Fathead minnow exposed to environmentally relevant concentrations of metformin for one life cycle show no adverse effects

Peristaltic pumps were used to deliver metformin to tanks of fathead minnow in a flow-through exposure lasting 177 d. Exposure was to metformin-HCl (metformin hydrochloride C₄H₁₁N₅–HCl, MW165.62, Pharmaceutical Secondary Standard Certified Reference Material). Flows to each 12 L tank were 24 L per day. Full details of the exposure methods are in Supplementary Material 1.

Metformin was measured every two weeks in each of the 20 individual exposure aquaria. Samples were filtered and an aliquot of 180 μL was diluted with 20 μL of either a diluted labelled solution of dimethyl D₆ metformin hydrochloride or ¹⁵N₄ guanylurea hydrochloride. The resulting sample was directly injected onto a Waters Xevo TQS tandem quadrupole mass spectrometer coupled to an Acquity ultra high performance liquid chromatograph. The limit of detection (LOD) for metformin was 0.0088 μg/L. The limit of quantification (LOQ) for metformin was 0.029 μg/L. The LOD and LOQ for guanylurea, the main metformin breakdown product, were 1.3 μg/L and 5.4 μg/L, respectively. Details of the metformin analysis are found in Supplementary Material 1 and Supplementary Material 2 Figs. S1–S4.

All fathead minnow exposures to metformin and sampling/euthanasia methods were approved under animal use protocol #1848 reviewed by the Department of Fisheries and Oceans/Environment Canada Joint Animal Care Committee for the Canada Centre for Inland Waters (Burlington, Ontario, Canada), This committee is operated under Canadian Council of Animal Care approval.

Detailed methods for fathead minnow exposures are found in Supplementary Material 1. Briefly, each tank contained 30 fathead minnow eggs <12 h post-fertilization (sources Aquatox, Aberfoyle, Ontario, Canada and Ontario Ministry of the Environment Lab, Etobicoke, Ontario, Canada), 10 in each of 3 egg cups. Exposure temperature was controlled at 25 °C by water bath surrounding aquaria. Photoperiod was 16 h light and 8 h dark with dawn and dusk dimming. Dilution water for the metformin was fish lab water, which was municipal water for the City of Burlington, Ontario, Canada, that was dechlorinated by charcoal filtration, followed by filtration for particulates, and UV sterilization. Water temperature, dissolved oxygen, pH, conductivity, and ammonia were measured once a week in all fish exposure aquaria. Larval fish were fed 2× per day newly hatched brine shrimp. As fish grew, diet was switched to 2-d-old hatched brine shrimp, with additional frozen and thawed brine shrimp added as fish grew (as described in Supplementary Material 2 Table S3).

Random culling of fish over time maintained appropriate densities for remaining fish to grow. At 78 dph, three breeding tiles were added to each tank to promote maturation of males and breeding/spawning behaviours. Eggs were counted and assessed for stage of development, deformities, and fertilization. Breeding was monitored until 168 dph. Up to 100 fertilized eggs from each batch were placed in screen-bottomed egg cups in separated flow-through aquaria containing clean lab water (aerated) and observed until hatch, or until 9 and 16 dph. For the entire experiment over 192 700 eggs were counted and observed (31 780 to 46 495 eggs from metformin treatments, which had 4 replicates, and 75 666 eggs from the control treatment, which had eight replicates).

At 169–172 dph, all of the adult fish from the controls and metformin-exposure aquaria were sampled and dissected. Methods for dissection are described in Supplementary Material 1. Concentrations of VTG in individual fish plasma samples were determined using a TECO Cyprinid Vitellogenin ELISA kit. Gonad sampling and processing methods were based on published methods (Ankley 2006; Tetreault et al. 2011). Methods for scoring and counting of gonads were based on Ankley (2006). Female tissues were imaged at 50× magnification. Oocytes were categorized as primary, cortical alveolar, early vitellogenic, late vitellogenic, or atretic and the number of each counted. Male tissues were scanned at 200× magnification for the presence of testis-ova (male testicular tissue with female oocytes present) and each sample staged as follows: stage 0—underdeveloped, stage 1—early spermatogenic, stage 2—mid spermatogenic, stage 3—late spermatogenic. On average about 20–25 slides were assessed per male fish. Details of the sampling, preservation and histological scoring of gonads are found in Supplementary Material 1.

Data were analyzed using Systat 11.0 (Systat Software Inc., San José, California, USA). Data were checked for normality by sex and by treatment (control, 3.0, 31, 322 μg/L metformin) for growth parameters, and by treatment for egg production and reproductive parameters (i.e., egg viability and hatching success) using the Shapiro–Wilk normality test. In about 90% of the cases, (i.e., 86% of the survival data, 92% of the growth data, 90% of the mature fish sampling data, 92% of the data on egg production and 95% of the data on egg quality, 91% of the data on larval survival and growth, 81% of the female oocyte stage data) the data were distributed normally, so analysis of variance (ANOVA) was performed using untransformed data. Mean VTG concentrations in plasma and mucous of adult fish were not normally distributed (33% of data were normal), so log VTG concentrations were used for statistics (once logged, the VTG data were 87.5% normal).

Growth parameters of length (mm), weight (g), condition factor (CF), liver-somatic index (LSI), gonadosomatic index (GSI), ovipositor area of females (mm²), male index, tubercle index, and log VTG were assessed for each sex for differences among treatments using ANOVA. This statistical analysis approach was also used to assess F1 parameters (i.e., % hatch, % larval deformities, % larval survival, larval lengths and weights at 9 and 16 dph) for differences among treatments. Significant differences from control treatments were assessed using two-sample t tests (separate variances, Bonferroni’s adjusted probabilities) to determine levels of significance. All p values for these tests are shown in the text, figures, and tables. Where ANOVA p values are used, they are indicated in the text.

Raw data for testes histological scores, when assessed for normality using Shapiro–Wilk normality test, were not normal. Individual tank mean scores were 0% normal (0 of 20 tank mean testes scores were normal), and treatment means were only 50% normal (2 of 4 treatment means were normal). In these cases, nonparametric statistics were used to compare ranked mean values. Kruskal–Wallis one-way ANOVA on ranked treatment means was used to assess treatments, followed by Kolmogorov–Smirnov two-sample test of two-sided probabilities which is a multiple-comparison-protected test of ranked treatment means to control means. In these cases the p values reported in the text were from Kolmogorov–Smirnov two-sample test (of two-sided probabilities when ranked treatment means were compared with ranked control means).

For the assessment of eggs and fry produced by the parent fish exposed to metformin, the data were summarized in two ways. Percentages of unfertilized eggs and dead eggs as well as % normal eggs (healthy eggs that were assumed viable) and % bad eggs were assessed in each batch of eggs and averaged per tank. These fertilization and egg quality data were also assessed by summing the incidence of these in all eggs laid in a given tank and dividing by the total number of eggs assessed for that tank. These values provided another way for the reproductive parameters to be analyzed per replicate tank and gave an overall assessment of the success of reproduction and hatch for that treatment.

Water quality parameters were stable during the flow-through exposures. Mean temperatures ranged from 24.9 to 25.1 °C, mean dissolved oxygen ranged from 7.8 to 7.9 mg/L, mean pH ranged from 8.15 to 8.23, mean conductivity ranged from 334 to 336 μS/cm, and mean free ammonia ranged from 0.0 to 2.5 μg/L. None of the water quality values differed between the controls and any of the three metformin exposure concentrations. Mean (± standard deviation) dilution water hardness was 129 ± 5 mg/L, and mean total alkalinity was 91 ± 4 mg CaCO₃/L. All water quality data are summarized in Supplementary Material 2 Tables S4 and S5.

Measured concentrations of metformin in fish exposure aquaria were about 100%–107% of nominal concentrations (mean = 103.5% ± standard deviation 3.5%). Table 1 summarizes the mean concentrations of metformin in each exposure aquarium, and the overall means for each treatment. Target nominal concentrations for the metformin were based on concentrations of metformin in rivers, in MWWE, and 10× MWWE. All metformin measurements are shown in Supplementary Material 2 Table S6. No guanylurea (the major breakdown product of metformin) was detected in the exposure aquaria. The detection limit for guanylurea was 1.3 μg/L, and all exposure water samples were analyzed for guanylurea. There was never any guanylurea detected in the exposure solutions.

During the life cycle exposure to metformin, a leak in the header tank (containing clean lab water, no toxicant) dripped down from a dusty top shelf/surface to contaminate two flow-through tanks. Two aquaria were affected before the leak was discovered and fixed (see Supplementary Material 2 Fig. S5 for tank set up and position of tanks affected by the leak). In the two affected tanks some larval and juvenile fish died, and so tanks 9 (3.0 μg/L metformin) and 19 (control tank) were removed from the survival-over-time data. (The reader is referred to Supplementary Material 2 Table S7a for a statistical assessment of survival over time with tanks 9 and 19 included). The remaining fish in these two affected tanks appeared to grow and breed normally, and so these two tanks were included in the growth data and in the reproductive phase of the life cycle study.

In this life cycle experiment, two sources of fathead minnow eggs were mixed. This was done to ensure genetic diversity among the minnows, and to reduce the possibility of siblings mating during the reproductive phase of the life cycle test. Extra eggs from each source were separately grown in clean water in another flow-through system until adulthood. After the life cycle metformin exposure, a skewed sex ratio was observed in the eggs sourced from one supplier (kept in clean water for growing out). The sex ratio was 72% male in the eggs sourced from the Ontario Ministry of the Environment and Energy. Only 4 breeding tiles from this supplier were used in the life cycle test (1/3 of the eggs for the life cycle exposure to metformin were sourced from this supplier), and so 4 sets of fathead minnow parents (i.e., 4 breeding tiles) produced these eggs. In the eggs sourced from the usual supplier, Aquatox (14 breeding tiles, which made up 2/3 of the eggs used in the life cycle test), the sex ratio was normal, i.e., 50% male. Because some of the eggs (4 of 14 tiles) were sourced from a 72% male biased colony, there was a skewed-towards-male sex ratio in the life cycle exposure. The calculated expected sex ratio as 56% male, 44% female in the overall study, which likely affected some of the maturation and breeding data, and there were fewer females than desired in some of the replicate tanks.

Life cycle exposures to the metformin did not affect survival of fish. There were no effects on egg mortality or fry hatching at the beginning of the life cycle exposures. Fathead minnow eggs exposed to metformin successfully hatched, and larval survival was not statistically different from the control treatment, as illustrated in Fig. 2 (see Supplementary Material 2 Table S7b). Overall, control fish survival was 89% from the fertilized egg stage until 172 dph, while fish exposed to metformin had an overall survival of 84%, 92%, and 90% for treatments 3.0, 31, and 322 μg/L (measured concentrations), respectively. Decreases in survival (of 8%) at 106 dph and afterwards for metformin 3.0 μg/L (light green triangles) in Fig. 2 were not significant (p ≥ 0.324) compared with control survival (blue open circles). When survival over time data were also assessed with the two accidentally contaminated tanks included (tank 9 which was a 3 μg metformin/L tank, and tank 19 which was a control tank), there were no differences in survival (ANOVA p > 0.142). Supplementary Material 2 Table S7a provides a statistical assessment and re-calculated means for survival over time with tanks 9 and 19 included. Whether these tanks were included or not made no differences to the conclusions: There were no significant differences in survival over time between the controls and any metformin treatment.

There were few effects of the metformin on growth of the fish over time. Mean weight (mg), length (mm), and CF of metformin-exposed fish that were sampled at 30, 64 and 78 dph did not differ from those of controls, as shown in Table 2 (see Supplementary Material 2 Table S8 for individual replicate weight and length details). The only significant differences seen were at 30 dph where fish exposed to metformin (31 μg/L) had significantly greater lengths (21.4 mm, p = 0.038) compared to control fish length (19.6 mm), and at 64 dph where metformin-exposed females (322 μg/L) had significantly higher condition factors (CF = 0.906, p = 0.033) compared to control fish condition factors (CF = 0.864).

There were no effects of exposure to the metformin on maturation of the fish during the life cycle test (Supplementary Material 2 Table S9) (Fig. 3). From their external characteristics, only a few fish were mature by 64 dph, with 2% of fish identified as either males or females at this time. At 78 dph, 13% of the fish were mature and breeding tiles were added to the exposure tanks to promote maturation. By 92 dph, 50% of fish were mature (and could be classified externally as males or females), and by 106 dph, 78% of the fish were mature. There were no differences in the % mature fish per tank in the metformin treatments versus the controls (p ≥ 0.226). The one exception was at 92 dph, when 3.0 μg/L metformin-exposed fish (light green triangle) were significantly more mature compared to control fish (blue open circles) (see asterisk, p = 0.005, Fig. 3).

At the end of the exposure period, when fish were sampled at 169–172 dph, there were no differences in growth of the mature males. Weight, length, and condition factors were not statistically different between control males and males from all metformin treatments (Table 3). Internal organ sizes relative to body weight (as represented by LSI and GSI) were also not statistically different between control males and males from all metformin treatments.

There were trends of decreased weight of mature female fish exposed to metformin. Adult females sampled at the end of the exposure (169–172 dph) mean weights were: controls (1.81 g), 3.0 μg/L metformin (1.74 g), 31 μg/L metformin (1.71 g), and 322 μg/L metformin (1.66 g) (Table 3). The highest metformin exposure concentration had females that weighed 8% less than controls, but these differences in body weight of females were not statistically-significant (p = 0.144). Although the females from 322 μg/L metformin had body weights that were 8% smaller than controls, their relative gonad (ovary) sizes were significantly larger than controls (Table 3, Fig. 4).

Females exposed to metformin had 23% increased ovary weights (relative to their body weights). The highest metformin treatment (322 μg/L) had females with significantly larger relative ovary size (16.5% body weight was ovary) compared to control females (with 13.4% body weight as ovary, p = 0.041, Table 3). Raw data for each replicate tank are presented in Supplementary Material 2 Tables S10 and S11. Raw data for each female fish are in Supplementary Material 3.

Mature male sex characteristics were unaffected by metformin, with the exception of males from the 31 μg/L metformin exposure concentration, that had 28% enhanced male sex characteristics (male index = 8.2) compared to control fish (control male index = 6.4, p = 0.045). Males from the highest metformin concentration (322 μg/L) had male indexes that were 11% elevated as well, but these were not significantly different from controls (p = 0.097) (Table 3). See Supplementary Material 2 Tables S10 and S11 for all data by replicate tank. Raw data for each male fish are in Supplementary Material 3. Female fathead minnows exposed to the highest metformin concentration (322 μg/L) had 16% larger ovipositors (p = 0.023) compared with control fish. Mean ovipositor areas of female control fish and metformin-exposed female fish ranged from 1.13 to 1.48 mm² (Table 3).

Some metformin-exposed fish began breeding later than controls. Breeding began in the control treatments at 89 dph (this was when the first of eight replicate tanks in the controls produced a clutch of 20 or more eggs). The mean time to first breeding was 97 dph for controls. (This was the average day post-hatch when breeding first occurred for the 8 replicate tanks.) In the metformin treatments, the first breeding began at 94, 104, and 95 dph for fish from the 3.0, 31, and 322 μg/L metformin treatments, respectively. Mean time to first breeding for the metformin treatments was 102, 107, and 106 dph for fish from the 3.0, 31, and 322 μg/L metformin treatments, respectively (Table 4). All egg production data for each replicate tank are shown in Supplementary Material 4. Based on other fathead minnow life cycle tests run previously in the same laboratory (Parrott and Bennie 2009; Parrott and Balakrishnan 2017; Parrott and Metcalfe 2017), the time-to-first-breeding was within the normal range for fish from the control treatments. There were 9–10 d delays in mean time to first breeding in the top two metformin exposure concentrations, but these were significant (p = 0.003) only for the 31 μg/L metformin-exposed fish (and not for the 322 μg/L metformin exposure concentration, p = 0.208) compared with controls.

Fish from all metformin treatments produced similar numbers of eggs in relation to the controls. Cumulative egg production over time was very similar for the controls and metformin treatments, when plotted per female fish (Fig. 5). See Supplementary Material 2 Table S12 for all breeding data per replicate tank).

When individual tanks were assessed, the variability in egg production was apparent. This was in part due to some tanks with fewer females. There was a trend of lower mean total egg production per tank and lower mean eggs per female in the 322 μg/L metformin treatment (Fig. 6) (Supplementary Material 2 Table S12), although these differences were not significantly different from fish in the control (i.e., p = 0.584, and p = 0.741, respectively for total eggs and eggs per female for Bonferroni’s adjusted p using separate variances). Total egg production per 4 replicates was 37 833 for controls and 46 495, 38 834, and 31 780 for fish exposed to the 3.0, 31, and 322 μg/L metformin, respectively. There were no statistically significant differences in mean egg production per female across all treatments, with 2485 for control fish, 2826, 2141, and 2021 for fish exposed to the 3.0, 31, and 322 μg/L metformin treatments. Mean clutch sizes were not statistically different (p ≥ 0.302) and ranged from 208 to 215 eggs per clutch. Mean number of clutches per treatment were also not statistically different between controls and the three metformin exposure concentrations (ANOVA p = 0.987, Table 5).

There were no negative effects in F1 offspring egg quality produced by parent fish exposed to metformin for a life cycle. The percent unfertilized eggs ranged from 4% to 8% and did not differ between controls and metformin treatments (Table 5). The percent of normal eggs increased in the low metformin exposure, with 3.0 μg/L metformin producing higher % of normal eggs (76.7%, p = 0.008) compared with controls (normal eggs = 69.7%). Two metformin treatments had lower % of dead fertilized eggs compared with controls: Mean number of dead eggs (calculated as the mean of every egg batch) and % dead eggs per tank (calculated using tank totals) were significantly reduced in offspring from parent fish exposed to 3.0 and 322 μg/L metformin (p = 0.012–0.034, see Table 5, bold values with asterisk). All data per replicate tank are shown in Supplementary Material 2 Table S13.

There were no differences in the hatch success or deformities of F1 offspring. Embryo hatching success and fry deformity rates from parent fish exposed to metformin were not statistically different from those of controls (Table 6). The % hatch and % fry deformities all showed no differences across treatments (see Supplementary Material 2 Table S14 for details by replicate tank of all parameters assessed in eggs and hatched fry) (Table 6).

All fertilized and healthy eggs (or a subset of 100 eggs from egg batches over 100) from each egg batch were placed in egg cups to assess hatching success and time to hatch. Hatching was monitored for a total of 17 700 eggs from the 8 replicate control aquaria, and 10 690, 9130, and 7560 eggs from the 3.0, 31, and 322 μg/L metformin treatments, respectively (having 4 replicate aquaria in each). For eggs produced by parent fish exposed to metformin for a life cycle, the time to hatch of F1 eggs was 5.1–5.2 d, not statistically different from control fish (p ≥ 0.812) (Table 6).

There were no significant differences in hatchability (i.e., % embryos that hatched) or hatching success (i.e., % of embryos that hatched into viable fry) (p ≥ 0.199). Hatching success of eggs was high and ranged from 92% to 94%. Percentages of fry with deformities were also not statistically different across the controls and metformin treatments (Table 6). Percentages of newly hatched fry with severe deformities was about 2%–4%, and mild to moderate deformities about 3%–4% (Table 6). The percentages of deformed fry at hatch (both severe deformities, and those with mild to moderate deformities) was not statistically different between controls and all metformin treatments (p ≥ 0.644). See Supplementary Material 2 Table S14 for data from all replicate tanks.

There were no changes in survival or growth of F1 larvae produced by parent fish exposed to metformin. Survival was very good (above 80%, Fig. 7) from the egg stage until 16 dph for all F1, with no significant differences among controls and metformin treatments (p > 0.844). As shown in Table 7, growth at 9 and 16 dph was not statistically different among controls and all metformin treatments (0.066 < p > 0.935). All data for growth of F1 in metformin replicate aquaria are shown in Supplementary Material 2 Table S15).

Plasma VTG in adult male fathead minnows was not affected by metformin exposure for a life cycle. Mean plasma VTG in male fish ranged from 0.77 to 1.9 μg/mL (Fig. 8) and there were no significant differences among metformin treatments and controls (statistics done on log VTG plasma, p ≥ 0.469). Full data for VTG in plasma of male fish are found in Supplementary Material 2 Table S16.

Life cycle exposure to metformin did not affect VTG in female fathead minnow blood plasma. Mean plasma VTG in female fish ranged from 12 500 to 13 300 μg/mL (see Supplementary Material 2 Table S16) which was over 6000 times the VTG in male plasma, as expected. There were no significant differences among metformin treatments and controls (statistics done on VTG in plasma, ANOVA p = 0.964). Full data for VTG in plasma of female fish are found in Supplementary Material 2 Table S16.

Assessment of gonads of male and female fish showed no differences in metformin-exposed fish and controls. Male testes were scored from 0 to 3, with 3 being fully mature male gonads. Mean ± standard deviation (SD) maturity stages for male gonads were 2.3 ± 0.3, 2.4 ± 0.2, 2.6 ± 0.2 for 3.0, 31, and 322 μg/L metformin, respectively, which were not statistically different from the maturation stages of controls (2.5 ± 0.3, Kolmogorov–Smirnov p = 0.875 for all three pairwise comparisons). On average 22 ± 5 slides were assessed per male fish (min 11, max 34 slides per fish). Out of over 1400 slides scanned, representing 69 males, no oocytes were observed in any male testes examined. All data for male testes maturity scoring by replicate tank are found in Supplementary Material 2 Table S17a.

Female gonads were scored by counts and proportions of oocytes at various stages of maturation. For each fish an average of 288 ± 119 oocytes were counted (min 50, max 838). There were no differences from controls and the three metformin treatments in any parameter monitored (all ANOVA p > 0.533) for percentages of primary, cortical alveolar, early vitellogenic, late vitellogenic, and atretic oocytes (Fig. 9). Scores for atretic oocyte severity had ANOVA p = 0.207, and comparison of 3.0 μg/L metformin to controls showed the differences were not significant (Bonferroni’s p = 0.131). Overall % mature oocytes (early vitellogenic + late vitellogenic) were similar among treatments with 24.1% ± 3.7%, 25.5% ± 6.3%, 27.0% ± 2.8% for 3.0, 31, and 322 μg/L metformin treatments, respectively, which were not statistically different from controls (24.7% ± 7.6%, ANOVA p = 0.906). All data for female ovary scoring by replicate tank are shown in Supplementary Material 2 Table S17b.

A summary of all statistically different changes seen in the metformin life cycle exposure is found in Supplementary Material 2 Table S18.

Concentrations of metformin were close to nominal concentrations over the 5.5-month exposure of fathead minnows. There was no observed degradation of metformin to guanylurea or nitrate/nitrite in the flow-through aquaria, which is not surprising as 24 L of fresh metformin solution flowed through each tank daily. Experimental biodegradation studies in the lab show that metformin is oxidized to guanylurea, which breaks down to nitrate (Roche 2015a, 2015b in Straub et al. (2019)). Laboratory degradation studies show that the oxidation to guanylurea is slow (occurring over days 7–21), and guanylurea breakdown to nitrate occurs over days 35–56 (Roche et al. 2015a, 2015b in Straub et al. (2019)). Metformin breakdown to guanylurea is efficient in municipal wastewater treatment plants with conventional activated sludge (Kosma et al. 2015), with degradation slower under anaerobic conditions (Tisler and Zwiener 2019). Although breakdown of metformin occurs efficiently in municipal wastewater treatment plants, once in a river system, the breakdown of metformin may be slower.

There were no significant changes in survival of fish exposed over a full life cycle to the metformin at environmentally relevant concentrations (i.e., 3.0 and 31 μg/L) or at higher concentrations (i.e., 322 μg/L metformin). Other researchers have seen similar results, with no effects on survival of Japanese medaka exposed for 133 d to metformin concentrations up to 360 μg/L (Lee et al. 2019). Rainbow trout sac fry showed no changes in survival after 108-d exposure to metformin through their juvenile stages (from 48 d post-fertilization (dpf) to 156 dpf) (Jacob et al. 2018).

Exposure to the metformin did not negatively affect the maturation of the fathead minnows over time. Juvenile fish exposed to all three metformin treatments developed sex characteristics at the same rate as control fish. Exposure to 31 μg/L metformin increased the secondary sex characteristics (i.e., male index) of adult male minnows sampled at the end of the life cycle test (Table 3). By contrast, other researchers have reported that exposure to metformin decreased (but not significantly) male secondary sex characteristics in fathead minnows (Niemuth and Klaper 2015).

Life cycle exposure to high concentrations of metformin caused trends of decreased body weight of female adult fathead minnows. There were dose-responsive trends of decreased weight of female fish sampled at the end of the life cycle test, but these were nonsignificant (Table 3). At the end of the exposures, female fish exposed to 3.0, 31, and 322 μg/L metformin weighed 96%, 94%, and 92% of control female mean weights, respectively.

The 8% decreases in weight of females observed in this study are consistent with effects observed by others. Most studies of metformin’s effects in fish report reduced fish wet weight. Fathead minnows exposed to 40 μg/L metformin for 360 d had decreased weight and condition factor compared with control males (Niemuth and Klaper 2015). Rainbow trout exposed (at 7 °C) to metformin from 48 dpf until 156 dpf had decreased wet weights at 10 and 100 μg/L, but not at 1000 μg/L, despite similar metformin uptakes into fish at 100 and 1000 μg/L (42 and 56 ng/g wet weight, respectively) (Jacob et al. 2018). Other researchers have shown low concentrations of metformin (3.2 μg/L) can decrease larval Japanese medaka growth (from the egg stage to 28 dph), and much longer exposures (egg stage to 165 dph) caused trends of lower growth, but these were not significant in the adult medaka (Ussery et al. 2018). However, some long-term exposures of metformin show no changes in fish growth: a two-generation study of Japanese medaka (lasting 360 d), exposed to 40, 120, or 360 μg/L metformin caused no changes in growth or in relative organ size (GSI and LSI) of parental fish or offspring (Lee et al. 2019).

Despite trends of 8% lower body weights of female fathead minnows exposed to high concentrations of metformin, their ovary weights were 23% larger than the control fish. Females exposed to metformin tended to weigh less, but that they had larger gonads (Fig. 4). Whether this was influenced by the sometimes uneven numbers of females per tank was examined. However, it appeared there was no relationship of female body weight to the number of females per tank (data not shown). Life cycle (19-week) exposure of female medaka to metformin caused no significant change in GSI (Lee et al. 2019). No other studies reporting metformin’s effects on female ovary size in fishes were found in the literature.

The significance of the 23% increased ovary size of female fish exposed to 322 μg/L metformin appears to be low, as there were no overall effects on egg production, egg quality, or F1 survival and growth. In addition, there were no metformin-related changes in the stages of oocytes in ovaries. Relative percentages of oocyte stages in the control and metformin-exposed females were similar to that found in control fathead minnows in other labs (Jensen et al. 2001).

Life cycle metformin exposure caused delays in breeding in fathead minnows. The 31 μg/L metformin treatment significantly delayed the start of breeding by 10 d, and the 322 μg/L delayed the start of breeding by 9 d (but this delay was statistically not significant). In previous studies with fathead minnow exposed to MWWE for a life cycle, similar delayed time to first egg production (delayed by 20–30 d) was observed, but in those studies the delays were accompanied by significantly reduced cumulative egg production in some MWWE treatments (Parker et al. 2014). In the case of the present life cycle exposure to metformin, breeding was delayed by 9–10 d, but the overall egg production per female was unaffected.

A 9–10 d delay in first breeding could be biologically significant to wild fish, especially in colder climates where minnow breeding times are brief. In Canadian northern lakes, fathead minnows grew and matured faster, and breeding time was earlier (Danylchuk and Tonn 2006) presumably so that larval fish had adequate time to reach sizes that could better survive the harsh winter conditions (Sogard 1997; Hurst 2007). Delays in spawning would shorten the time larval fish had to reach these optimal weights. Delayed spawning may also affect larval fish growth and survival due to limits on food availability. Larval minnows may not have adequate food of the correct size available (Laurich et al. 2003; Nunn et al. 2012) if they hatch later, and this would also affect their abilities to reach body sizes that would over-winter successfully (Sogard 1997; Hurst 2007).

Concentrations of VTG in plasma of male and female fathead minnows were not affected by metformin exposure for a life cycle. Mean plasma VTG in male fish ranged from 0.77 to 1.9 μg/mL and in female fish ranged from 12 500 to 13 300 μg/mL (Fig. 8). The concentrations of VTG in fathead minnows in the present metformin exposure compared well with concentrations of VTG measured in plasma of mature control fathead minnows by other researchers in several labs (Table 8).

Life cycle exposure to 3.0, 31, or 322 μg/L metformin caused no significant changes in egg production compared with control fish (Fig. 5). There were no statistically significant differences in mean egg production per female across all treatments, with 2485 for control fish, 2826, 2141, and 2021 for fish exposed to the 3.0, 31, and 322 μg/L metformin treatments. Fathead minnow breeding groups typically have variable egg production (Jensen et al. 2001; Watanabe et al. 2007). However, when the egg production per female in each tank/replicate was examined, a few tanks did produce fewer than 1500 eggs per female (Fig. 6). In the present experiment several tanks had fewer females than the desired 4–5, which may have influenced the egg production even after the lower number of females was accounted for. Figure 10 shows that there appears to be a good relationship (r² = 0.72) for control fish between number of eggs per female and the number of females per tank. Thus, the control tanks with fewer females produced lower numbers of eggs per female, even when per-female egg production was assessed. Because each tank had 8 fish, tanks with low numbers of females had higher numbers of males and (or) immature fish that turned out to be males. The relationship in Fig. 10 may be caused by the increased number of males/immature males in the tank chasing the three dominant males that occupy and defend their nests (breeding tiles) or chasing the females; male fathead minnows often chase and compete with others in the tank (McMillan and Smith 1974; Salierno and Kane 2009; Villeneuve et al. 2010). This may explain the lower number of eggs per female in two of the 322 μg/L metformin tanks (Fig. 10, orange arrows), but it does not explain the low numbers of eggs per female for the 31 μg/L metformin tank (Fig. 10, black arrow).

A previous life cycle study with fathead minnows exposed to metformin showed decreased reproduction (Niemuth and Klaper 2015), although in that study the exposure was from the juvenile (30 dph) to the adult stage, not from egg to adult as in the present study. The Niemuth and Klaper (2015) study also observed VTG induction and intersex in male minnows, and this is discussed below. Other differences between these two life cycle exposures of fathead minnows to metformin include the age of the breeding fathead minnows, which were 320–360 dph in the Niemuth and Klaper (2015) study, and 90–168 dph in the present study. Also, the previous study was static renewal (with metformin-exposure solutions renewed every 3–4 d) compared with the present study’s flow-through design (where 24 L per day flowed through each 12 L tank). The fish diets were also different in the two studies. Brine shrimp (newly hatched then frozen) was the only food provided to fish in the present study, compared to mainly Tetramin tropical fish food in the Niemuth and Klaper (2015) study. Two other fish life cycle studies have shown no effects of metformin on Japanese medaka reproduction (up to 360 μg/L for 133 d, (Lee et al. 2019), and 3.2 μg/L metformin for 165 d (Ussery 2018), despite the observation of some changes in gene expression in both of these metformin-medaka studies.

In the present life cycle exposure to metformin there was no incidence of intersex in any of the 69 mature male fish, despite over 1400 testes slides being assessed (20–25 slides per male). The diameter of primary oocytes in fathead minnows is about 60–120 μmol/L, while late vitellogenic oocytes are over 500 μmol/L (Jensen et al. 2001). So each male testis in the present study was sampled into sections of 4–5 μm thickness (with no more than 28–35 μm spacing between groups of sections) to maximize the potential to detect any oocytes present. Previous life cycle exposures to metformin have observed high incidences (84%) of males with testis-ova in 40 μg/L metformin, as well as 13% testis-ova in control males (Niemuth and Klaper 2015). It is unknown whether older control male fathead minnow, such as those assessed in the Niemuth and Klaper (2015) study, develop testis-ova. Studies of control male Japanese medaka (raised in distilled tap water) show 0% incidence of intersex at 6 months of age, but 11.1% incidence of intersex at 9 months of age (Grim et al. 2007). Another possible cause of intersex in control fish is the presence of endocrine-active substances (such as phytoestrogens) in flake-food diets (Grim et al. 2007).

Generally testis-ova condition is less common in fathead minnows than other commonly used fish test species, as demonstrated by the absence of a photo of this condition in a detailed gonadal histology atlas of fathead minnow (Ankley 2006). It would be valuable to assess whether other labs have observed intersex (testis-ova) in control fathead minnow males, and if so, whether the presence of testis-ova increases with the age of the males or may be influenced by the type of feed, as it was in Japanese medaka (Grim et al. 2007).

There were no differences in egg quality, % fertilization, or hatching success in eggs of fathead minnows exposed to metformin. This is similar to Lee et al. (2019) who saw no effect on fecundity, hatching rate, or larval abnormalities in a two-generation exposure of Japanese medaka to 40, 120, and 360 μg/L metformin (Lee et al. 2019). Metformin exposure of parent fathead minnows for a life cycle caused no changes in growth or survival of their F1 offspring up to 16 dph (when hatched in clean water flow-through conditions). There are no published literature studies of F1 survival and growth from life cycle metformin-exposed parent fish.