Lysyl oxidase polymorphisms inuence the risk of cancer: an update meta-analysis

The genetics of cancer metastasis is important for designing optimal therapeutic strategies. The lysyl oxidase (LOX) gene has been found important in the metastatic process, with roles in setting the microenvironment for future metastatic sites. Associations between the LOX polymorphisms (473G/A and -22G/C) have been examined in several studies, however, results were inconsistent, prompting a meta-analysis in order to obtain more precise estimates. Searches of six databases yielded 14 articles (15 studies) that examined associations of 473G/A and -22G/C with cancer. We examined ve cancer groups: breast, lung, bone (osteosarcoma), GIC (gastrointestinal cancers) and GYC (gynecological cancers). For each cancer group, we calculated pooled odds ratios (ORs) and 95% condence intervals (CIs) using standard genetic models. High signicance (P a < 0.00001), homogeneity (I 2 = 0%) and high precision of effects (CI difference < 1.0 [upper CI-lower CI]) comprised the three criteria for strength of evidence (SOE). Multiple comparisons were Bonferroni-corrected. Sensitivity analysis assessed robustness of the outcomes. Thirteen signicant associations indicating increased risk (OR > 1.00) were found in all cancer groups except breast (P a = 0.10-0.91). Of the 13, two were in osteosarcoma where the -22G/C effects (ORs 4.05-4.07, 95% CIs 1.30-12.70, P a = 0.02) were homogeneous (I 2 = 0%) but imprecise (CIDs 11.4) and did not survive the Bonferroni correction. In contrast, the Bonferroni-surviving dominant/codominant outcomes in lung cancer (OR 1.44, 95% CI 1.19-1.74) and GYC (ORs 1.52-1.62, 95% CIs 1.26-1.88) met all three SOE criteria (P a = 0.00001, I 2 = 0%, CIDs 0.49-0.56). In summary, associations of LOX 473G/A with lung, ovarian and cervical cancers indicate 1.4-1.6-fold increased risks. These outcomes were underpinned by robustness and high statistical power at the aggregate level. (i) precision was low in the homozygous/recessive models but not in the dominant/codominant models; (ii) GYC outcomes were homogeneous (I 2 = 0%) but not in GIC (I 2 = 30-61%).


Introduction
Between 70-90% of cancer deaths result from metastasis, whereby the cancer has spread through the body [1]. In metastasis, cancer cells form new tumors far from the location where cancer was rst detected (primary tumor) [2].
Metastasis occurs when cancer cells from the primary tumor invades the surrounding tissue, use the lymph and/or blood to travel through the body, then enter a distant organ (extravasate), settle in the new microenvironment and proliferate to form a secondary tumor [3]. Ability of the extravasated cancer cells to grow depends on features that are inherent to both the cancer cells and target organ and the active interplay between these two [4]. These interactions underpin the complexity of metastasis, given that this systemic process involves nonmalignant host cells in both primary and secondary sites [5]. Metastatic transformation is a driving factor in cancer research because treatments are more successful before metastasis has occurred than after. Thus, the pivotal role of metastasis in determining the success of cancer treatments depends on thorough understanding of this cancer phenomenon [6]. Metastasis results from genetic and epigenetic alterations in pathways involving proteins that mediate cell invasion, survival outside of the primary tumor microenvironment, and colonization at a distant organ site [7]. Lysyl oxidase (LOX) is a protein that is involved in the etiology of cancer metastasis because of its functional role affecting signaling, transcription and translation, which alters cell adhesion, motility and proliferation resulting from increased extracellular matrix (ECM) deposition [8]. Elevated expression of LOX was found to signi cantly correlate with increased metastasis and reduced patient survival [9]. Thus, involvement of LOX in multiple stages of metastasis [10] and its role the metastatic milieu of various cancers [11][12][13][14] renders this protein a useful clinical target [15]. Furthermore, LOX accumulation in future metastatic sites [9] renders the gene for this protein important in understanding its emissary role in metastasis.
The LOX gene has seven exons that encode several functional domains of the LOX protein [16]. LOX undergoes a series of transformations with size changes expressed in kilo Daltons (kDa) from a preproenzyme (46 kDa) to a proenzyme (50 kDa) to a propeptide (18 kDa) and ends up as a functional protein (32 kDa) in the ECM [17]. The LOX gene has an important single-nucleotide polymorphism (SNP) located at exon 1 of chromosome 5q23.1-q23 (rs1800449). At this location, the open reading frame at position 473 contains the guanine (G)-adenine (A) bases [18]. A shift from 473G to 473A changes the amino acid arginine (Arg) at residue 158 to glutamine (Gln) (Arg158Gln) in the LOX propeptide [16]. Since it was discovered [16], LOX polymorphisms (473G/C and -22G/C) have been closely studied for their relationship with carcinogenesis [5,10,19]. At the gene level, single-study reports of LOX SNP associations with cancer have not been consistent. It is thus opportune to statistically synthesize the ndings of these studies using meta-analysis. Here, we examine the role of the LOX SNPs in the risk of cancer metastasis, which might guide potential future directions in cancer genetics. To obtain less ambiguous, clearer estimates of the role of SNPs in this investigation, we assessed the strength of evidence (SOE) using statistical and meta-analytical criteria. This study aims to highlight the genetic role of LOX polymorphisms in cancer metastasis and to provide information that could be useful in clinical decision making.

Selection of studies
We searched MEDLINE using PubMed, Google Scholar, Scopus, Mednar, Wanfang and CNKI (China National Knowledge Infrastructure) databases for association studies as of August 11, 2020. The terms used were "Lysyl oxidase", "protein-lysine 6-oxidase", "LOX", "polymorphism" and "cancer" as medical subject headings and text. References cited in the retrieved articles were also screened manually to identify additional eligible studies. In case of duplicates, the article with the most recent date was selected. Inclusion criteria were (i) case-control studies evaluating the association between the LOX polymorphisms and cancer risk and (ii) su cient genotype frequency data presented to calculate the odds ratios (ORs) and 95% con dence intervals (CIs). The exclusion criteria were as follows: (i) reviews; (ii); (ii) articles that were not case-control studies; and (iii) studies with genotype data that could not be used to calculate ORs and 95% CIs.

Data extraction
Two investigators (RM and NP) independently extracted data and arrived at consensus. The following information was obtained from each publication: cancer group, family name of the rst author, year of publication, the country of origin, ethnicity, LOX SNP, primary tumor site, study-speci c association of the LOX SNP with cancer from each publication with their respective 95% CIs and P-values, status of the controls, genotyping platform, basis for matching the controls with cases, and study features needed to tally scores for the Newcastle-Ottawa Scale (NOS).

LOX polymorphisms and cancer groups
We examined two LOX polymorphisms in ve cancer groups: -22G/C in (i) osteosarcoma (bone cancer) and 473G/A in the other four cancer groups that included (ii) breast, (iii) lung, (iv) gastrointestinal cancers (GIC) and (v) gynecological cancers (GYC). Three and two cancer types comprised GIC (oral, gastric and colorectal) and GYC (cervical and ovarian), respectively.

Quality of the studies
The NOS [20] was used to assess quality of the included studies. NOS scoring is based on three broad perspectives: selection, comparability, and exposure in case-control studies. The star rating system has scores ranging from zero (worst) to 9 (best). Scores of 5-6 and ≥7 stars indicate moderate and high quality, respectively.
Statistical power and Hardy-Weinberg equilibrium (HWE) Using the G*Power program [21], we evaluated statistical power. Meta-analyses in cancer genetics have used the ORs of 1.2 and 1.5 to assess statistical power [22]. Thus, at these OR levels with a genotypic risk level of α = 0.05 (twosided) and 5% minor allele frequency (maf), power was considered adequate at ≥ 80%. HWE was assessed with the application in https://ihg.gsf.de/cgi-bin/hw/hwa1.pl. A P-value of < 0.05 indicated deviation from the HWE.

Data synthesis
Examining two LOX polymorphisms (473G/A and -22G/C) warranted the use of a common notation indicating var and wild-type (wt) alleles. Supplementary Table S2 includes a column for the minor (var) allele in both polymorphisms. After estimating cancer risk (OR) for each study, pooled ORs with 95% CIs were calculated for each of the ve cancer groups in the following genetic models: (i) homozygous: (var-var and wt-wt) genotypes compared with wt-wt, (ii) recessive: (var-var versus wt-var + wt-wt), (iii) dominant: (wt-wt versus wt-var + var-var), and (iv) codominant: (var versus wt). Three indicators were used for strength of evidence (SOE): First, highly signi cant P-values (P a = 0.00001) most likely to survive the Bonferroni correction, which was performed with Microsoft Excel (Microsoft, Redmond, WA, USA). Second, highly precise effects were assessed with the con dence interval difference (CID = upper CI-lower CI). High (> 1.0) and low (< 1.0) CID values indicate low and high precision, respectively [23]. Third, homogeneity was assessed with the I 2 metric, expressed as 0% [24]. In meta-analysis, however, studies differ from each other [25]. This heterogeneity was estimated with the c 2 -based Q test [26] where signi cance was set at P HET < 0.10. The randomeffects model (DerSimonian-Laird) [27] was used in the presence of heterogeneity [24] and the xed-effects model (Mantel-Haenszel) [28] in its absence. Summary effects that met the SOE criteria were tested for robustness, with use of sensitivity analysis, which involves serial omission of the studies followed by recalculation of the pooled OR.
Signi cant outcomes (P a < 0.05) with ≥ 10 studies warranted assessment for publication bias. Except for heterogeneity estimation [26] two-sided P-values of £ 0.05 were considered signi cant. Data for the meta-analysis were analyzed using Review Manager 5.3 (Cochrane Collaboration, Oxford, England), SIGMASTAT 2.03, and SIGMAPLOT 11.0 (Systat Software, San Jose, CA).

Results
Characteristics of the included studies Figure 1 outlines the selection process in a owchart based on guidelines from the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [29] with a checklist detailing the description of this meta-analysis (Supplementary Table S3). A total of 504 citations were identi ed from the initial search, the screening of which yielded 22 full-text articles. Of the 22, eight were excluded for not conforming to the inclusion criteria. Table 1 lists the 14 articles [30][31][32][33][34][35][36][37][38][39][40][41][42][43] included in this study , seven [30,31,33,36,38,40,43] of which were new additions to the metaanalysis literature on LOX-cancer. Two Chinese language publications [44,45] were duplicates (excluded from this study) of the English language article [36] included in this study. This article [36] examined two cancer types (lung and colorectal), which were treated as two studies. Subjects were all Asians except in two publications in breast cancer. Three, two and two articles focused on breast, lung and bone, respectively. Four articles each examined GIC and GYC.
Age (mean ± standard deviation years) of the patients were predominantly 50s to 60s in all cancers except two studies in GYC (38.2 ± 9.2y) and osteosarcoma ( Aggregate statistical power (ASP) in the ve cancer groups were adequate at an OR = 1.5 (82.5%-99.9%), but not at an OR = 1.2, where only GIC was adequately powered (91.8%). Five studies were HW-non-compliant covering lung and bone cancers and all but one in GYC (Supplementary Table S2).

Overall and subgroup analysis
This meta-analysis yielded 28 comparisons (Tables 2 and Supplementary Table S4), of which 18 were non-signi cant (P a > 0.05), found in breast, bone and lung cancers. Thirteen outcomes were signi cant (P a < 0.05), all indicating increased risk (ORs 1.36-4.07). The low number of studies precluded assessment of publication bias.

Lung cancer and osteosarcoma
The lung cancer (473G/A) and osteosarcoma (-22G/C) comparisons were each based on two studies collectively yielding eight outcomes (Table 2). Of the eight, ve were signi cant (P a < 0.05), three of which survived the Bonferroni correction, all in lung cancer (P a < 0.0001). Of the three, only the codominant result was homogeneous (I 2 = 0%) which, with high precision (CID 0.55), met all three SOE criteria. In osteosarcoma, two signi cant outcomes (P a = 0.02) in the homozygous/recessive models had high magnitude (ORs 4.05-4.07). However, their imprecise effects (CIDs 11.35-11.39) and failure to survive the Bonferroni correction warrant caution in interpreting the risk that -22G/C poses for bone cancer.

GIC and GYC
Of the eight GIC and GYC signi cant outcomes, seven survived the Bonferroni correction (Table 2). These highly signi cant (P a < 0.0001) pooled ORs presented a dichotomy of precision effects, low in homozygous/recessive (CIDs of 1.63-3.14), high in dominant/codominant (CIDs 0.40-0.56). Figure 2 visualizes of the difference between low and high precision studies in GYC. The diamond was broader and horizontal lines from each study in the homozygous plot were longer (CID: 1.78, low precision) compared to the shorter lines (CID: 0.49, high precision) and narrower diamond in the codominant plot.

Summary of ndings
Given the different clinical manifestations, etiologies and progression in the ve cancer groups, we conducted the meta-analysis by cancer group, which reduced the number of studies (n = 2-4). However, each study contributed to the aggregate sample size that resulted in adequate to high ASP in all ve cancer groups (82.5-99.9% at an OR = 1.5) (Supplementary Table S2). This OR level has been used in previous studies that explored associations of genetic polymorphisms with cancer [46]. Breast cancer was the only comparison to yield non-signi cance (P a = 0.10-0.91) in all genetic models (Table 2 and Supplementary Table S4). Study-speci c ORs from the three component studies were unsurprisingly non-signi cant for the var 473G/A genotype in this ethnically heterogeneous cancer group (Table 1). These three articles have examined the in uence of ER status in breast cancer risk, where two reported signi cant outcomes in their expression studies. Min et al [31] showed signi cantly higher expression levels of LOX in ER-breast cancers compared to ER+ ones (P a < 0.05). Friesenhengst et al [30] favored the greater prognostic role of LOX expression over that of the 473G/A genotype. In contrast to the breast cancer ndings, GIC and GYC increased risk effects were signi cant in all genetic models (Table 2), which presented contrasts according the genetic model. Homozygous and recessive odds in GIC indicated 3.0 to 3.3-fold risks, more than double the odds in the dominant/codominant models (1.4-fold). In GYC, the homozygous/recessive odds were 2.7 and 2.5-fold, while that in the dominant/codominant models were 1.5-1.6-fold. Thus, for both GIC and GYC, homozygous/ recessive odds were higher than the dominant/codominant odds. Between these two cancer groups, GIC may pose greater increased risks (3.3-fold) than those in GYC (2.7-fold). However, other meta-analytical evidence need to be considered for a more complete picture of LOX genetic associations with cancer. Thus, two dichotomies delineated effects between the genetic models and cancer groups of GIC/GYC. (i) precision was low in the homozygous/recessive models but not in the dominant/codominant models; (ii) GYC outcomes were homogeneous (I 2 = 0%) but not in GIC (I 2 = 30-61%). Between the non-signi cant breast cancer and signi cant GIC/GYC outcomes in all genetic models were signi cance in some, not all genetic models of osteosarcoma and lung cancer. In the -22G/C polymorphism of osteosarcoma, the codominant null outcome agreed with the lack of signi cant association in glioma [19] but contrasted with our moderately signi cant homozygous/ recessive nding. In lung cancer, the homozygous/recessive outcomes were highly signi cant (P a = 0.00001) but imprecise (CIDs 2.61-2.71). In contrast, the codominant pooled OR met all SOE criteria (high signi cance + high precision [(CID 0.55)] + zero heterogeneity). This centralized the codominant lung cancer and dominant/codominant GYC outcomes, with evidence of association between LOX 473G/C with risk of cancer. Scaffolds that underpinned the SOE were robustness and high statistical power.
Comparison with a previous meta-analysis Table 4 details the differences between a previous meta-analysis [37] and ours. Table 1 identi es which articles were and were not in Gao et al [37]. Of note, the article on glioma [19] was in Gao et al [37] but not in ours on account of our cancer group study design. Differences in study design (overall analysis: cancer groups in our study versus pooled cancer types in Gao et al [37]) between the two meta-analyses precluded direct comparisons of the results.

Role of LOX gene and LOX protein in cancer metastasis
Metastasis is the last stage of cancer progression that warrants a good understanding of its genetic etiology. Literature on the association between LOX polymorphisms and cancer metastasis, particularly 473G/A are uncommon, with outcomes that may require more clarity. Of the 14 articles in this meta-analysis, three examined lymph node metastasis [30,32,43], where signi cant associations with the LOX genotypes (P = 0.02) were found in the ovarian cancer study of Yang et al [43] but not for breast cancer (P = 0.41). In the breast cancer study of Friesenhengst et al [30], however, their ndings involving 473A-carriers among ER-patients showed that 473G/A may increase the risk for breast cancer, particularly in ER-women with weaker outcome that involved metastasis. In their osteosarcoma study, Liu et al [34] found that the AA genotype and A allele were higher in patients with metastasis than those without metastasis indicating a signi cant 1.5 to 2.4-fold increased risk (P = 0.02-0.03) but failed the Bonferroni correction. In their ovarian cancer study, Wang et al [41] posited that 473G/A reinforces LOX signaling which may affect metastasis. At the mRNA level, high LOX expression was reported to favor metastasis and disfavor patient survival [47][48][49]. These ndings underpin the ability of LOX as a potent predictor of cancer metastasis.
Moreover, interventions that involved silencing of LOX gene expression and targeting the hypoxia pathway have been reported to suppress [50], even reverse metastasis in breast and pancreatic cancers [51]. These differential clinical outcomes underpin the complex role of LOX in cancer metastasis. Despite the complex role of LOX in cancer metastasis, this gene remains an appealing therapeutic target [10,47,52,53].

Strengths and limitations
We identi ed four limitations in our study: First, majority (12/14: 86%) of the studies had Asian subjects, indicating an underrepresentation of other ethnic groups. The two studies [30,31] with non-Hispanic Caucasian and African-American ethnicities warrant more of these two ethnic groups in future studies. Second, imprecise effects and failure to survive the Bonferroni correction of the signi cant -22G/C outcomes in the homozygous/recessive models of osteosarcoma may have decommissioned this polymorphism as a genetic risk factor for cancer, but future studies might modify this conclusion. Third, we did not explore gene-environment interactions. Four [32,36,37,39] articles mentioned gene-environment interactions but did not provide data for further analysis. However, four articles explored the LOX polymorphism associations with cigarette smoking and cancers of the lung [35,36], bone [33] and cervix [40] as well as bisphenol A (an environmental estrogen) and osteosarcoma [33]. Fourth, the core GYC and lung cancer outcomes had HW-deviating studies [35,[40][41][42], which may have posed methodological and representation bias. On the other hand, the strengths of our study include: (i) combinability of the component studies where most (54%) of the comparisons (15/28) were xed-effects and 60% (9/15) had zero heterogeneity (I 2 = 0%); (ii) most controls (13/14: 93%) were uniformly de ned (healthy or cancer-free); (iii) most tissue sources were blood specimens (12/14: 86%); (iv) most (11/14: 79%) of the articles had controls that were matched with cases, with 80% (eight articles based on age); (v) all signi cant core outcomes were robust.

Conclusion
We have presented evidence for the role of the LOX polymorphisms in increasing cancer risk, GYC and lung cancer in particular, which suggest that 473G/A might be a useful susceptibility cancer marker. However, a single locus effect on cancer will likely be small given the involvement of other factors, such as gene-gene interactions. All 14 publications focused only on LOX. Functional studies have shown that other genes such as hypoxia-inhibiting factor 1 (HIF-1) transforming growth factor -beta (TGFβ), and interferon-gamma (IFNγ) interact with LOX to regulate metastasis [10,15,54,55]. More studies based on sample sizes commensurate with the detection of small genotypic risks should allow more de nitive conclusions about the association of the LOX polymorphisms and cancer.  GIC: gastrointestinal cancers (oral, gastric, colorectal); GYC: gynecological cancers (cervical, ovarian); all cancer groups examined 473G/A unless otherwise specified; n: number of studies; OR: odds ratio; CI: confidence interval; CID: confidence interval difference; P a : P-value for association; P HET : P-value for heterogeneity; I 2 is a measure of variability attributed to heterogeneity; values in bold indicate significant associations; *survived the Bonferroni correction. Table 3 Main outcome summary of lysyl oxidase 473G/A and cancer