Metal(loid)s and human semen quality: The LIFE Study

Abstract

Multiple studies have demonstrated a global population-wide decline in semen quality, with sperm concentrations having fallen 50 % over the past 50 years. Several metal and metalloid (“metal(loid)”) compounds are known to have testicular toxicity, raising concerns about their contribution to rising infertility. In the male reproductive tract, metal(loid)s can reduce semen quality and disturb function both directly, by inducing tissue damage, and indirectly, by disrupting hormone production and secretion. This study assessed associations between 15 creatinine-adjusted metal(loid)s and 7 measures of semen quality among 413 reproductive-aged men recruited from 16 U.S. counties between 2005–2009. Multi-metal(loid) multivariable linear regression models estimated associations between semen quality endpoints and urinary concentrations of chromium, cobalt, copper, molybdenum, selenium, zinc, antimony, arsenic, barium, cadmium, lead, thallium, tin, tungsten, and uranium. LASSO regression was employed to select model variables and account for multicollinearity of the metal (loid)s. A positive association was observed between tin and sperm morphology (β = 4.92 p = 0.045). Chromium (β = 1.87, p = 0.003) and copper (β= −1.30, p = 0.028) were positively and negatively associated with total sperm count, respectively. With respect to DNA fragmentation, cadmium (β = 12.73, p = 0.036) was positively associated and chromium was negatively associated (β = −5.08, p = 0.001). In this cohort of U.S. population-based men, there was evidence of both positive and negative associations between specific metal(loid)s and semen quality. Additional research is needed to determine interactions between metal(loid)s within a mixture, consistent with typical human exposure, and identify sperm effects resulting from cumulative metal(loid) exposures.

1. Background

Multiple studies have demonstrated a global population-wide decline in semen quality, with sperm concentrations reportedly having fallen 50 % over the past 50 years [1–5]. Male factors, including sperm concentration, account for approximately 40–50 % of all infertility [6], and 12 % of U.S. men between the ages of 15 and 44 report experiencing infertility [7]. While the causes of these declines are largely unknown, environmental chemicals have been suspected given the rapid reduction in reported trends. Further, recent data demonstrate that semen quality is associated with overall health and mortality, reinforcing the need to identify the root causes of population-wide sperm declines [8].

Industrialization has dramatically increased environmental distribution of and subsequent human exposure to metal and metalloid chemicals (“metal(loid)s”). Several metal(loid) compounds are known to have testicular toxicity, raising concerns about their contribution to rising infertility [9,10]. In the male reproductive tract, metal(loid)s can reduce semen quality and disturb function both directly, by inducing tissue damage, and indirectly, by disrupting hormone production and secretion [11]. Metal(loid)s can also induce oxidative stress, which is known to contribute to testicular pathogenesis and male infertility [12]. Several metal(loid)s, including lead (Pb) and cadmium (Cd), can increase production of reactive oxygen species and contribute to DNA oxidative damage [13,14]. DNA fragmentation in human spermatozoa, an important indicator of semen quality, is closely correlated to oxidative stress and evidence of impaired spermiogenesis [15].

Among metal(loid)s, Cd, Pb, and arsenic (As) are well-studied for their human endocrine effects, and negative associations between these metal(loid)s and various semen quality endpoints have been identified [16–27]. Other metal(loid)s, such as tin (Sn), have not been well studied. Most research to date has relied upon occupational or clinical study populations with limited attention to general population sampling of men with metal(loid) concentrations reflective of typical environmental exposures experienced by U.S. men serving, in part, as an impetus for study. The available occupational and clinical data suggest that Cd and Pb are associated with harmful changes in semen parameters such as sperm concentration [20,28] and semen volume [20,29]. In animal studies, Pb and Cd are associated with a wide range of male reproductive effects, including reduction in gonadal size/weight [30,31], testicular necrosis [32,33], oxidative damage [34–37], spermatocyte DNA damage [36,38], and reduced spermatogenesis [39].

Importantly, some metal(loid)s (e.g., zinc (Zn), copper (Cu), and magnesium (Mg)) have known essential biological functions at trace levels, while others (e.g., As and Cd) do not. In terms of human health impacts, standard criteria for determining whether an element is essential are based upon whether untoward functional or structural abnormalities manifest in the absence or reduction of that compound, and if a reversal of the state occurs with supplementation of the element [40]. Both non-essential and essential metal(loid)s, when accumulated in large amounts, may cause metabolic disruptions that interfere with testicular function [41]. Isolating health effects associated with a single metal(loid) is challenging because 1) many metal(loid) exposures are correlated and rarely occur in isolation, and 2) many metal(loid)s have both beneficial and harmful health effects, with the outcome being heavily dependent on concentration. Furthermore, metal(loid) compounds may behave either agonistically or antagonistically depending on the biological pathway [42]. When metal(loid) exposures occur in combination (i.e., as mixtures), interaction effects can highly influence the overall impact on male reproductive function. These possibilities may account for the difference in findings across studies focusing on a single metal(loid).

Only a limited number of studies have applied variable selection statistical techniques to the evaluation of toxicity of mixtures [22,43, 44]. Some studies have considered the effects of various metal(loid)s on semen quality parameters, though none to our knowledge has done so with a focus on U.S. men recruited from the general population and for whom metal(loid) mixtures have been quantified. This study context is important for exploring whether environmental mixture exposures are associated with semen quality parameters among men who are not presenting to fertility clinics and have no known fertility impairments.

2. Methods

2.1. Study population and data collection

The study population comprised men participating in the Longitudinal Investigation of Fertility and the Environment (LIFE) Study, a population-based preconception cohort study, for whom both urinary concentrations of trace elements and semen quality endpoints were quantified. The study population and data collection methods for the parent cohort have been described in detail elsewhere [45]. Briefly, the referent cohort comprised 501 couples from 16 counties in Michigan and Texas who were discontinuing contraception in an attempt to achieve pregnancy with their partner. Eligibility criteria required participants to be at least 18 years of age, married or in a committed relationship, able to communicate in English or Spanish, and without physician-diagnosed infertility. Upon enrollment, men underwent an interview to assess medical history and lifestyle, as well as an anthropometric assessment for measurement of body mass index (BMI). Urine and semen samples were also collected. Among the 501 eligible men in the referent cohort, 28 (6 %) did not provide semen samples, 55 (11 %) did not have residual urine for quantification of metal(loid)s, and 5 (1 %) were azoospermic in both semen samples, and as a result could not be evaluated. Therefore, the study cohort comprised 413 (83 %) men.

2.2. Semen collection

Men were provided with two at-home semen collection kits and instructed in their proper use. Briefly, men collected the first semen sample the day after the enrollment interview (following two days of abstinence) and a second sample approximately one month later. Samples were shipped to the LIFE Study’s andrology laboratory via over-night delivery for next day analysis following laboratory inspection of specimen integrity. All samples were found acceptable for subsequent analysis using established protocols [45].

Semen analysis included quantification of 7 endpoints: five general measures (semen volume (mL), sperm concentration (×10⁶/mL), total sperm count (×10⁶/ejaculate), next day motility (%), and traditional morphology (%)) and two DNA measures (% fragmentation index and % high DNA stainability). While higher values are associated with better semen quality, in general, for the five general parameters, the reverse direction pertains to the two DNA measures.

Laboratory processing of semen samples included measuring temperature, turbidity, color, and liquefaction (data not shown). Semen volume was measured to the nearest 0.1 ml. Sperm motility was assessed using the HTM-IVOS (Hamilton Thorne Biosciences, Beverly, MA) computer assisted semen analysis (CASA) system. Sperm concentration and morphometry were measured using the IVOS system and the IDENT^™ stain (Hamilton Thorne Biosciences, Beverly, MA), and slides for morphology assessments were prepared by Fertility Solutions^® (Cleveland, OH). Sperm morphology was determined on a fixed, stained semen smear and classified using both strict (Tygerberg) and traditional [40] criteria. DNA fragmentation was quantified using the Sperm Chromatin Structure Assay (SCSA)^® method as applied by Evenson et al. [46]. Semen was diluted by adding 100 μl of whole semen to 500 μl of TNE buffer and frozen at 70 °C until analysis. The SCSA^® procedure was conducted on a Coulter Epics Elite Flow Cytometer using the SCSA^® program (SCSA diagnostics, Brookings, SD).

2.3. Urine collection

Urine samples were collected upon completion of the baseline interview and shipped to the New York State Department of Health (NYS DOH) Wadsworth Center (Albany, NY) for the quantification of 8 essential metal(loid)s (cobalt (Co), chromium (Cr), Cu, manganese (Mn), molybdenum (Mo), nickel (Ni), selenium (Se), and Zn) and 12 non-essential metal(loid)s (antimony (Sb), As, barium (Ba), Beryllium (Be), Cd, Pb, platinum (Pt), tellurium (Te), thallium (Tl), tin (Sn), tungsten (W), and uranium (U)). Quantification was performed using ICP-MS methods, including quality control procedures, developed for bio-monitoring studies [47].

In human biomarker analysis, there is a need to quantitatively measure contaminants at very low levels; consequently, there is a threshold below which a value cannot be accurately quantified [48,49]. To reduce potential bias in exposure measurements associated with censored data and in keeping with other published works from the LIFE Study, we used instrument-reported values without substitution of observations below the limits of detection (LOD) for all statistical analyses [50–52].

Geometric means (GM) and distributions of urinary metal(loid) concentrations of male LIFE Study participants were compared to those reported for a nationally representative population of male participants in NHANES for cycle years 2005–2010 [53] and were found to be comparable. Mean concentrations were slightly higher for As and Ba and slightly lower for Sn in male LIFE Study participants compared to male NHANES participants.

2.4. Statistical analysis

Exploratory data analysis was conducted by assessing missing data and influential observations and examining distributions of metal(loid)s and semen endpoints. All metal(loid) concentrations were creatinine-adjusted and log-transformed for normality. The original referent cohort (n = 501) was compared with the study cohort (n = 413; 82 %) to determine whether there were any systematic differences between men with and without biospecimens for analysis.

In situations where multiple variables are correlated (i.e., multicollinearity), multivariable linear regression models may yield unreliable parameter estimates [54]. Multicollinearity, evidenced by examining Pearson correlation coefficients and Variance Inflation Factors (VIF), was observed across the metal(loid)s. To assess which individual metal(loid)s were associated with each semen endpoint while simultaneously adjusting for the other metal(loid)s, LASSO regression modeling was used [55]. LASSO imposes a penalty on the absolute value of the coefficients, shrinking coefficients by a constant factor, and selecting a subset of predictors by shrinking coefficients exactly to zero for the predictors that have the least predictive value [55]. LASSO also conducts parameter estimation simultaneously with predictor selection, a computational advantage compared to traditional two-step procedures in which these processes instead occur sequentially. Because specific variable inclusion cannot be forced in LASSO analyses, metal(loid)s were adjusted for creatinine prior to modelling. Creatinine-adjusted metal(loid) concentrations, expressed in μg/g creatinine, were calculated by dividing metal(loid) concentrations (μg/L) by creatinine concentrations (100*mg/dL).

After examining metal(loid) concentration distributions, a two-step process was applied to the data: 1) penalized LASSO regression models were fitted to identify and select metal(loid) variables most likely to be predictive of each semen quality endpoint; and 2) unpenalized linear regression models were fitted with only the metal(loid) variable(s) selected in the previously implemented LASSO regression included in the models. For the penalized models, LASSO regression was run using 10-fold cross-validation (CV) to determine the optimal degree of penalization. Inputs for model selection included all detectable metal (loid)s and the following variables, chosen a priori as potential confounders based upon prior evidence: abstinence time (days) [56]; age (years) [57]; race/ethnicity (non-Hispanic white, other race/ethnicity) [58]; alcohol consumption upon enrollment (frequency per month) [59,60]; body mass index (BMI; weight in kg/height in m²) [1, 61]; educational attainment (<high school, high school diploma, some college, college degree); household income (USD) [62,63]; having previously fathered a pregnancy (yes/no) [64]; urinary creatinine (mg/dL) [65]; and current smoking status (serum cotinine (ng/mL)) [66]. The final model yielded the minimum mean-squared error (MSE) of prediction resulting from performing repeated 10-fold cross validation.

A LASSO modeling technique selects the most important predictors for each semen quality endpoint; however, it does not test the significance of the selected predictor variables (i.e., no p-values or confidence intervals are provided by the model). The purpose of the second step is to provide these values. Using exposures and covariates selected via LASSO, linear regression models were fitted to obtain unpenalized, adjusted coefficient estimates. To simplify the models and prevent over-fitting, the linear regression model included only creatinine-adjusted metal(loid)s and potential confounders that were selected by the LASSO modeling procedure. If more than one metal(loid) variable was selected by LASSO modeling for a given semen quality endpoint, all selected metal(loid)s were included in the linear regression model for that endpoint.

To examine associations between each metal(loid) and semen quality endpoint separately, multivariable linear regression was used to estimate associations between single metal(loid)s (unadjusted for creatinine) and each semen quality endpoint. All single-metal(loid) models were adjusted for the entire set of relevant covariates, including creatinine as an individual covariate in each model [65]. Eleven metal(loid)s (Sb, As, Ba, Cd, Pb, Sn, Cu, Co, Mo, Se, and Zn) were log (x+1)-transformed, and 4 (W, U, Mn, Ni) were square root-transformed for best fit. Two outlying concentrations each for Sn, Se, Cu, and Mn and one each for W, Co, Mo, and Zn were substituted with the median.

In a separate analysis intended to impose the most conservative assumptions possible, a Bonferroni correction was applied to the p-values associated with each individual test to impose an α level at 0.05 over all tests [67]. This was performed to test the robustness of the observed associations and examine the potential for ‘experiment-wise’ error rates that can result from multiple comparisons.

3. Results

Men were largely nonsmoking (62 %), college educated (62 %) and residing in households with incomes >$70,000 (67 %) (Table 1). There were no significant differences in sociodemographic characteristics between men in the referent and study cohorts.

Table 1.

Demographic characteristics of study cohort.

	Referent Cohort (n = 501)		Study Cohort (n = 413)
Characteristics	n	%	n	%
Age category (years)
19–29	176	35.1	142	34.4
30 – 34	186	37.1	158	38.3
35 – 51	139	27.7	113	27.3
College graduate or higher
No	190	37.9	153	37.0
Yes	311	62.1	260	63.0
Household income (USD)
<30 000	22	4.4	19	4.6
30 000–49 999	56	11.2	44	10.7
50 000–69 999	90	18.0	71	17.2
≥ 70 000	333	66.5	279	67.6
Alcohola
No	73	14.6	58	14.0
Yes
<Weekly	151	30.1	130	31.5
Weekly	108	21.6	89	21.6
> Weekly	169	33.7	136	32.9
Ever smoker (>100 cigarettes/lifetime)
No	311	62.1	258	62.5
Yes	187	37.3	153	37.0
Ever fathered a pregnancy
No	215	42.9	181	43.8
Yes	285	58.9	232	56.2
Characteristics	Mean	SD	Mean	SD
Age, years	31.8	4.9	31.8	4.8
BMI, kg m2	29.8	5.5	29.9	5.7
Serum cotinine, ng/mL (n = 466)	54.9	136.1	54.5	135.7
Abstinence time, # days (n = 460)	4.0	4.4	4.0	4.7

NOTE: Differences between the referent and study cohorts were evaluated by chi square test for categorical variables and by ANOVA for age in years, BMI, serum cotinine and abstinence time. None of the above differences between referent and study cohort achieved significance (p < 0.05).

^aIn the year prior to enrollment.

Fifteen metal(loid)s were frequently detected: As (96 %), Ba (91 %), Cd (99 %), Cr (69 %), Co (99 %), Cu (99 %), Mo (100 %), Pb (99 %), Tl (100 %), Sb (93 %), Se (90 %), Sn (85 %), W (59 %), U (87 %) and Zn (100 %) (Table 2), and five other metal(loid)s (Be, Mn, Ni, Pt, and Te) were sparsely detected (<20 %) and were not included for further analysis.

Table 2.

Distribution of Urinary Metal(loid)s, (n = 413).

	Median (IQR)		% > LODa
	Unadjusted Metal (loid)s (μg/L)	Creatinine-Adjusted Metal(loid)s (μg/g)a
Metal(loid)s -Known Essentiality
Cr	0.62 (0.66)	0.51 (0.4)	69.0
Co	0.29 (0.28)	0.21 (0.14)	99.3
Cu	10 (8.78)	8.05 (3.74)	98.8
Mn	0.1 (0.14)	0.07 (0.11)	18.2
Mo	46.65 (59.88)	38.36 (32.26)	100
Ni	4.6 (5.77)	3.56 (4.25)	15.3
Se	38.84 (46.53)	29 (19.34)	89.6
Zn	240.5 (346.7)	218.26 (191.89)	100
Metal(loid)s – Unknown Essentiality
Sb	0.06 (0.08)	0.05 (0.05)	93.2
As	11.59 (16.61)	9.12 (11.71)	95.6
Ba	1.86 (2.52)	1.56 (1.68)	91
Be	0.03 (0.18)	0.03 (0.13)	8.2
Cd	0.19 (0.24)	0.16 (0.1)	98.8
Pb	0.41 (0.53)	0.32 (0.31)	98.5
Pt	−0.01 (0.01)	−0.01 (0.01)	0
Te	0.08 (0.23)	0.05 (0.19)	0.7
Tl	0.14 (0.13)	0.11 (0.07)	100
Sn	0.32 (0.5)	0.27 (0.33)	85
W	0.06 (0.12)	0.05 (0.07)	58.6
U	0.004 (0.006)	0.003 (0.004)	87.2
Creatininea (mg/dL)	139.13 (130.49)

NOTE: Machine read values were used for all analyses.

^an = 371 %>LOD – percent greater than the limit of detection.

The distributions of semen quality endpoints are presented in Table 3. Participants’ semen quality values were largely within normal ranges (>88 % normal) per the World Health Organization standards [68,69].

Table 3.

Distribution of Semen Quality Endpoints.

	% Normala	Percentile
		5th	25th	50th	75th	95th
Total Sperm Count (*106)	91.3	27.7	98.3	182.7	318.1	566.3
Volume (mL)	90.6	1.0	2.2	3.2	4.4	6.8
Sperm Concentration (*106/mL)	91.8	9.6	34.7	60.8	94.7	189.0
24-hour Motility (%)	88.7	0.0	2.0	8.0	18.0	40.0
Traditional Morphologyb (%)	92.3	10.0	21.7	30.5	39.0	51.0
Sperm DNA Fragmentation (DFI; %)	93.7	5.1	8.5	12.4	18.9	33.4
High DNA Stainability (HDS; %)	99.3	2.2	3.8	5.9	9.6	19.5

^aSemen endpoints deemed normal based on WHO 5th edition criteria.

^bMeasured using traditional morphology criteria (WHO 3rd edition).

Penalized metal(loid) coefficients from the LASSO regression models are presented in Table 4. Thirteen (87 %) metal(loid)s were selected by LASSO for inclusion across all 7 semen quality endpoints: total sperm count (β ≠ 0; Cr, Co, Cu, Ba, Sn, W), semen volume (β ≠ 0; Cr, Co, U), sperm concentration (β ≠ 0; Sn), sperm motility (β ≠ 0; Cr, Mo, As, Pb, Sn), traditional morphology (β ≠ 0; Sn), DFI (β ≠ 0; Cr, Co, Cd), and HDS (β ≠ 0; Co, Cu, Sb, Cd).

Table 4.

Penalized Regression Models^a between Urinary Metal(loid) Concentrations (μg/g creatinine) and Semen Endpoints and Covariates, (n = 356).

Creatinine-Adjusted Metal (loid)s (μg/g)	Total Sperm Count (*106)	Semen Volume (mL)	Sperm Concentration (*106/mL)	Motility (%)	Traditional Morphologyb (%)	SCSA DFI (%)	SCSA HDS (%)
	Beta	Beta	Beta	Beta	Beta	Beta	Beta
Cr	1.53	0.06	0	0.34	0	−1.14	0
Co	0.44	0.04	0	0	0	0.26	−0.19
Cu	−0.56	0	0	0	0	0	−0.65
Mo	0	0	0	0.26	0	0	0
Se	0	0	0	0	0	0	0
Zn	0	0	0	0	0	0	0
Sb	0	0		0	0	0	−0.78
As	0	0	0	0.14	0	0	0
Ba	0.08	0	0	0	0	0	0
Cd	0	0	0	0	0	3.04	−1.41
Pb	0	0	0	−0.31	0	0	0
Tl	0	0	0	0	0	0	0
Sn	0.96	0	0.08	0.33	0.79	0	0
W	−0.41	0	0	0	0	0	0
U	0	−6.57	0	0	0	0	0
Other selected covariates:	abstinence time, body mass index, education	age, body mass index, education, income	abstinence time, education	race/ethnicity, study site	serum cotinine	age, abstinence time, income	abstinence time, alcohol consumption, body mass index, income, serum cotinine, study site

^aLASSO regression for each outcome included the following potential predictors for variable selection: age (continuous), abstinence time (continuous), alcohol consumption (categorical), body mass index (continuous), log-transformed urinary creatinine (continuous), education (categorical), income (categorical), previously fathered pregnancy (continuous), serum cotinine (continuous), study site (categorical), and race/ethnicity (categorical).

^bTraditional morphology standards (WHO 3rd edition); CI- Confidence Interval; SCSA- sperm chromatin structure assay; DFI- DNA fragmentation index; HDS- High DNA stainability.

Unpenalized multi-metal(loid) associations between selected metal (loid)s and semen quality endpoints are presented in Table 5. The magnitude of associations increased after adjustment for selected covariates in unpenalized models. Creatinine-adjusted urinary Cr and Cu were significantly associated (p < 0.05) with total sperm count in the adjusted models, with Cr having a positive (or beneficial) association and Cu having a negative (or harmful) association. For every one percent increase in Cr and Cu, total sperm count increased by 1.87*10⁴ and decreased by 1.30*10⁴, respectively (β_Cr = 1.87; β_Cu = −1.30). Creatinine-adjusted Cr was negatively associated with SCSA DFI (suggesting benefit) (β_Cr = −5.08, p < 0.001), while creatinine-adjusted Cd was positively associated with DFI (suggesting harm) (β_Cd = 12.73, p < 0.05). A positive association was observed between Sn and traditional morphology (β_Sn = 4.92, p < 0.05). No other associations were observed for sperm motility, sperm concentration, semen volume, or SCSA HDS in the unpenalized multi-exposure models.

Table 5.

Unpenalized Multi-Exposure Adjusted Associations between Urinary Essential Metal(loid) Concentrations and Semen Endpoints (n = 356).

	Creatinine-Adjusted Metal(loid)s (μg/g)
	Beta	95 % CI	p-value
Total Sperm Count (*106) a
Ba	0.21	−0.75, 1.17	0.6611
Cr	1.87	0.62, 3.12	0.0034
Co	1.03	−0.04, 2.10	0.0594
Cu	−1.30	−2.47, −0.14	0.0280
Sn	1.06	−0.40, 2.52	0.1557
W	−1.24	−4.96, 2.47	0.5098
Sperm Traditional Morphology b
Sn	4.92	0.12, 9.72	0.0445
SCSA DFI (%) c
Cd	12.73	0.85, 24.61	0.0358
Cr	−5.08	−8.07, −2.09	0.0009
Co	1.49	−0.64, 3.62	0.1702
Motility (%) d
As	0.14	−0.16, 0.44	0.3613
Cr	0.55	−0.16, 1.26	0.1307
Pb	−0.79	−1.95, 0.36	0.1789
Mo	0.07	−0.3, 0.44	0.7161
Tl	2.10	−2.08, 6.27	0.3238
Sn	0.49	−0.34, 1.31	0.2446
W	−0.20	−2.29, 1.9	0.8523
U	24.21	−27.61, 76.02	0.3586
Sperm Concentration (*106/mL) e
Sn	0.74	−0.24, 1.72	0.1407
Semen Volume (mL) f
Cr	0.15	−0.07, 0.37	0.2017
Cr	0.04	−0.13, 0.21	0.6473
U	0.14	−0.002, 0.024	0.1102
SCSA HDS (%) g
Sb	−2.97	−12.2, 6.26	0.5275
Cd	−2.16	−8.70, 4.38	0.5164
Co	−0.17	−1.46, 1.12	0.7936
Cu	−0.91	−2.44, 0.62	0.2433

NOTE: Next-day semen analysis was utilized. Separate models were run for each semen quality endpoint. Significant findings are in boldface.

^aModel adjusted for abstinence time, body mass index, education.

^bTraditional morphology standards (WHO 3rd edition), model adjusted for serum cotinine.

^cModel adjusted for age, abstinence time, income.

^dModel adjusted for race/ ethnicity, study site.

^eModel adjusted for abstinence time, education.

^fModel adjusted for age, body mass index, education, income.

^gModel adjusted for abstinence time, alcohol consumption, body mass index, income, serum cotinine, study site; CI- Confidence Interval; SCSA- sperm chromatin structure assay; DFI- DNA fragmentation index; HDS- High DNA stainability.

In the comparative analysis using single-exposure multivariable analyses (Table 6), four metal(loid)s (Co, Cu, Sn, U) were significantly associated (p < 0.05) with semen quality endpoints in models fully adjusted for all covariates of interest, of which all were selected (β ≠ 0) in the LASSO regression model for at least one semen endpoint. The associations for all metal(loid)s in these models were positive (suggesting benefit). After applying p-value corrections to adjust for the number of comparisons, associations between single-metal(loid)s and semen quality did not remain statistically significant at the 0.05 level.

Table 6.

Adjusted Single-Exposure Associations between Urinary Non-Essential Metal Concentrations and Semen Endpoints, (n = 356).

	Total Sperm Count (*106)		Semen Volume (mL)		Sperm Concentration (*106/mL)		Motility (%)		Traditional Morphologya (%)		DFI (%)		HDS (%)
	Beta	95 % CI	Beta	95 % CI	Beta	95 % CI	Beta	95 % CI	Beta	95 % CI	Beta	95 % CI	Beta	95 % CI
Cr	1.26	−0.39, 2.91	0.15	−0.15, 0.44	0.39	−0.73, 1.51	0.56	−0.29, 1.42	2.33	−3.17, 7.84	−0.20	−4.23, 3.83	0.59	−1.5, 2.68
Co	0.81	0.03, 1.58	0.10	−0.04, 0.24	0.21	−0.32, 0.74	0.28	−0.13, 0.69	0.54	−2.07, 3.15	0.00	−1.90, 1.90	−0.65	−1.63, 0.33
Cu	0.07	−0.96, 1.1	−0.03	−0.21, 0.15	−0.07	−0.77, 0.63	0.63	0.08, 1.18	1.88	−1.56, 5.32	−2.15	−4.64, 0.34	−1.58	−2.87, −0.29
Mo	−0.03	−0.63, 0.58	0.03	−0.08, 0.14	−0.13	−0.54, 0.28	0.21	−0.12, 0.54	−0.30	−2.35, 1.74	0.73	−0.75, 2.20	−0.04	−0.81, 0.72
Se	0.59	−0.17, 1.35	0.09	−0.05, 0.23	0.12	−0.40, 0.64	0.01	−0.39, 0.42	0.18	−2.37, 2.73	−0.16	−2.01, 1.69	−0.07	−1.03, 0.89
Zn	0.04	−0.56, 0.63	−0.05	−0.16, 0.05	0.07	−0.33, 0.47	0.27	−0.03, 0.58	−0.03	−1.99, 1.94	−0.05	−1.5, 1.4	−0.47	−1.22, 0.28
Sb	−2.28	−6.94, 2.38	0.09	−0.75, 0.93	−2.24	−5.4, 0.91	1.49	−0.93, 3.92	4.31	−11.64, 20.26	−8.14	−19.45, 3.17	−3.19	−9.06, 2.68
As	−0.08	−0.65, 0.49	−0.05	−0.16, 0.05	0.09	−0.29, 0.48	0.23	−0.07, 0.53	−0.08	−1.96, 1.8	0.12	−1.26, 1.50	−0.27	−0.99, 0.44
Ba	0.01	−0.77, 0.79	0.10	−0.04, 0.24	−0.35	−0.88, 0.18	0.12	−0.3, 0.53	1.95	−0.58, 4.48	−0.98	−2.89, 0.93	−0.97	−1.96, 0.01
Cd	0.03	−0.75, 0.81	0.02	−0.12, 0.16	−0.15	−0.67, 0.38	0.30	−0.12, 0.71	0.05	−2.52, 2.61	0.81	−1.08, 2.71	−0.51	−1.49, 0.47
Pb	0.09	−1.75, 1.93	0.07	−0.26, 0.40	−0.46	−1.71, 0.78	0.51	−0.46, 1.47	2.97	−3.04, 8.98	−3.29	−7.75, 1.18	−1.98	−4.3, 0.33
Tl	1.65	−3.93, 7.23	0.03	−0.97, 1.03	1.15	−2.63, 4.93	2.69	−0.22, 5.6	12.63	−7.09, 32.34	−3.81	−17.40, 9.79	1.10	−5.95, 8.15
Sn	1.30	−0.03, 2.64	−0.08	−0.32, 0.16	1.10	0.20, 2.00	0.78	0.06, 1.49	5.30	0.90, 9.70	−1.77	−5.01, 1.48	−1.12	−2.8, 0.56
W	1.38	−3.09, 5.85	0.07	−0.74, 0.87	0.35	−2.68, 3.38	0.29	−2.09, 2.68	1.03	−13.53, 15.58	−2.61	−13.50, 8.28	−0.62	−6.26, 5.03
U	−0.91	−4.98, 3.16	−0.45	−1.18, 0.28	−0.21	−2.98, 2.55	0.88	−1.25, 3.01	6.09	−7.39, 19.58	−9.97	−19.81, −0.14	−1.74	−6.87, 3.39

NOTE: Next day semen analysis was utilized. All models adjusted for age (continuous), abstinence time (continuous), alcohol consumption (categorical), body mass index (continuous), log-transformed urinary creatinine (continuous), education (categorical), income (categorical), previously fathered pregnancy (continuous), serum cotinine (continuous), study site (categorical), and race/ethnicity (categorical). Separate models were run for each metal(loid) and semen endpoints. Significant findings are in boldface.

^aTraditional morphology standards (WHO 3rd edition).

4. Discussion

Among this sample of men recruited from the general population, evidence of association between mixtures of environmentally relevant metal(loid) concentrations and semen quality were observed. Beneficial associations were observed between urinary concentrations of Cr and Sn and semen quality, as measured by higher sperm count, higher percent normal morphology, and lower DNA fragmentation. Harmful associations were observed between Cu and Cd and semen quality, as measured by lower sperm count and higher DNA fragmentation.

The harmful associations observed for Cd and Cu in this study are consistent with associations reported in other studies. Animal studies have demonstrated that the testis is highly sensitive to Cd toxicity [70, 71], possibly through disruption of the blood-testis barrier (BTB) or by affecting the cell adhesion properties of epithelial cadherins. In the testis, to facilitate spermatocyte transit, epithelial cadherins are in front of the tight junction, a structuring that is slightly different than other blood barriers, leaving the BTB uniquely susceptible to the adhesion disrupting effects of Cd [70]. In human studies, Cd is negatively correlated with semen quality when measured in serum [16,25], seminal plasma [18,72–74], and urine [28]; and is associated with declines in sperm motility, concentration, count [16,18,23,25], and normal morphological forms [16,27,75].

Cu has long been associated with male infertility [76]. In animal studies, Cu overload is associated with increased lipid peroxidation in the testis, antioxidant and hormone imbalances, and altered testicular cell number and function [77]. In humans, it has been negatively correlated with semen quality when measured in seminal plasma [78] and urine [28]. These effects may result from cellular level disruptions in ATP production and alterations in mitochondrial membrane potential that can occur in the presence of cuprous ions [41].

Both Cu and Cr (in the trivalent state, Cr³⁺) are categorized as trace essential metals and have a role in the metabolism of cholesterol, fat, and glucose [41,79]. Cr³⁺ is nontoxic but can be converted to the toxic hexavalent state (Cr⁶⁺), which is a strong oxidizing agent [80]. Animal research evaluating Cr effects on semen quality has documented spermatotoxic effects of the Cr⁶⁺ form [81,82], whereas little is known about the biological effect of Cr³⁺. Studies, while conflicting, have found that Cr³⁺ may affect protective antioxidant status [83,84], which has a well-established impact on semen quality [85].

Like Cr, Sn has long been thought to have essential roles in the human immune, nervous and endocrine systems [86–89], but the role or effects of Sn on the reproductive system are largely unknown and few human studies have examined the association between Sn and semen quality. Guzikowski et al. [90] found a correlation between increased Sn concentrations in seminal plasma and reduced sperm count (r = 0.32) in the semen of men with limited fertility potential. Wang et al. [26] found a significant positive association between Sn seminal plasma concentrations and the percentage of necrotic sperm in men from subfertile couples.

There is evidence that the presence of Sn leads to decreases in Cu [91], which is a male reproductive toxicant known to promote oxidative stress. Although the nutritional importance of Sn is uncertain, there appears to be some beneficial bioactivity in nutritional or supra-nutritional amounts. Dietary deficiency of Sn in animals has been reported to depress growth, alter the mineral composition of several organs, and cause hair loss [89]. Cr deficiencies in the diet can produce decreased sperm counts and reduced fertility [79]; however, Cr deficiency is not common [92], and supplementation of Cr is not generally advised.

The beneficial associations observed between Sn and Cr and semen parameter endpoints in this study are similar to the results of studies of the association between semen quality and the essential trace metal Zn, which has been observed to have protective and null associations with semen quality in humans [21,22,25,26,28]. It is unclear whether the properties of Sn are similar to other essential nutrients such as phytonutrients and omega-3 fatty acids [93], and there is uncertainty surrounding the essentiality, function, and mode of action of Cr [80,92]. In order to determine whether any of the effects seen in this study reflect essential functions, a mechanism of action must be identified to explain the sperm-enhancing effects of Sn and Cr. It is also important to identify the dietary intake levels that provide optimal response.

Our findings need to be interpreted carefully given the large body of evidence indicating negative or null associations with semen quality for several metal(loid)s previously evaluated. Aside from Cd and Cu, we observed no harmful associations between metal(loid)s and semen quality, possibly given the relatively low concentrations found in men. This contrasts with other studies where male participants had higher concentrations of metal(loid)s and reported changes in semen quality [16,18–27]. The absence of negative findings in this study for some metal(loid)s with known toxicity could be due to differences in the study methods compared to prior studies. Most studies assessed metal(loid)s individually (vs. in mixture), evaluated men with known or suspected fertility impairments, and controlled for few to no potential confounding factors. Importantly, some studies have found that metal(loid)s may have different biological effects as a mixture relative to as an individual compound [83]. However, our null findings for As and Pb are comparable to other studies that found no association between sperm parameters and As [22,27] and Pb [22,27,28,44,74,94].

The results of this study are not without limitations. Our analyses controlled for a wide array of factors that may confound the association between metal(loid)s and semen quality. However, use of LASSO regression in the first step of our mixture analysis precluded our ability to “force” confounders into the model (e.g., serum cotinine where smoking is known to be a confounder). Residual confounding remains a consideration, including factors associated with other aspects of lifestyle or metabolic status. Of note, metal(loid)s have potential to interfere with metabolism and metabolic status (e.g., central adiposity, hypertension, reduced HDL, etc.) [95,96] that can directly influence sperm production and function [97] however metabolic impacts could not be taken into account in these analyses. Purposeful mechanistic research focusing on metal(loid) mixtures may help inform our findings.

Other limitations prompting cautious interpretation of our findings include multiple comparisons for the many single-metal(loid) models (i. e., one model for each combination of metals and semen endpoints) despite application of Bonferroni correction. Still, Bonferroni is the most conservative of corrections, and its application resulted in a very small α, which greatly reduced the power of each test and therefore decreased the likelihood of making any true discoveries (i.e., making a Type II error).

Our analysis also relied on next-day semen analysis due to at-home data collection for this population-based study and motility findings should be interpreted with caution because they are not directly comparable to same-day clinical semen analysis. However, only motility is affected by next-day analysis, and several other studies have validly assessed semen quality using remote semen collection methods [98,99].

With such limitations recognized, this study included numerous methodological strengths. LASSO techniques foster the modelling of health outcomes more in keeping with the nature in which humans are exposed to a multitude of environmental agents [100,101]. It is worth noting that this work specifically aimed to pinpoint the health endpoints associated with individual chemicals within a mixture. Men are ubiquitously exposed to toxicants in their ambient environment. In an attempt to identify the individual exposures that are most predictive of semen quality endpoints among the multitude, a LASSO regression approach was used for simultaneous variable selection and effects shrinkage [55,102]. This approach accounts for uncertainty related to variable selection by shrinking some coefficients toward zero. The two-step process of combining LASSO and multivariate regression is a promising approach to quantify health impacts of multiple environmental chemicals because it provides a more detailed and realistic picture of the association between metal(loid)s and semen quality than single-exposure models. Additional research is needed to determine interactions between metal(loid)s within a mixture, consistent with typical human exposure, and identify sperm effects resulting from cumulative metal(loid) exposures [100].

A major distinction between these findings and previous analyses is the use of a population-based sample, as earlier work has largely relied upon occupational, clinical or uniquely exposed populations. Few have utilized as extensive an array of metal(loid)s and confounders as considered in this work or as complete an assessment of semen quality (i.e., DNA). The majority (88 %) of men in this study were categorized as having semen quality within clinical norms [69,103], which is not surprising given that they were not recruited from clinic settings. If metal(loid) exposure impairs semen quality, prompting men to seek care, then clinical sampling may result in men with higher exposures relative to men from the general population. Most human exposure to metal(loid)s like Cd, Pb, Cu, and Sn results from anthropogenic activities, such as mining and smelting operations, industrial production and use, combustion of fossil fuels and wastes, and domestic and agricultural use of metal(loid)s and metal(loid)-containing compounds [104,105]. People in the general population living near these activities or industries and in areas where metal(loid)s are naturally occurring may be exposed to higher-than-average levels. Importantly, the metal (loid) concentrations in male LIFE Study participants are comparable to those of male participants in NHANES, which is representative of the U.S. general population [53].

We observed mixed associations between metal(loid)s at environmentally-relevant concentrations and semen quality. Whether Sn and Cr, particularly Cr³⁺, have positive effects on male fecundity as measured by semen quality is an intriguing question awaiting corroboration. Given the paucity of data on this topic relative to concerns about declining semen quality, further targeted investigations into these common environmental contaminants will help answer lingering data gaps.

5. Study conclusions

In this cohort of U.S. population-based men, there was evidence of both harmful and beneficial associations between specific metal(loid)s and semen quality. Additional research is needed to determine specific interactions between metal(loid)s within a mixture, consistent with typical human exposure, and to identify sperm effects resulting from cumulative metal(loid) exposures.

Abbreviations:

BMI

body mass index

CI

confidence intervals

DFI

DNA fragmentation

GM

geometric mean

HDS

high DNA stainability

SCSA

sperm chromatin structure assay

USD

US dollar