178, R1119Q) with an increased risk of CC in a Mexican Following this procedure, we ended up with six SNPs to evaluate population. That work links AKNA directly, as a transcription factor (rs115287453, rs73656049, rs2636898, rs10122672, rs10817594 of immune response co-stimulatory molecules, and as a negative and rs10817595). regulator of inflammation processes, with CC. Taking all this Three SNPs (rs115287453, rs73656049 and rs2636898) resulted together, it is possible to place akna as a potential CC to be non-polymorphic with a minor allele frequency o1%. We susceptibility genetic factor. However, functional data concerning could not carry out multinomial logistic regression analysis with AKNA and its complex transcriptional regulation is scarce, not to them, so we only show the OR with SIL and with CC for the mention the data about its role along CC natural history. To fully genotypes and alleles of the three remaining selected SNPs support the notion of akna as a CC susceptibility genetic factor, we (rs10122672 (−392C4T), rs10817594 (−1116C4T) and rs10817595 need to better understand its function and the role of its (−1372C4A); Table 3). We found a highly significant negative polymorphisms (coding and non-coding) in CC. association for CC with both − 392C4T and − 1372C4A minor Therefore, we aimed to assess the association of AKNA F1 allele homozygotes (OR = 0.31, 95% CI 0.12–0.80; and OR = 0.31, isoform promoter region SNPs with squamous intraepithelial 95% CI 0.12–0.79, respectively), with a significant negative trend lesion (SIL) and CC, as well as their effect over akna mRNA for both (P = 0.017 and P = 0.014, respectively). For − 1372C4A expression levels in peripheral blood mononuclear cells (PBMC) in minor allele homozygotes, we also found a significant negative both stages of the disease by measuring the number of transcripts association with SIL (OR = 0.37, 95% CI 0.16–0.86), along with a coming from each of both alleles simultaneously (feature known significant negative trend (P = 0.016), but we did not find a as allelic expression imbalance (AEI)). significant association for the − 392C4T minor allele homozygotes. These associations were maintained when the analysis by alleles of these two variants was performed. More importantly, this RESULTS analysis revealed a significant negative association with SIL and Known reproductive and sexual lifestyle risk factors for CC were − 392C4T minor allele (OR = 0.64, 95% CI 0.42–0.99). On the other confirmed to be statistically different between study groups: on hand, we did not find a significant association for either genotypes average, women with CC were older, had higher parity, had their or alleles of the − 1116C4T polymorphism with SIL or with CC first sexual intercourse at a younger age and reported a higher (Table 3). Upon doing the haplotype analysis, we observed that frequency of not using contraceptive methods than women with the negative associations seen in the genotype and allele analysis SIL and HPV-positive (HPV+) patients with non-cervical lesions were enhanced in the simultaneous presence of − 392C4T and (NCL) (Table 1). Most of the women had high-risk HPV infection, − 1372C4A minor alleles (TCA haplotype). However, in the had less than three lifetime sexual partners and reported a higher haplotype where the three minor alleles for the three SNPs of frequency of not having any cancer family history independently interest are present (TTA haplotype), this negative association did of the diagnosis (Table 1). The 79% of the women with SIL had a not prove to be significant (Table 3). We observed that the SNPs of low-grade dysplasia (cervical intraepithelial neoplasia 1), and most interest were in LD between − 1372C4T polymorphism with a D′ of the women with CC had stage II tumors. We found a significant value of 0.99 and 0.80 for − 392C4T and − 1116C4T, respectively. positive association with age at first sexual intercourse, age at first Akna mRNA expression level distribution normalized with childbirth, contraceptive method used and the number of HPRT-1 (hypoxanthine-guanine phosphoribosyltransferase-1) childbirths and both SIL and CC diagnosis (Table 2). History of mRNA expression levels (expressed by relative expression units previous sexually transmitted diseases (STDs) behaved as a risk (REU)) and stratified by the study group in PBMC and cervix are factor only for SIL, whereas that for CC behaved in the opposite shown in Figure 1. Akna expression levels in PBMC were higher in way, as well as cancer family history (Table 2). Given the fact that patients with SIL than in HPV+ NCL patients (P = 0.003); in contrast, these two covariates did not behave as expected, and that the in PBMC from CC patients, akna expression levels were lower best score in the goodness-of-fit tests performed for all the logistic (median of 1.04 REU) than in PBMC of HPV+ NCL patients models evaluated was achieved upon incorporating these two (Po0.001). Akna expression level tendency across the study variables, plus patient age and HPV type, further multinomial groups seen at the systemic level was maintained at cervix; logistic regression analyses were adjusted by these four however, the REU magnitude was significantly lower, and the covariates. difference between HPV+ NCL and SIL and between the former For the SNP selection, we first performed a bioinformatics and CC was not pronounced and was not significant. Nonetheless, analysis with the Neural Network Promoter Prediction (NNPP) the difference of akna expression levels between SIL (median of software to confirm the six putative transcription initiation sites 1.02 REU) and CC patients (median of 0.45 REU) was statistically (TIS) predicted by Sims-Mourtada et al.9 and found two additional significant (Po0.001) (Figure 1). TIS and their corresponding TATA boxes for the AKNA F1 isoform After performing the multinomial logistic regression analysis for in particular. To delimit the F1 isoform promoter region, we took akna REU tertiles at the systemic and cervix levels with SIL and CC 2 kb further upstream of the last TIS predicted, which left us with a diagnosis, we observed a significant high-magnitude positive 2.2-kb long sequence spanning from the 9:117124595 to the association for the higher PBMC akna expression tertile and SIL 9:117122379 base (this sequence was recently updated to (OR = 4.03, 95% CI 1.87–8.69) along with a significant positive 9:114362315-114360099). We then input this sequence in the trend across the tertiles (Po0.001). Contrarily, for CC there were remaining three transcription factor (TF) response element (RE) significant high-magnitude negative associations for median and predictor programs (SiteGA, TF Search and P-Match) and obtained high PBMC akna expression tertiles (OR = 0.21, 95% CI 0.09–0.49; a total of 1546 REs for 50 TFs. Detailed information about all the RE OR = 0.005, 95% CI 0.0005–0.04; respectively), also with a predicted that had a validated SNP in its sequence is shown in the significant negative trend (Po0.001). Even though there was no Supplementary Table S1. Of the SNPs present in the delimited significant difference between akna expression levels distribution promoter region, 16 were validated and fell in at least one of the in cervix across the diagnosis groups (Figure 1); we found a predicted RE. Two of these were not polymorphic, and 12 were in significant OR of 0.23 (95% CI 0.06–0.88) between the higher akna linkage disequilibrium (LD) with at least one other candidate SNP. expression tertile in cervix and CC (Table 4).

Table 1. Clinical and clinicopathological characteristics of the study population

Variable Total (n = 420) NCL (n = 109) SIL (n = 149) CC (n = 162) P-value

Age (years) Median (5–95%) 39 (22–66) 34 (22–66) 34 (21–66) 51 (32–76) o0.01a

Onset of menarche (years) Median (5–95%) 13 (11–15) 13 (11–15) 13 (10–15) 13 (11–16) 0.27a

Age at ﬁrst sexual intercourse (years) Median (5–95%) 18 (14–25) 20 (15–27) 18 (14–24) 17 (13–24) o0.01a

Age at ﬁrst childbirth (years) Nulliparous 12.75 52.73 7.25 3.75 o0.01b o19 39.38 14.55 44.2 43.75 19–21 23.51 7.27 27.54 25.62 ⩾ 22 24.36 25.45 21.01 26.88

Parity (%) Nulliparus 12.53 33.96 6.76 3.73 o0.01b ⩽ 3 47.47 55.66 52.7 37.27 43 40 10.38 40.54 59.01

Number of lifetime sexual partners (%) ⩽ 3 84.95 79.61 87.16 86.34 0.20b 4–9 12.86 18.45 9.46 12.42 ⩾ 10 2.18 1.94 3.38 1.24

Contraceptive method None 26.67 11.11 11.72 49.69 o0.01b Non-hormonal 38.52 51.52 35.86 32.92 Hormonal 34.81 37.37 52.41 17.39

History of previous STD None 55.94 53.33 24.49 88.16 o0.01b Other STD 34.9 35.24 60.54 9.87 HPV 9.16 11.43 14.97 1.97

Cancer family history (%) No 70.05 58.88 71.43 76.25 o0.01b Yes 29.95 41.12 28.57 23.75

HPV type (%) Low risk 8.49 14.58 15 0 o0.01b High risk 91.51 85.42 85 100

Dysplastic degree (only for the SIL group) CIN 1 ——79.19 —— CIN 2 ——10.07 — CIN 3 ——10.74 —

Cancer stage (only for the CC group) I ———12.5 — II ———56.25 III ———25 IV ———6.25 Abbreviations: CC, cervical cancer; CIN, cervical intraepithelial neoplasia; HPV, human papillomavirus; NCL, non-cervical lesions; SIL, squamous intraepithelial lesion; STD, sexually transmitted disease. Bold values denotes signiﬁcant P-values (Po0.05). aKruskal–Wallis test P-value. bΧ2 test P-value.

In order to explore whether the SNPs of interest had a allele homozygotes. Consistent with the analysis done by relationship with the different akna expression levels seen in CC genotypes, we observed a mean estimated difference of almost patients, we stratified them by polymorphism genotypes one REU for the minor allele of the three polymorphisms of (Figure 2). The difference in akna REU level distribution between interest in CC patients. This mean estimated difference was of minor allele homozygous and ancestral allele homozygous CC more than one REU (β = 1.14; 95% CI 0.30–1.99) when the three patients was only statistically significant for − 392C4T and minor alleles of the SNPs of interest were simultaneously present − 1372C4A polymorphisms; as for the − 1116C4T polymorphism, and exhibited a significant positive trend as the number of minor the carriers of one minor allele had higher levels than the carriers alleles present increased (P = 0.008) (Table 5). of two copies of the ancestral allele (Figure 2). The mean On the other hand, we did not find a statistically significant estimated difference of akna REU in CC patients was 1.97 association between the presence of the minor allele of the three (0.54–3.41) REU for both − 392C4T and − 1372C4A minor allele SNPs of interest and akna expression levels in SIL patients; homozygotes and 2.12 (0.45–3.78) REU for − 1116C4T minor however, for the − 1116C4T polymorphism we found that the

Table 2. Association analysis of conventional reproductive and sexual lifestyle risk factors with squamous intraepithelial lesions and cervical cancer in the study population

Risk factor NCL/SIL/CC ORa (95% CI)

(n = 109/149/163)b Squamous intraepithelial lesion (SIL) Cervical cancer (CC)

Onset of menarche (years) o12 12/30/2029 1 1 ⩾ 12 95/118/131 0.50 (0.24–1.02) 0.47 (0.22–1.01)

Age at ﬁrst sexual intercourse (years) ⩾ 18 77/75/78 1 1 o18 27/73/83 2.77 (1.61–4.79)c 3.22 (1.77–5.84)c

Age at ﬁrst childbirth (years) ⩾ 22 14/29/43 1 1 Nulliparous 29/10/6 0.12 (0.04–0.34) 0.17 (0.05–0.57) 19–21 4/38/41 3.92 (1.15–13.38) 4.71 (1.36–16.26) o19 8/61/70 3.29 (1.23–8.81)c 3.05 (1.13–8.23)c

Parity Nulliparous 36/10/6 1 1 ⩽ 3 59/78/60 7.10 (3.13–16.11) 3.95 (1.48–10.54) 43 11/60/95 49.01 (16.57–144.96)c 20.87 (6.63–65.68)c

Number of lifetime sexual partners ⩽ 3 82/129/139 1 1 43 21/19/22 0.57 (0.28–1.13) 0.91(0.43–1.91)

Contraceptive method Non-hormonal 51/52/53 1 1 Hormonal 37/76/28 2.00 (1.14–3.46) 0.80 (0.41–1.55) None 11/17/1980 1.58 (0.67–3.71) 4.38 (2.00–9.58)c

History of previous STD None 56/36/134 1 1 Other STDs 37/89/15 3.75 (2.11–6.65) 0.22 (0.11–0.47) HPV 12/22/2003 2.85 (1.26–6.48)c 0.14 (0.04–0.56)c

Cancer family history No 63/ 105/122 1 1 Yes 44/42/38 0.57 (0.34–0.97) 0.44 (0.24–0.79) Abbreviations: CI, confidence interval; HPV, human papillomavirus; NCL, non-cervical lesions; OR, odds ratio; STD, sexually transmitted disease. Bold values denotes significant P-values (Po0.05). aAdjusted by age. bSome numbers do not equal the total sample size because of missing values. cStatistically significant P-values for trend (Po0.05).

mean estimated difference of akna REU in NCL HPV+ patients significant association of akna expression levels at the cervix and was of 24.02 (2.88–45.16) for the minor allele and of 43.39 in PBMC with CC diagnosis, and (iii) the fact that its expression (11.67–75.11) for this allele homozygotes. For the other two differs across SNP genotypes in CC patients. polymorphisms of interest, there was no significant association Even though the precise role of AKNA in CC has not been with akna expression levels in NCL HPV+ patients (Table 5). reported, the evidence showing the linkage of its locus with this To further explore the putative − 1372C4A polymorphism disease and the previous report of a coding SNP in its amino effect over akna transcriptional regulation, we carried out an AEI terminal AT-hook domain that increases the risk for CC place akna analysis in the heterozygous patients. We could only perform the as a potential genetic susceptibility factor.7,17 Furthermore, the fi AEI analysis in this SNP, because we did not nd suitable capacity to regulate the expression of immune response co- − 4 − 4 synonymous SNPs for 392C T and 1116C T polymorphisms. stimulatory molecules and its role in the negative regulation of We found a clear difference in the number of mRNA coming from inflammation processes entail AKNA loss of function with known the two alleles of interest (Log2 (A allele/C allele) = 1.58; 95% CI emerging and enabling hallmarks of cancer (specifically, avoiding 1.48–1.69). On average, the higher AEI was seen in SIL patients immune destruction and tumor promoting—inflammation followed by HPV+ NCL and CC patients; however, there was no 18 fi hallmarks, respectively). In agreement with this notion, signi cant statistical difference between the diagnosis groups − 4 (P = 0.44) (Figure 3). we have found two non-coding akna polymorphisms, 392C T (rs10122672) and − 1372C4A (rs10817595), which are negatively associated with SIL and CC. In addition, these associations are not DISCUSSION masked by the previous reported associated SNP (rs3748178), as The main findings of this study were (i) the identification of a these two polymorphisms are not in LD with it, as the 1000 significant negative association between SIL and CC with akna genomes project reports.19 This finding supports the idea of akna − 392C4T, and − 1372C4A polymorphisms, (ii) together with a being a CC susceptibility gene.

Table 3. Association analysis of polymorphisms of interest with squamous intraepithelial cervical lesion and cervical cáncer

Polymorphism Frequency (%) ORa (95% CI)

NCL (n = 109) SIL (n = 149) CC (n = 162) SIL CC

− 392C4T (rs10122672) C/C 26.6 36.51 28.22 1 1 C/T 45.74 44.44 53.37 0.62 (0.30–1.31) 0.71 (0.31–1.60) T/T 27.66 19.05 18.4 0.43 (0.18–1.01) 0.31 (0.12–0.80)b C (0.55) 49.47 58.73 54.91 1 1 T (0.45) 50.53 41.27 45.09 0.64 (0.42–0.99) 0.56 (0.35–0.90) P-valuec 0.7

− 1116C4T (rs10817594) C/C 42.31 46.67 34.57 1 1 C/T 45.19 38.52 54.32 0.74 (0.39–1.41) 1.37 (0.68–2.75) T/T 12.5 14.81 11.11 0.88 (0.36–2.12) 0.59 (0.21–1.64) C (0.64) 64.9 65.93 61.73 1 1 T (0.36) 35.1 34.07 38.27 0.88 (0.58–1.35) 0.91 (0.57–1.44) P-valuec 0.82

− 1372C4A (rs10817595) C/C 25 37.96 27.61 1 1 C/A 46.15 43.8 53.99 0.54 (0.26–1.11) 0.69 (0.31–1.55) A/A 28.85 18.25 18.4 0.37 (0.16–0.86)b 0.31 (0.12–0.79)b C (0.55) 48.08 59.85 54.91 1 1 A (0.45) 51.92 40.15 45.09 0.60 (0.39–0.90) 0.57 (0.36–0.90) P-valuec 0.67

Haplotypes CCC 48.48 61.09 55.47 1 1 TCA 17.58 6.79 6.2 0.33 (0.15–0.71) 0.18 (0.08–0.41) TTA 33.94 32.13 38.32 0.74 (0.45–1.23) 0.63 (0.36–1.10) Abbreviations: CC, cervical cancer; CI, confidence interval; NCL, non-cervical lesions; OR, odds ratio; SIL, squamous intraepithelial lesion. Bold values denotes significant P-values (Po0.05). aAdjusted by age, human papillomavirus type, history of previous sexually transmitted disease and cancer family history. bStatistically significant P-value for trend (Po0.02). cHardy–Weinberg equilibrium test in the study population.

Regarding the akna expression levels, the only previous report and increased risk of high-grade HPV lesions.23–25 This is of their indirect association in vivo with a human disease is with an consistent with the phenotypes observed in the akna knockout autoimmune disorder, Vogt–Koyanagi–Harada syndrome. In a case and constitutes another potential mechanism that explains the control study, Mao et al.20 found significant lower levels of AKNA low expression levels seen in CC patients in comparison to HPV+ protein in CD4+ T cells obtained from patients than in the cells NCL patients. Having lower levels of akna expression could from healthy controls. This syndrome is characterized by a chronic enhance the expression of proinflammatory cytokines and inflammation of highly pigmented tissues, and the decrease in neutrophil-specific chemokines promoting a chronic inflammation AKNA protein concentration seen in these patients is consistent that favors HPV persistence, resulting in a higher risk of CC. The with the lack of inflammatory response intensity regulation increase in infiltrated inflammatory cells concentration (that is, observed in the akna knockout mice. Nonetheless, the molecules neutrophils) could result in a greater release of reactive oxygen and mechanisms involved in akna expression regulation under species, well-known mutagenic chemicals, promoting the transi- normal and pathological conditions have not been identified to tion of nearby cancer cells to a heightened malignant state.18 date. This study presents the first evidence of akna expression On the other hand, we found high akna expression levels in behavior in CC patients at a systemic level as well as in tumor PBMC of SIL patients compared with HPV+ NCL patients. This can biopsies. be explained by the fact that immune response in patients with We found lower akna expression levels in PBMC as well as in SIL is still fully active trying to achieve clearance of the HPV lesion; tumor biopsies of CC patients than in PBMC and in cervix after all, it is well established that the majority of HPV lesions scrapings of HPV+ NCL patients. It has been reported that eventually regress and, even though all the immune response interaction of CD4+ T-cell CD40L with CD40 of dendritic cells is mechanisms that have a role in this phenomenon are not needed for CD4+ T cells to help generate an efficient CD8+ T cells’ completely understood, there is evidence of key elements of the response in vivo and consequently is crucial for the development cellular immune response (that is, high CD4:CD8 ratio, a more of antitumor immunity.21 Furthermore, in the particular case of CC, prominent T helper type 1 response, T-cell response against early it has been shown that despite the presence of infiltrating CD8+ viral antigens and so on) that favors the cytological remission of T cells, they have poor activity.22 Low levels of akna expression HPV lesions in patients.22,23,26–30 In contrast, in CC this response could compromise the amount of CD40 and CD40L (or other co- has been counteracted by the action of tumor cells. Previous stimulatory molecules) present on the surface of dendritic cells reports have shown that the tumor cells are capable of secreting and CD4+ T cells, implicating a deficient antitumor immunity in CC anti-inflammatory cytokines (interleukin-10, transforming growth patients. factor-β and interleukin-4) suppressing the cellular immune In a somewhat opposite way, there is evidence suggesting that response.22,31 chronic inflammation, with sustained inflammatory infiltrates of Moreover, it is highly likely that a great percentage of women macrophages and neutrophils, is associated with HPV persistence with SIL still have an active HPV infection, meaning that the viral

© 2015 Macmillan Publishers Limited Genes and Immunity (2015) 43 – 53 Akna in cervical cancer GA Martínez-Nava et al 48 genome has not integrated yet into the genome of host cells, during viral infections. Nonetheless, it is not completely clear unlike in CC, where the integration of the viral genome is whether the increase of akna expression seen is a general presumed. The high expression of akna in PBMC of SIL and HPV+ mechanism activated by virus infection.32 NCL patients is probably counteracting not only the lesion Even though we did not aim to identify the cells expressing produced by HPV but also the active viral infection. It has been akna in cervical scrapings and tumor biopsies, we hypothesize shown that in Huh 7.5 cells infected with hepatitis C virus akna that, as akna REU levels were considerably lower in the cervix than expression is substantially upregulated 72 h. after infection; this in PBMC, the number of cells expressing akna present in the fi supports the idea that akna expression is differentially regulated scrapings and biopsies was signi cantly small. We based this reasoning on the fact that expression of this gene has been reported only in lymphoid secondary organs and in immune system cells. We speculate that infiltrating lymphocytes, dendritic * cells or macrophages present in the cervix and the tumor 400 microenvironment, and not in the epithelial cells or tumor cells, are the ones responsible for akna expression levels seen in situ. * Efforts should be addressed to elucidate which cells are 300 expressing akna in the cervix in order to clarify the potential role of this gene in the pathology of the disease, as the functions attributed to akna are highly likely cell type and environment 200 dependent. Interestingly, carriers of the two minor alleles for each of the three SNPs of interest had higher akna expression levels than 100 AKNA REU in PBMC carriers of the respective two ancestral alleles, suggesting that the presence of these polymorphisms enhance akna expression. The biological impact of having one more akna REU given the 0 haplotype TTA is not easily inferred in CC; that is because of the NCL SIL CC intuitively opposite functions reported for AKNA, the complex role that CD40 has along cervical carcinogenesis33–35 and because * of the controversial role of inflammation in HPV persistence 23–25 40 and CC. AEI assessment allows us to know if there is a difference in the number of transcripts coming from each allele of a certain 30 polymorphism. It is expected that in a heterozygote both alleles (DNA strands) are transcribed at the same rate, unless the presence of a certain allele favors its transcription over the other 20 one. In this last case, the SNP responsible has a functional role on the nearby gene expression regulation, meaning that it is a cis- expression quantitative trait locus (eQTL). There are plenty of SNPs in several promoter gene regions associated with CC; however, 10 AKNA REU in cervix few are assessed to be a functional cis-acting polymorphism. Recently, Chen and Gyllensten36 evaluated a cis-eQTL of HLA-DRB1 gene in CC patients, and their findings provide a possible 0 explanation for the variable patterns of associations with HLA- NCL SIL CC B*07 and DRB1*1501-DQB1*0602 and the consistent pattern Figure 1. Akna REU level distribution over the diagnosis group in of association with DRB1*1301-DQA1*0103-DQB1*0603 in CC. PBMC (a) and in cervix-scraped cells and tumor-biopsied cells (b). Nonetheless, there is no other study in CC patients that, by Asterisk represents a statistically significant P-value for the assessing AEI, points out the existence of a potential cis-acting Kruskal–Wallis test adjusted by multiple comparisons (Po0.01). polymorphism. It is of high relevance to identify those noncoding

Table 4. Odds ratios for akna REU and squamous intraepithelial lesion and cervical cáncer

akna REU NCL/SIL/CC ORa (95% CI)

(n = 109/149/163)b SIL CC

PBMC (n = 412) o2.35 35/27/132 1 1 2.35–20.16 35/37/26 1.99 (0.87–4.56) 0.21 (0.09–0.49) 420.16 34/85/1 4.03 (1.87–8.69)c 0.005 (0.0005–0.04)c

Cervix (n = 269) o0.39 11/27/1949 1 1 0.39–1.61 11/57/36 1.54 (0.47–5.09) 0.33 (0.09–1.18) 41.61 11/47/20 1.36 (0.41–4.46) 0.23 (0.06–0.88) Abbreviations: CC, cervical cancer; CI, confidence interval; NCL, non-cervical lesions; OR, odds ratio; PBMC, peripheral blood mononuclear cells; REU, relative expression units; SIL, squamous intraepithelial lesion. Bold values denotes significant P-values (Po0.05). aAdjusted by age, human papillomavirus type, history of previous sexually transmitted disease and cancer family history. bSome numbers do not equal the total sample size because of poor quality of the sample. cStatistically significant P-value for trend (Po0.02).

* * 30 30

20 20

10 10 AKNA REU in PBMC AKNA REU in PBMC

0 0 C/C C/T T/T C/C C/T T/T

* 30

10 AKNA REU in PBMC

0 C/C A/C A/A Figure 2. PBMC akna REU level distribution stratiﬁed by genotypes of the SNPs of interest in CC patients. Minor allele homozygotes of − 392C4T(a) and − 1372C4A(c) polymorphisms exhibited higher levels of akna REU. For the − 1116C4T polymorphism genotypes (b), the distribution of akna REU was statistically different only between the heterozygotes and homozygotes of the ancestral allele. Asterisk represents a statistically signiﬁcant P-value for the Kruskal–Wallis test adjusted by multiple comparisons (Po0.01).

Table 5. Estimated mean difference of PBMC akna REU between polymorphism genotypes and alleles stratiﬁed by diagnosis

Polymorphisms βa coefﬁcient (95% CI)

Total NCL SIL CC

− 392C4T (rs10122672) C/C —— —— C/T 6.64 (−3.39–16.68) 30.77 (−10.47–72.02) − 0.08 (−15.66–15.49) 0.96 (−0.14–2.06) T/T 1.81 (−10.43–14.06) 9.85 (−35.06–54.76) − 8.12 (−28.19–11.96) 1.97 (0.54–3.41)b C —— —— T 1.41 (−4.67–7.49) 2.41 (−19.20–24.03) − 3.53 (−13.17–6.11) 0.93 (0.25–1.62)b

− 1116C4T (rs10817594) C/C —— —— C/T 5.40 (−4.06–14.86) 43.39 (11.67–75.11) − 9.06 (−23.88–5.76) 1.00 (−0.05–2.04) T/T 6.91 (−6.99–20.81) 38.20 (−7.08–83.47)b − 10.75 (−30.80–9.30) 2.12 (0.45–3.78)b C —— —— T 3.93 (−2.45–10.32) 24.02 (2.88–45.16) − 6.83 (−16.56–2.90) 0.92 (0.21–1.63)

− 1372C4A (rs10817595) C/C —— —— C/A 6.79 (−3.02–16.60) 32.16 (−7.10–71.41) 2.19 (−12.66–17.05) 0.96 (−0.14–2.06) A/A 3.07 (−8.88–15.02) 15.22 (–27.03–57.48) − 6.60 (−25.53–12.35) 1.97 (0.54–3.41)b C —— —— A 2.03 (−3.93–7.99) 5.52 (−15.13–26.17) − 2.62 (−11.85–6.62) 0.93 (0.25–1.62)b

Haplotypes CCC —— —— TAC − 3.02 (−15.17–9.14) − 21.57 (53.25–10.11) 2.01 (−19.30–23.32) 0.78 (−0.92–2.48) TAT 3.08 (−4.40–10.56) 13.96 (−12.20–40.11) − 6.38 (−17.89–5.13) 1.14 (0.30–1.99)c Abbreviations: CC, cervical cancer; CI, confidence interval; NCL, non-cervical lesions; PBMC, peripheral blood mononuclear cells; REU, relative expression units; SIL, squamous intraepithelial lesion. Bold values denotes significant P-values (Po0.05). aAdjusted by age, human papillomavirus type, history of previous sexually transmitted disease and cancer family history. bStatistically significant P-value for trend (Po0.02), assessed by genotypes. cStatistically significant P-value for trend (Po0.02), assessed by haplotypes.

© 2015 Macmillan Publishers Limited Genes and Immunity (2015) 43 – 53 Akna in cervical cancer GA Martínez-Nava et al 50 indeed with CC risk and not with HPV infection risk; this is due 6 to the incorporation of only HVP+ patients to the study sample. This selection allowed us to determine the role of akna polymorphisms and its different expression profiles over what distinguishes patients with an active HPV infection and those that fi 4 developed SIL or CC. Furthermore, con rmation of the presence of known reproductive risk factors for CC (age at first sexual intercourse, age at first childbirth, contraceptive method and number of childbirth) supports the external validity of our results. The fact that history of previous STD and cancer family history 2 variables in our population behaved in an opposite way as have been reported could be a sign of a potential measurement bias. As Log2(A allele/ C allele) we used a previously built database, we could not control the data collection processes; therefore, we could not prevent such bias by 0 study design. However, all multinomial logistic regression analyses NCL SIL CC performed were adjusted by these two variables (HPV type and patient age) in order to control the potential confounding effect that history of previous STD and cancer family history covariates 1.8 could be providing, so it seems unlikely that the results obtained are being affected by this fact. Moreover, the SNPs of interest were all in Hardy–Weinberg equilibrium, meaning that the allelic 1.7 frequencies seen in our study population are equal to the ones

1.61 expected given no selection pressure or genetic drift. Additionally, 1.6 1.58 the minor allele frequency observed for each SNP of interest were 1.56 consistent with the frequency reported for Mexican ancestry 19 1.5 population in the 1000 genomes project, so we can assure there was no bias that favored the presence of the polymorphism in the study sample. 1.4 Regarding limitations, given the epidemiological design of this

Mean Log2(A allele/ C allele) study we cannot assure the akna expression level differences seen 1.3 here are a cause or an effect of the diagnosis status. However, the NCL SIL CC possibility that the akna expression levels are partially due to the presence of the polymorphism indirectly gives temporality to Figure 3. Akna AEI in − 1372C4A polymorphism heterozygotes. a the association observed between akna expression levels and CC. ( ) Distribution of the number of transcripts ratio coming from each The last thing to consider is that given the close proximity of the − 1372C4A polymorphism alleles, stratified by diagnosis status. (b) On average, the number of transcripts derived from the A allele three SNPs of interest and the LD between them seen in the study log ðÞAallele : population, the synonymous SNP selected for the AEI assays was almost three ðAallele ¼ 2 2 Callele ¼ 21 58 ¼ 2:99Þ times larger than Callele was also in LD with − 1116C4T (rs10817594) and − 392C4T the number of transcripts derived from C allele. Error bars represent (rs10122672) SNPs; so the real functional cis-eQTL could be any of the s.d. in each group. the three SNPs of interest. However, we consider that, given the consistent results obtained for the − 1372C4A (rs10817595) variants that have a function over the abnormal spatiotemporal polymorphism throughout the entire study, and the fact that expression patterns seen in CC patients. This will give insights of running the bioinformatics analysis with the same software used molecules and genes not yet associated with the disease, and a for the SNP selection with the respective minor allele in this better understanding of the role of known susceptibility genetic particular location eliminates the predicted MZF1 motif, it is highly factors in the neoplastic process. likely that this polymorphism is the responsible for the AEI In the present study, we have shown that the number of mRNA observed. copies from the − 1372 A allele is nearly three times larger than In conclusion, the present study suggests that transcriptional the mRNA copies from the C allele. This finding shows that akna regulation of akna has an important role in the natural history of expression is enhanced by the presence of the − 1372C4A minor CC disease; however, its effect is likely to be spatiotemporal allele, meaning that there is an AEI, which is a key signature of cis- dependent, and further studies are needed to comprehend acting polymorphism.37 The motifs predicted in the bioinformatics AKNA function in normal conditions and cancer. Supporting the analysis in this particular location correspond to myeloid zinc notion of akna being a genetic susceptibility factor for CC, we finger 1 (MZF1), CCAAT/enhancer binding protein beta (C/EBPB) have found an allelic imbalance in akna expression. We suggest and caudal type homeobox 1 (Cdx1) TFs (Supplementary Table that the significant negative association observed for the S1); the latter is an intestine-specific TF, so it is highly unlikely to − 1372C4A (rs10817595) variant and the high levels of akna be responsible for akna transcriptional regulation in PBMC. MZF1 expression with CC are explained by the effect of its minor and C/EBPB are TFs known to be related to immune system cells. allele over akna transcriptional regulation. We believe this work Both TFs, MZF1 and C/EBPB, have been associated with different will set the basis for in vitro evaluation of akna transcriptional – types of cancer38 41 but only MZF1 is known to participate in the regulation in carcinogenesis and other immune response-related invasiveness of human CC cells.42 However, with the evidence diseases. generated in this study we cannot conclude which TF is the one regulating akna expression. Further in vitro analyses are required to corroborate the presence of this cis-regulating element, to MATERIALS AND METHODS know which TF may be regulating akna transcription and in which Study population way it is doing so (as a repressor or an activator). This study employed samples from a previously built biological sample Regarding methodological strengths and limitations, one of the bank described elsewhere.43 Briefly, all women who assisted to the Centro study strengths lies in the fact that the associations reported are de Atención para la Salud de la Mujer de los Servicios de Salud del Estado

Genes and Immunity (2015) 43 – 53 © 2015 Macmillan Publishers Limited Akna in cervical cancer GA Martínez-Nava et al 51 de Morelos, México (CAPASAM) between June 2008 and December 2011, Biosystems 79000 Real-Time PCR System instrument. If the triplicate and to the gynecology service of the Instituto Nacional de Cancerología de showed a VC42%, the experiment was repeated. With the equation 2 − Δct, México (INCan) between September 2010 and December 2011, and met we calculated the REU of akna mRNA at the cervix and systemic level. the inclusion criteria (recent cytological, colposcopic and histopathological diagnosis; period of residence in Mexico 45 years; aged ⩾ 18 years; and had not initiated treatment) were invited to participate. After the women AEI assessment signed the informed consent and answered a lifestyle, socio-demographic For each associated SNP, we selected a synonymous SNP in LD (D′ ⩾ 0.99) and hormonal factors’ questionnaire, peripheral blood was collected to with each one of them to evaluate akna AEI. We found a suitable obtain genomic DNA (gDNA) and total RNA. Also, cervical scrapings and synonymous SNP (rs3748177) only for the − 1372C4A (rs10817595), tumor biopsies were obtained from women diagnosed with NCL or SIL and therefore, we could only determine the allelic imbalance for one of the CC, respectively, in order to further extract the total RNA. Complementary three associated SNPs. We designed a TaqMan probe for the selected DNA (cDNA) synthesis was carried out from total RNA extracted from synonymous SNP, taking the F1 isoform transcript sequence as template, PBMC, cervical scrapings and tumor biopsies. Women who suffered from where the reverse primer fell on the 9 and 10 exon junction, assuring non- autoimmune or chronic inflammation diseases or had a STD active gDNA amplification. infection (besides HPV) during the sampling were excluded from the PBMC cDNAs from the heterozygous women for the − 1372C4A original study (0.54%). polymorphism were used to perform, by duplicate, qPCR assays with Because we aimed to identify association with higher CC risk, and not custom-designed TaqMan probes for each one of the SNP alleles in an with HPV infection, the present study was carried out only in the HPV+ Applied Biosystems StepOnePlus Real-Time PCR System instrument, women from the described biological bank. Of these 420 HPV+ women, following the manufacturer’s instructions. The results were analyzed with 109 had NCL, 149 had SIL and 162 had CC. the Applied Biosystems Step One Software v2.2.2 (Life Technologies). The protocol of this study was approved by the National Institute of According to the method proposed by Chen et al.50 to calculate the ratio Public Health Institute of Mexico Bioethics and Research committees. of mRNA expression between two alleles, we used the following equation: alleleY log - ¼ a ´ ðCt - Ct Þþðb - b Þ SNP selection 2 alleleX 2 1 2 1 Due to the absence of previous association reports of regulatory SNPs or eQTL in the akna gene region, we decided to perform a bioinformatics In order to do so, we generated a standard curve with 8:1, 4:1, 2:1, 1:1, 1:2, analysis to help us select the SNPs that most likely affect akna F1 isoform 1:4 and 1:8 ratio dilutions of cDNAs from two samples with homozygous transcription rate. We used four distinct free online programs, each one genotypes (−1372 C/C, and − 1372 A/A) that had similar (±0.5) HPRT1 Ct with a different approach to predict both TATA boxes and TIS (NNPP44)or values in the previous qPCR assay. The linear equation of the standard 2 TF RE (SiteGA,45 TF Search46 and P-Match47). curve generated was f(x) = 1.436x+0.060 and yielded a R of 0.97. Because we wanted to validate the previously TIS predicted for the AKNA F1 isoforms, we used the same bioinformatics tool (NNPP) and the Statistical analysis whole genomic sequence of akna as Sims-Mourtada et al.9 did. This way Known reproductive and sexual lifestyle factors for CC risk (that is, age at we delimited the sequence used for the three remaining programs; we fi fi took 2 kb upstream from the last TIS predicted by the NNPP program, rst sexual intercourse, age at rst childbirth, parity, STD history, fi fi which left us with a 2.2-kb promoter sequence to analyze further. contraceptive method and HPV type) were rst con rmed in the study We used only the matrices for vertebrate organisms, when the program population by bivariate multinomial logistic regression models adjusted by – had such an option, and a minimal score of 0.8 as a cutoff point for each age. Hardy Weinberg equilibrium was corroborated using the allelic bioinformatic tool employed. Also, specifically for the SiteGA tool we opted frequencies of the whole study population. LD between the selected SNPs was calculated according to the standardized values proposed by to use both strands (forward and reverse) and to include all the TF 51 available for analysis. For P-Match, we selected the minimizing false- Lewontin; likewise the ORs for the genotypes, the alleles and haplotypes negative option to establish the cutoff point and the option to use only the of interest with SIL and CC were calculated using the multinomial logistic high-quality matrices. As we could use specific profile matrices for distinct regression models adjusted by age, HPV type, history of previous STD and cell types with P-Match, we performed two separate analyses: one without family history of cancer. a specific profile and one using the predefined immune system cell profile. PBMC and cervix akna REU were compared across the groups of study – We then crossed the data obtained from the bioinformatics analysis with using the Kruskal Wallis test adjusted by multiple comparisons. Tertiles for the information of all SNPs reported in the dbSNP48 and Ensembl49 both akna REU were created according to the respective observed databases on the delimited promoter region (data accessed on November distribution in the HPV+ NCL group to evaluate their association with SIL 2012). We selected the SNPs with the higher RE prediction score of each and CC by multinomial logistic regression models. Distribution of akna REU group of SNPs in LD that were previously validated by the HapMap project in PBMC was compared by genotypes and alleles of interest to determine or the 1000 genomes project and that had a minor allele frequency ⩾ 1% whether those variants modified the corresponding akna REU levels in a Mexican ancestry population. observed in the CC group with the Kruskal–Wallis test and the Wilcoxon– Mann–Whitney test in the case of the alleles. The mean estimated difference of PBMC akna REU between alleles of interest and haplotypes, Genotyping stratified by study group, was assessed by linear regression models The allelic discrimination assays for the six SNPs selected were carried out adjusted by age, HPV type, history of previous STD and cancer family from PBMC gDNA by traditional PCR using TaqMan probes in an Applied history. Biosystems 79000 Real-Time PCR System instrument, according to the The potential AEI was assessed by calculating the mean value of the manufacturer’s instructions. All assays were done in duplicate with a base two logarithms of the allele2/allele1 ratio and their respective 95% CI concordance of 100% and were analyzed and determined with the across the groups of study. We also compared their distribution across the ‘Sequencing Detection System’ software (SDS 2.3, Applied Biosystems, Life groups with the Kruskal–Wallis test. Technologies, Carlsbad, CA, USA) using a quality call rate of no o0.99. In all the regression models performed, only the observations with complete information of the variables of interest were included. All the Akna expression level assessment statistical analyses were performed using the STATA program, version 12.1 (StataCorp, College Station, TX, USA). Quantitative PCR (qPCR) was performed from PBMC, cervical epithelial- scraped cells and tumor-biopsied cDNA using TaqMan probes for gene expression assays (Applied Biosystems) to measure akna mRNA (probe ID: CONFLICT OF INTEREST Hs00363936_m1) in both levels (systemic and in cervix). We decided to use fl the HPRT-1 gene (probe ID: Hs01003267_m1) as a endogenous control, for The authors declare no con ict of interest. it showed no variation among the three study groups. Both Taqman probes used were carefully selected to ensure that one of their respective primers fall on an exon junction to prevent gDNA amplification. ACKNOWLEDGEMENTS The qPCR assays for both genes (target and endogenous) were carried We thank all the patients for their participation in the study and MD Guillermina out in triplicate with a variation coefficient (VC) of o2% in an Applied López-Estrada, Karina Delgado-Romero and David Cantú for the gynecological

© 2015 Macmillan Publishers Limited Genes and Immunity (2015) 43 – 53 Akna in cervical cancer GA Martínez-Nava et al 52 sampling and patient care that constitute the biological sample bank used. This work by TGF-beta1 and IL-10 in cervical cancer patients. J Clin Immunol 2009; 29: was supported by the Instituto Nacional de Salud Pública de México (INSP) and 532–544. grants from CONACYT--FONSEC SSA/IMSS/ISSSTE-2011-01-161710, Mexico. GAM-N 23 Hibma MH. The immune response to papillomavirus during infection persistence was recipient of a PhD fellowship from CONACYT. This study was also supported by and regression. Open Virol J 2012; 6: 241–248. the Consejo Nacional de Ciencia y Tecnología (CONACyT) and the Fondo Sectorial de 24 Castle PE, Hillier SL, Rabe LK, Hildesheim A, Herrero R, Bratti MC et al. Investigación en Salud y Seguridad Social, México. The supporting agency had no An association of cervical inflammation with high-grade cervical neoplasia in role in the study design; in the collection, analysis and interpretation of the data; in women infected with oncogenic human papillomavirus (HPV). Cancer Epidemiol – the preparation of the report; or in the decision to submit the article for publication. Biomarkers Prev 2001; 10: 1021 1027. This work was submitted in partial fulfillment of the requirements for the PhD degree 25 Hammes LS, Tekmal RR, Naud P, Edelweiss MI, Kirma N, Valente PT et al. fl of GAM-N from the Doctoral Program in Health Sciences of the School of Public Macrophages, in ammation and risk of cervical intraepithelial neoplasia (CIN) progression-clinicopathological correlation. Gynecol Oncol 2007; 105: 157–165. Health of Mexico. 26 Nobbenhuis MA, Helmerhorst TJ, van den Brule AJ, Rozendaal L, Voorhorst FJ, Bezemer PD et al. Cytological regression and clearance of high-risk human REFERENCES papillomavirus in women with an abnormal cervical smear. Lancet 2001; 358: 1782–1783. 1 Globocan 2012 (IARC) Section of Cancer Information. Available at: http://globo 27 Monnier-Benoit S, Mauny F, Riethmuller D, Guerrini JS, Căpîlna M, Félix S et al. can.iarc.fr/Pages/fact_sheets_cancer.aspx. Immunohistochemical analysis of CD4+ and CD8+ T-cell subsets in high risk 2 Schlecht NF, Kulaga S, Robitaille J, Ferreira S, Santos M, Miyamura RA et al. human papillomavirus-associated pre-malignant and malignant lesions of the Persistent human papillomavirus infection as a predictor of cervical intraepithelial uterine cervix. Gynecol Oncol 2006; 102:22–31. – neoplasia. JAMA 2001; 286: 3106 3114. 28 de Jong A, van Poelgeest MI, van der Hulst JM, Drijfhout JW, Fleuren GJ, Melief CJ 3 Dalstein V, Riethmuller D, Pretet JL, Le Bail Carval K, Sautiere JL, Carbillet JP et al. et al. Human papillomavirus type 16-positive cervical cancer is associated with Persistence and load of high-risk HPV are predictors for development of high- impaired CD4+ T-cell immunity against early antigens E2 and E6. Cancer Res 2004; grade cervical lesions: a longitudinal French cohort study. Int J Cancer 2003; 106: 64: 5449–5455. 396–403. 29 Peng S, Trimble C, Wu L, Pardoll D, Roden R, Hung CF et al. HLA-DQB1*02- 4 Insinga RP, Perez G, Wheeler CM, Koutsky LA, Garland SM, Leodolter S et al. restricted HPV-16 E7 peptide-specific CD4+ T-cell immune responses correlate Incident cervical HPV infections in young women: transition probabilities for CIN with regression of HPV-16-associated high-grade squamous intraepithelial lesions. and infection clearance. Cancer Epidemiol Biomarkers Prev 2011; 20:287–296. Clin Cancer Res 2007; 13: 2479–2487. 5 Magnusson PK, Lichtenstein P, Gyllensten UB. Heritability of cervical tumours. Int J 30 Woo YL, van den Hende M, Sterling JC, Coleman N, Crawford A, Kwappenberg KM Cancer 2000; 88: 698–701. et al. A prospective study on the natural course of low-grade squamous intrae- 6 Zhang X, Zhang L, Tian C, Yang L, Wang Z. Genetic variants and risk of cervical pithelial lesions and the presence of HPV16 E2-, E6- and E7-specific T-cell cancer: epidemiological evidence, meta-analysis and research review. BJOG 2014; responses. Int J Cancer 2010; 126: 133–141. 121:664–673. 31 Alcocer-Gonzalez JM, Berumen J, Tamez-Guerra R, Bermudez-Morales VH, 7 Engelmark MT, Ivansson EL, Magnusson JJ, Gustavsson IM, Beskow AH, Peralta-Zaragoza O, Hernandez-Pando R et al. In vivo expression of immunosup- Magnusson PK et al. Identification of susceptibility loci for cervical carcinoma by pressive cytokines in human papillomavirus-transformed cancer cells. Viral genome scan of affected sib-pairs. Hum Mol Genet 2006; 15: 3351–3360. Immunol 2006; 19:481–491. 8 Siddiqa A, Sims-Mourtada JC, Guzman-Rojas L, Rangel R, Guret C, Madrid-Marina V 32 Papic N, Maxwell CI, Delker DA, Liu S, Heale BS, Hagedorn CH. RNA-sequencing et al. Regulation of CD40 and CD40 ligand by the AT-hook transcription analysis of 5' capped RNAs identifies many new differentially expressed genes in factor AKNA. Nature 2001; 410:383–387. acute hepatitis C virus infection. Viruses 2012; 4: 581–612. 9 Sims-Mourtada JC, Bruce S, McKeller MR, Rangel R, Guzman-Rojas L, Cain K et al. 33 Altenburg A, Baldus SE, Smola H, Pfister H, Hess S. CD40 ligand-CD40 interaction The human AKNA gene expresses multiple transcripts and protein isoforms as a induces chemokines in cervical carcinoma cells in synergism with IFN-gamma. result of alternative promoter usage, splicing, and polyadenylation. DNA Cell Biol J Immunol 1999; 162: 4140–4147. 2005; 24: 325–338. 34 Hill SC, Youde SJ, Man S, Teale GR, Baxendale AJ, Hislop A et al. Activation of CD40 10 Aravid L, Landsman D. AT-Hook motifs identified in a wide variety of DNA-binding in cervical carcinoma cells facilitates CTL responses and augments chemotherapy- proteins. Nucleic Acid Res 1998; 19: 4413–4421. induced apoptosis. J Immunol 2005; 174:41–50. 11 Ma W, Ortiz-Quintero B, Rangel R, McKeller MR, Herrera-Rodriguez S, Castillo EF 35 Huang Q, Qu QX, Xie F, Zhang T, Hu JM, Chen YG et al. CD40 is overexpressed by et al. Coordinate activation of inflammatory gene networks, alveolar destruction HPV16/18-E6 positive cervical carcinoma and correlated with clinical parameters and neonatal death in AKNA deficient mice. Cell Res 2011; 21: 1564–1577. and vascular density. Cancer Epidemiol 2011; 35: 388–392. 12 Moliterno AR, Resar LM. AKNA: another AT-hook transcription factor ‘hooking-up’ 36 Chen D, Gyllensten U. A cis-eQTL of HLA-DRB1 and a frameshift mutation of MICA with inflammation. Cell Res 2011; 21: 1528–1530. contribute to the pattern of association of HLA alleles with cervical cancer. Cancer 13 Helmrich A, Stout-Weider K, Hermann K, Schrock E, Heiden T. Common fragile Med 2014; 3: 445–452. sites are conserved features of human and mouse chromosomes and relate to 37 Jones BL, Swallow DM. The impact of cis-acting polymorphisms on the human large active genes. Genome Res 2006; 16: 1222–1230. phenotype. Hugo J 2011; 5:13–23. 14 Savas S, Liu G. Genetic variations as cancer prognostic markers: review 38 Hromas R, Morris J, Cornetta K, Berebitsky D, Davidson A, Sha M et al. Aberrant and update. Hum Mutat 2009; 30: 1369–1377. expression of the Myeloid Zinc Finger gene, MZF-1, is oncogenic. Cancer Res 1995; 15 Moriarty HT, Webster LR. Fragile sites and bladder cancer. Cancer Genet Cytogenet 55: 3610–3614. 2003; 140:89–98. 39 Gaboli M, Kotsi PA, Gurrieri C, Cattoretti G, Ronchetti S, Cordon-Cardo C et al. Mzf1 16 Callahan G, Denison SR, Phillips LA, Shridhar V, Smith DI. Characterization of the controls cell proliferation and tumorigenesis. Genes Dev 2001; 15:1625–1630. common fragile site FRA9E and its potential role in ovarian cancer. Oncogene 40 Albergaria A, Resende C, Nobre AR, Ribeiro AS, Sousa B, Machado JC et al. CCAAT/ 2003; 22: 590–601. enhancer binding protein β (C/EBPβ) isoforms as transcriptional regulators of the 17 Perales G, Burguete-García AI, Dimas J, Bahena-Román M, Bermúdez-Morales VH, pro-invasive CDH3/P-cadherin gene in human breast cancer cells. PLoS ONE 2013; Moreno J et al. A polymorphism in the AT-hook motif of the transcriptional 8: e55749. regulator AKNA is a risk factor for cervical cancer. Biomarkers 2010; 15:470–474. 41 Rask K, Thörn M, Pontén F, Kraaz W, Sundfeldt K, Hedin L et al. Increased 18 Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell 2011; expression of the transcription factors CCAAT-enhancer binding protein- 144: 646–674. beta(C/EBBeta) and C/EBzeta (CHOP) correlate with invasiveness of human 19 1000 Genomes Project Consortium, Abecasis GR, Auton A, Brooks LD, DePristo colorectal cancer. Int J Cancer 2000; 86:337–343. MA, Durbin RM et al. An integrated map pf genetic variation from 1092 human 42 Tsai SJ, Hwang JM, Hsieh SC, Ying TH, Hsieh YH. Overexpression of myeloid genomes. Nature 2012; 491:56–65. zinc finger 1 suppresses matrix metalloproteinase-2 expression and reduces 20 Mao L, Yang P, Hou S, Li F, Kijlstra A. Label-free proteomics reveals decreased invasiveness of SiHa human cervical cancer cells. Biochem Biophys Res Commun expression of CD18 and AKNA in peripheral CD4+ T cells from patients with Vogt- 2012; 425:462–467. Koyanagi-Harada syndrome. PLoS ONE 2011; 6: e14616. 43 Torres-Poveda K, Burguete-García AI, Cruz M, Martínez-Nava GA, 21 French RR, Chan HTC, Tutt AL, Glennie. MJ. CD40 antibody evokes a cytotoxic Bahena-Román M, Ortíz-Flores E et al. The SNP at − 592 of human IL-10 gene is T-cell response that eradicates lymphoma and bypasses T cell help. Nat Med 1999; associated with serum IL-10 levels and increased risk for human papillomavirus 5: 548. cervical lesion development. Infect Agent Cancer 2012; 7:32. 22 Díaz-Benítez CE, Navarro-Fuentes KR, Flores-Sosa JA, Juárez-Díaz J, Uribe-Salas FJ, 44 Reese MG. Application of a time-delay neural network to promoter annotation in Román-Basaure E et al. CD3zeta expression and T cell proliferation are inhibited the Drosophila melanogaster genome. Comput Chem 2001; 26:51–56.

Genes and Immunity (2015) 43 – 53 © 2015 Macmillan Publishers Limited Akna in cervical cancer GA Martínez-Nava et al 53 45 Levitsky VG, Ignatieva EV, Ananko EA, Turnaev II, Merkulova TI, Kolchanov NA 48 Sherry ST, Ward MH, Kholodov M, Baker J, Phan L, Smigielski EM et al. et al. Effective transcription factor binding site prediction using a combination of dbSNP: the NCBI database of genetic variation. Nucleic Acids Res 2001; 29: optimization, a genetic algorithm and discriminant analysis to capture distant 308–311. interactions. BMC Bioinformatics 2007; 8: 481. 49 Flicek P, Ahmed I, Amode MR, Barrell D, Beal K, Brent S et al. Ensembl 2013. Nucleic 46 Heinemeyer T, Wingender E, Reuter I, Hermjakob H, Kel AE, Kel OV et al. Databases Acids Res 2013; 41 (Database issue): D48–D55. on transcriptional regulation: TRANSFAC, TRRD, and COMPEL. Nucleic Acids Res 50 Chen X, Weaver J, Bove BA, Vanderveer LA, Weil SC, Miron A et al. Allelic 1998; 26: 364–370. imbalance in BRCA1 and BRCA2 gene expression is associated with an increased 47 Chekmenev DS, Haid C, Kel AE. P-Match: transcription factor binding site search breast cancer risk. Hum Mol Genet 2008; 17: 1336–1348. by combining patterns and weight matrices. Nucleic Acids Res 2005; 33(Web 51 Lewontin RC. The interaction of selection and linkage. I. General considerations; Server issue): W432–W437. heterotic models. Genetics 1964; 49:49–67.

Supplementary Information accompanies this paper on Genes and Immunity website (http://www.nature.com/gene)