This analytic sample consisted of 338 individuals (1:1 matched cases and controls) who participated in the baseline collection (2009–2015) of the Atlantic PATH study, which is part of the largest prospective cohort study in Canada, the Canadian Partnership for Tomorrow’s Health (CanPath, formerly the Canadian Partnership for Tomorrow Project). Data holdings included questionnaire data, physical measures and biosamples, which consisted of blood, urine, saliva, and toenail samples, collected both in person and at home. Details on the study population and data collection in the Atlantic PATH cohort is described in detail elsewhere (38, 39). This study used a case–control study design, matched on sex and age (+/−5 years) for each of the cancer types, individually. The cancer types of interest were breast, cervical, prostate, and skin cancer. Cancer cases were randomly selected for inclusion if participants had a history of the cancer of interest, but no other health conditions. Among the cases, there were 41 individuals with a history of breast cancer, 41 with a history of cervical cancer, 44 with a history of prostate cancer, and 43 with a history of skin cancer. The randomly selected healthy matched controls must have never had a history of cancer, diabetes, or cardiovascular disease. Participants’ toenail samples were analyzed for metallomes and arsenic species and compared between cases and controls (Figure 1).

Participant demographics and characteristics are provided in Table 1. The mean ages across the groups ranged from 50–60 years, and the skin cancer group was 60% women. Participants largely sourced their water from municipal sources (34–51%) or private wells (12–27%), had moderate to high levels of physical activity, were overweight or obese, and were not current smokers. Data collection and participant characteristics are described in greater detail in the Atlantic PATH Cohort Profile (38).

The method used to digest and analyze toenails has been previously described and validated (33, 34). In brief, toenail samples were weighed to an approximate mass of 50 mg and transferred to a 10 mL quartz vial for cleaning. Toenails were submerged, sonicated, and rinsed with acetone, and this process was repeated using deionized water (Milli-Q Advantage A10 unit, MilliporeSigma, ON Canada). Cleaned samples were then dried in a Heratherm 60 L gravity convection oven (Thermo Scientific, MA USA) at 105°C overnight. Toenail samples were digested using a Discover SP-D microwave digestor (CEM Corporation, NC USA). In each sample vial 100 μL of concentrated nitric acid (HNO₃), 500 μL of hydrogen peroxide (H₂O₂), and 400 μL of water were added. Three blanks were created with the same solution and analyzed alongside each batch. Digested samples were transferred to 15 mL polypropylene tubes and diluted tenfold with water; the final concentration of HNO₃ was 1% (v/v). For the speciation analysis, a 995 μL aliquot of each sample and the blanks were transferred to 1.8 mL polypropylene vials (Thermo Scientific, MA United States). Finally, 5 μL of 20 μg/L arsenobetaine (AsB, Sigma Aldrich, MO United States) in 1% v/v HNO₃ was added to each vial as an internal standard. AsB was chosen as internal standard as it represents a small fraction of total As found in toenails and is often undetectable (33, 40, 41). Samples were vortexed (Maxi Mix I, Thermo Scientific, MA United States) immediately before Inductively Coupled Plasma-Mass Spectrometry (ICP-MS) and High Performance Liquid Chromatography/ICP-MS (HPLC-ICP-MS) analysis to ensure homogeneity.

Total concentration of arsenic and 22 other heavy metals (Li, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Ga, Se, Rb, Sr, Ag, Cd, Sb, Ba, Hg, Tl, Pb, Th, and U) in the toenail samples was measured using an ICP-MS (iCAP Q, Thermo Scientific, MA United States). An internal standard of 50 μg/L scandium (Sc, AccuStandard, CT USA) in 1% (v/v) HNO₃ was added. All metallomes were analyzed in kinetic energy mode, with the exception of selenium (Se), which was measured in standard mode. High purity helium (>99.999% He) was used as the collision gas. The multi-element calibration standard was diluted to concentrations of 0.01, 0.05, 0.1, 0.5, 1, 5, 10, 50, and 100 μg/L in 1% (v/v) HNO₃ to form a nine-point calibration curve. Quality control check standards (1 and 10 μg/L) were measured every 15 samples. Qtegra Intelligent Scientific Data Solution software (version 2.7, Thermo Scientific, MA United States) was used to collect and process data from total arsenic analysis.

To determine the levels of arsenic species, a separate analysis was performed using HPLC coupled with ICP-MS using a previously validated method (33). Briefly, the HPLC was fit with a P4000 pump, AS3000 autosampler, and SN4000 interface module. PEEK tubing was used to pair the HPLC with the ICP-MS. Sample injection and measurement was coordinated using the Qtregra Intelligent Scientific Data solution software. Arsenic species were separated with an anion exchange column (IonPac AS7, 250×2 mm, Thermo Scientific, MA, United States) and guard column (IonPac AG7, 50×2 mm, Thermo Scientific, MA, United States). Ammonium carbonate was used as the mobile phase using a gradient solution between 20 mM and 200 mM to achieve adequate separation for the arsenic species measured. A calibration stock solution was made by mixing individual arsenic (III), arsenic (V), MMA, and DMA standards. The stock solution was then diluted to 0.02, 0.05, 0.1, 0.2, 0.5, 1, and 2 μg/L in 1% (v/v) HNO₃ to form a seven-point calibration curve.

The method was previously validated and published (33, 34), but is briefly described hereafter. To date, there is no certified reference material for trace metal concentrations or arsenic speciation in toenails. Validation of the analytical methods was performed instead using certified human hair and frozen urine. NIES No. 13 Human Hair (National Institute for Environmental Studies, Tsukuba, Japan) samples were prepared using the toenail protocol. The method was tested for different sample masses (as little as 5 mg), as not all toenail samples had sufficient mass (50 mg). NIST 2669 Arsenic Species in Frozen Urine were used to validate the arsenic speciation method.

Method detection limits (MDL) were calculated using United States Environmental Protection Agency procedures (EPA 821-R-16-006) and the method blanks that were analyzed alongside the toenail samples (Supplementary Table S1). Previous studies (42–45) have found that replacing values below MDL with MDL/√2 does not introduce bias when the percentage of values below MDL is below 25%. As such, values for arsenic species that fell below MDL were replaced with the MDL/√2 for that species before being normalized by sample mass.

After toenail analysis, the four cancer case–control groups were analyzed independently: breast cancer (n = 82), cervical cancer (n = 82), prostate cancer (n = 88), and skin cancer (n = 86). Stata 15 was used to generate descriptive statistics and analyze the data (46). Figures were generated in RStudio using the ggplot package (47, 48). Demographic and participant characteristics are described with frequencies and were compared with chi-squared tests. Arsenic species and total metallome were non-normally distributed, however, in line with the Central Limit Theorem, we can assume that for sample sizes greater than 15, the errors will be normally distributed. As such, student’s t-tests were used to compare total concentrations of nine metallomes of interest (As, Mn, Fe, Ni, Cu, Zn, Se, Cd, and Pb) and arsenic species between cases and controls in each cancer group. Multivariate analysis of variance (MANOVA) was used to compare the profiles of arsenic species (%MMA, %DMA, %iAs, PMI, SMI) between cases and controls in each group. The models were then replicated using multivariate analysis of covariance (MANCOVA) to determine whether the covariates were significant predictors of the profiles of arsenic species. The correlations of metallomes (18) for cases and controls in each group were evaluated. All statistical tests were evaluated with a significance level of α = 0.05.

The predominant form of arsenic found in toenails was iAs, accounting for more than 80% of total arsenic. Moreover, when compared to their respective controls, the concentrations of iAs were significantly higher among both cervical and skin cancer cases (Figure 2). Mean concentrations of arsenic species (Supplementary Table S2) were more varied between cancer cases than between controls: iAs ranged from 0.054–0.074 μg/g, MMA ranged from 0.0045–0.0055 μg/g, and DMA ranged from 0.0039–0.0049 μg/g. While the concentrations of MMA and DMA in toenails did not significantly differ between cases and controls in any of the groups, the PMI (ratio of MMA to iAs) and SMI (ratio of DMA to MMA) was significantly different between cases and controls in the skin and cervical cancer groups, respectively.

In multivariate analysis, arsenic speciation profiles were significantly different between cases and controls in the breast cancer (p = 0.0330), cervical cancer (p = 0.0228), and skin cancer (p = 0.0085) groups (Supplementary Table S3). The arsenic speciation profiles patterns were not uniform across groups, and significant mean differences were observed among different variables across the groups. Adjustment did not appreciably alter the observed mean differences, but many increased in magnitude (Figure 3). The adjusted models for the breast, cervical, and skin cancer groups indicated the selected covariates were not significant predictors of arsenic speciation. By contrast, the adjusted model for the prostate cancer group was significant (p = 0.0373); significant predictors of arsenic speciation were province of residence (p = 0.0004) and smoking status (p = 0.0095). However, after adjustment, none of the arsenic speciation variables were statistically significantly different between prostate cancer cases and controls.

Significantly higher concentrations of zinc were measured among cases compared to controls in the breast (p = 0.0412), prostate (p = 0.0116), and skin (p = 0.0011) cancer groups (Figure 4). In the breast cancer group, lead concentrations were higher among cases compared to controls (p = 0.0385). In the cervical cancer group, higher concentrations of iron were observed among cases compared to controls (p = 0.0291). Prostate cancer cases had statistically significantly higher selenium concentrations (p = 0.0116) when compared to their controls (Figure 4). The means and standard deviations (SD) of the arsenic species (μg/g), methylation indices (PMI, SMI), and total metallomes (μg/g) are provided in Supplementary Information (Supplementary Table S1).

In multivariate analysis, profiles of metallomes were significantly different between cases and controls in the prostate cancer (p = 0.0244) and skin cancer (p = 0.0321) groups in unadjusted analyses, whereas the breast and cervical cancer models were non-significant. However, significantly higher mean concentrations of zinc were observed in the breast cancer (p = 0.041), prostate cancer (p = 0.012), and skin cancer (p = 0.001) cases, higher mean iron was observed in cervical cancer cases (p = 0.029), and higher mean selenium was observed in prostate cancer cases (p = 0.012). The mean differences did not appreciably change between unadjusted and adjusted analysis (Supplementary Table S4). In the MANCOVA analysis, only the prostate cancer model remained statistically significant (p = 0.0194), and significant differences in zinc and selenium concentrations between cases and controls persisted. None of the covariates in this model were significant predictors of metallome profiles.

Of the 23 elements measured, only 18 were consistently above the MDLs (V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Ga, As, Se, Rb, Sr., Cd, Tl, Pb, Th, and U). The concentrations of several metals (metallomes) were significantly correlated in many of the groups (Figure 5). Among breast cancer cases, significant positive correlations were observed in the concentration of arsenic-manganese (r = 0.61, p = 0.001) and manganese-lead (r = 0.55, p = 0.0081), whereas positive correlations between cadmium-lead (r = 0.52, p = 0.0157) and arsenic-cadmium (r = 0.60, p = 0.0010) were observed among the cervical cancer cases. Significant correlations were also observed between copper-lead in the breast cancer control group (r = 0.54, p = 0.0082) and in cervical cancer control groups (r = 0.49, p = 0.0379). In addition, the concentrations of manganese-iron were positively and significantly correlated in both the cervical cancer control group (r = 0.59, p = 0.0015) and the prostate cancer control group (r = 0.47, p = 0.0465), while manganese-lead were correlated in both the cervical cancer (r = 0.54, p = 0.0091) and skin cancer (r = 0.50, p = 0.0250) control groups. Significant positive correlations were also observed in the concentrations of iron-copper (r = 0.49, p = 0.0295) and iron-lead (r = 0.54, p = 0.0072) in the skin cancer control group.

The proportions of arsenic species measured in toenail samples are markedly different than those previously measured in urine, but are comparable to those found in toenails (33, 49, 50). Our previous work has found that toenails better capture arsenic speciation profiles and exposure to iAs when compared to urine samples (33). The compositional differences of arsenic species in these biomarkers may be explained by differential exposure levels or abilities to process iAs between study populations, or differences in the properties of the metabolites themselves. For example, iAs (III) has been shown to have a strong affinity to keratin (51). It is likely these biomarkers are complementary measures of arsenic exposure, but given our previous findings, toenails appear to be a more advantageous biomarker for evaluating arsenic methylation capacity, and thus, the risk of associated cancers.

Results here show significantly higher toenail zinc concentrations among breast, prostate, and skin cancer cases compared to controls. The current evidence linking zinc and cancer in general highlights the adverse effect of deficiencies (26). One study (63) measuring trace metals in fingernails of breast cancer cases and controls found similar zinc concentrations in their control group (103.24, SD = 38.96 μg/g versus 105.15, SD = 12.10 μg/g in the present study), but much lower concentrations among breast cancer patients (70.06, SD = 24.12 μg/g versus 111.85, SD = 16.76 μg/g in the present study). Another study found comparable zinc nail concentrations in prostate cancer patients (114.2, standard error = 4.53 μg/g) but non-significant differences between cases and controls (64). The differences between cases and controls we observed here may be explained by the varied zinc levels across populations, as well as differences in case definition (29, 65). For example, using a metallomics approach, Schilling and colleagues (2020) explained that higher urinary zinc concentrations may be caused by the downregulation of metal-binding transporters that promote the development and growth of cancer cells. In the results we report here, higher nail-zinc concentrations could be indicative of poor uptake of zinc, and homeostasis may be important in pathogenesis (24). However, critical gaps remain on zinc toxicokinetics and the associations between zinc levels and its association with cancer. Previous research has indicated that the degree of zinc exposure may play a key role in determining its association with cancer. Specifically, the effect of zinc deficiency has been shown to be different from overexposure, but both may increase cancer risk. While a deficiency in zinc may be a risk factor due to its antioxidant properties, over-exposure to zinc has been linked with skin cancers (66).

The findings of this research showed higher toenail-lead concentrations among breast cancer cases compared to controls. Lead concentrations were higher among cervical and prostate cancer cases compared to their respective controls, but the difference was not statistically significant. Another study found prostate cancer cases had statistically non-significant higher lead concentrations in nails compared to controls, but the concentrations were much higher (100-fold) in both groups compared to the present study (64).

Despite limited evidence from human and animal studies, the IARC has classified lead as a possible carcinogen (1, 67), and there is some evidence to suggest there may be an association with lung, stomach, and brain cancers (32). In addition, little research has been conducted on the use of nails as a biomarker for lead exposure. Our findings suggest there may be an association between lead concentrations and breast cancer. This result is consistent with a previous study of metallome in breast cancer patients (25). However, further research is required to ascertain the ability of toenails to both measure lead body burden, and to assess potential interactions between lead and other metals, such as arsenic.

The present study found no association between toenail cadmium concentration and cancer, despite there being some limited evidence to suggest cadmium is associated with lung, prostate, bladder, and breast cancers (68, 69). Some observational studies using nail samples have found cadmium concentrations to be higher in prostate cancer cases than in healthy controls (64, 70). Nail samples are considered a reliable biomarker for cadmium exposure, especially as they capture cadmium accumulation over a long period of time (71, 72). However, in the present study, toenail cadmium concentrations (0.007–0.010 μg/g) were much lower compared to those (1.11 ± 0.83 μg/g) previously found in nail samples (68). Considered together, the lack of observed association between cadmium and the cancer types in this study may be the result of low cadmium exposure among Atlantic Canadian. However, further research is needed.

Results we obtained show no association between toenail total arsenic concentrations and cancer. Arsenic exposure has been previously associated with liver, lung, urinary bladder, skin, kidney, breast, and prostate cancers using other biomarkers (8, 73). Recent evidence, however, has found that arsenic metabolism is the key mechanism behind arsenic-related cancers via the genetic polymorphisms involved in the process and the varied toxicity of the metabolites (74, 75). Our findings that arsenic speciation profiles—not total arsenic concentrations—are linked to history of several cancer types, provide further evidence for the important role of arsenic methylation and the need to use relevant, long-term biomarkers that can capture the variation in these profiles.

The strengths of this study include the use of a novel toenail biomarker, new analytical methods for toenail arsenic species and metallomes, and the large Atlantic PATH cohort to assess arsenic and trace metal exposure and their roles in less studied arsenic-related cancers. Currently, urine is the gold-standard biomarker to measure total concentrations of trace metals, however the use of urine is limited as it only captures a short exposure window. For the purposes of monitoring chronic exposure, as is the case for cancer risk factors, this is a critical shortcoming. The advantages of using toenails include non-invasive sample collection, and the ease and duration of sample storage. In addition, this research has used a newly developed analytical methods to measure arsenic species in toenails, which yields higher extraction efficiencies than previously observed. Moreover, this approach undergone method validation using two certified reference materials. To our knowledge, this study fills a major gap in the literature and is the first to investigate arsenic exposure and arsenic speciation among cervical cancer cases, and among the first to do so with toenails among breast, prostate, and skin cancers. Furthermore, this research is some of the first to employ a metallomics approach using toenails to compare profiles between cases and controls in several cancer groups.

Nonetheless, we note several limitations. The data in this study are cross-sectional and rely on prevalent cancer cases, and without temporal order, there is no way to ascertain causality between arsenic exposure/speciation and cancer. This limitation will be addressed in subsequent studies thanks to the longitudinal nature of Atlantic PATH, which will allow the use of incident cases. In addition, the data do not distinguish between melanoma and non-melanoma skin cancers. Many of the covariates used in this study were sourced from the questionnaire and some had missing data. Given the nature of the Atlantic PATH cohort, these findings may not be generalizable to the entire general population. The observed differences in the Atlantic PATH participants compared to the general population include higher overall socioeconomic status and higher representations of women and people with university degrees. These differences are important as those with higher incomes and levels of scientific literacy may be more likely to engage in remediation strategies for well water as a primary source of drinking water. Finally, we believe that drinking water is the primary source of exposure to arsenic, but do not have information on the duration of the exposure.

Despite these limitations, this study makes important contributions to the current literature. Exposure to arsenic and other heavy metals through contaminated drinking water remains a problem experienced in Atlantic Canada and around the world. While the mechanisms of arsenic toxicity and carcinogenicity are yet to be fully elucidated, this research provides further evidence of the potential mediating role that arsenic species may play in cancer development. The findings of our work provide further evidence that arsenic species, rather than total arsenic, is critical in understanding cancer risk associated with arsenic exposure. In particular, our findings support the hypothesis that individuals with cancer have differential arsenic speciation and metallomics profiles compared to healthy controls, indicating that environmental arsenic exposure plays a role in carcinogenesis. In addition, the research we present here suggests that arsenic speciation profiles may differ slightly by cancer type. Further research is required to confirm these findings, to better understand the mechanisms underlying this variation, and to develop upstream population health monitoring tools in response to environmental arsenic and metallome exposure.

Future research should attempt to gain a better understanding on the association between arsenic and metal exposures and cancer. Specifically, more exposure biomarker studies using a biological sample for chronic exposure are needed to understand how arsenic exposure induces cancer. Furthermore, additional method development for arsenic speciation analysis in toenails should focus on the ability to differentiate between trivalent and pentavalent arsenicals in biomarkers that would provide a more comprehensive understanding of methylation and its role in toxicity. From a population health and epidemiological perspective, the next steps include incident case analysis, accounting for the time elapsed since the first cancer diagnosis, and a larger sample. Our future research will include the use of incident cases and larger sample numbers which help to determine more definitive inferences about potentially causal mechanisms. Furthermore, our findings were consistent with previous evidence that metallome concentrations are interrelated and may generate synergistic or protective effects. As such, future research should analyze arsenic speciation and important metallomes together in profile analysis to determine whether there are any associations or interactions with methylation capacity. Lastly, given some higher arsenic and metallome profiles found in this study among cancer survivors compared to controls and their link with oxidative stress, future studies should attempt to examine their role as a possible environmental mediator of poor oncological outcomes following active forms of cancer treatments.