We selected juvenile fish to represent the most likely size range of consumption by predators. All fishes were euthanized, weighed to the nearest 0.1 g, measured to the nearest mm for both total length (TL) and standard length (SL), and then flash frozen before being placed into a -70 °C freezer until each fish could be homogenized to a whole body fine powder on dry ice. Thiaminase I activity was measured using a 4-NTP assay on homogenated samples following the methods of Hanes et al. (2007) and Kraft et al. (2014). Briefly, homogenate (0.5 g wet weight) was suspended in phosphate buffer (1.25 mL of 100 mM, pH 6.5), bead beat to lyse cells and centrifuged at 4 °C for 10 min at 16,000 x g. A 750 μL aliquot of each supernatant was transferred to 0.8 mL Pierce centrifuge columns containing 30 μm pore size polyethylene filters in new 1.5 mL centrifuge tubes and centrifuged a second time for 5 min at 5,000 x g at 4 °C to remove any particulates. Filtrates were placed on ice and immediately assayed for thiaminase activity using Tris (2-carboxyethyl) phosphine hydrochloride (TCEP, 10 mM, pH 6.9) buffer with 4-NTP (80 µM), both with and without thiamine (400 µM), in 96-well microtiter plates. Wells of the microtiter plate were filled with 97 μL of solution using a multi-channel pipette; three wells with the solution without thiamine and three with solution with additional thiamine. Enzymatic reactions were initiated by the addition of 3 μL of sample supernatant to each well. Readings of absorbance at 411 nm were taken every minute over 60 minutes total at 37 °C in a Synergy HT Multi-Detection Microplate Reader (BioTek Instruments, Winooski, VT). Activity was calculated by linear regression of the change in absorbance over time, corrected for non-thiaminase activity by the difference in slopes (with and without thiamine added), and converted to concentration using the molar extinction coefficient for 4-NTP (13,650 M^-1 cm^-1). We did not correct thiaminase activity for protein concentrations, so activity measurements are reported as nmol/min/g-sample wet weight.

All assays were performed in triplicate. Solution blanks and positive control samples (Paenibacillus thiaminolyticus supernatant) were run with each plate of samples as a QA/QC check. The positive control (P. thiaminolyticus) is a known thiaminase-producing bacterium isolated from alewife gut (Honeyfield et al., 2002). The bacterium was grown in 50 mL of terrific broth at 37 °C set to shake at 225 revolutions per minute for 24 hours or until optical density was approximately 1. Cultures were centrifuged at 5000 x g for 60 minutes at 4 °C, and then the supernatant was filtered through 10 kD molecular weight cutoff filters (Fisher Scientific, Waltham, MA, USA) rinsed with sterile phosphate-buffered saline (PBS). The original volume was concentrated to 750 µL of P. thiaminolyticus filtered supernatant and was brought up to 1 mL volume with 60% sterile glycerol stock and stored at -20 °C. Across all plates run for this study, the average activity of P. thiaminolyticus supernatant control samples was 480 nmol/g/min (± 26 standard deviation ( stdev)) and the coefficient of variation was 5%.

All statistical analyses were completed in R v4.5.1 (R Core Team, 2025). Enzyme activity data are often highly variable due to enzyme production being up- or down-regulated for specific functions. There are often non-detections in the data. We accounted for this using functions specifically designed for datasets that include censored (i.e., below detection limit) data. We assessed differences among species using censored ANOVA fit with the cenanova function in the ‘NADA2’ package (Julian and Helsel, 2024) with a log-link function and 10,000 permutations. This function performs a parametric maximum likelihood test of differences among group means including censored data, followed by a parametric Tukey’s multiple comparison test.

For testing between culture- and pond-sourced fishes, we used the cenpermanova function in the ‘NADA2’ package (Julian and Helsel, 2024). Permutation tests are more robust to small or uneven group sizes, and the cenpermanova function allows for testing differences between two group means that include censored data in a non-parametric way. We tested the two species with both indoor and outdoor/pond cultures, silver and grass carp, for differences in thiaminase activity between pond and indoor culture sources as separate species-specific models. We fit the models separately because there was little overlap in thiaminase activity of grass and silver carp. We ran all cenpermanova models with 10,000 permutations.

We assessed whether fish body condition (K, eq 1) or size (standard length in mm) would relate to thiaminase activity.

Where w = mass (g), SL = standard length (mm), and 100,000 is a conversion factor for metric units.

Because the four invasive carps are from three different genera, we fit separate models for each species to examine effects of body length or body condition on thiaminase activity. We fit Bayesian linear models fit in the ‘rstanarm’ package (Goodrich et al., 2024) exploring how body length or condition affected the ln-transformed thiaminase activity for all values above the limit of quantification. For each model, we ran four chains for 20,000 iterations and discarded the first 5,000 posterior samples as warm-up, leaving 60,000 posterior samples for analysis. We assessed fit by examining traceplots for convergence and using leave-one-out cross-validation using function ‘loo’ in the R package ‘rstan’ (Stan Development Team, 2024). All data are available in a data release (Rowland et al., 2025).

Overall, we assessed 145 individual carp for thiaminase activity (Table 1) including bighead carp from indoor culture (n = 7), black carp from indoor culture (n = 38), grass carp from both indoor culture (n = 37) and outdoor pond culture (n = 13), and silver carp from indoor culture (n = 37) and outdoor pond culture (n = 13). Censored ANOVA indicated that there were statistical differences among species (d.f. = 3, Χ²= 170.8, p < 0.0001; Fig 2). The model estimated the mean thiaminase activity was highest in grass carp (60.1 nmol/g/min), followed by silver carp (14.7 nmol/g/min), bighead carp (2.7 nmol/g/min), and lowest in black carp (2.2 nmol/g/min). Tukey contrasts with lognormal censored data indicated that grass carp had statistically higher thiaminase activity than silver carp (z = 8.626, p < 0.001), bighead carp (z = 8.430, p < 0.001), and black carp (z = 16.613, p < 0.001). Silver carp had statistically higher thiaminase activity than bighead carp (z = 4.207, p < 0.001) and black carp (z = 9.121, p < 0.001). There were no differences between bighead and black carp (z = 0.948, p = 0.768). Notably, no grass carp samples, 8% of silver carp, and 14% of bighead carp samples were below detection limits (Table 1), whereas the majority (~66%) of black carp fell below assay detection limits.

Censored permutation ANOVAs suggested there were no differences between fish sourced from outdoor rearing ponds or indoor rearing spaces for either grass carp (mean_pond = 65.9, mean_indoors = 58.8; p = 0.235 to 0.235) or silver carp (mean_pond = 14.9, mean_indoors = 14.7; p = 0.851 to 0.880).

Bayesian linear models indicated that there were no relationships between thiaminase activity and standard length of bighead or grass carp (Fig. 3a,c; Table 2). Black carp and silver carp had statistically negative relationships between standard length and thiaminase activity (Table 2, Fig 3b,d). All 95% credible intervals strongly overlapped zero for the relationship between carp body condition and thiaminase activity (Fig 4, Table 2), suggesting that body condition does not predict thiaminase activity in juvenile bighead carp, black carp, grass carp, and silver carp.

Scientists do not fully know why some fishes produce thiaminase (Rowland et al., 2023), nor why there is such wide variation in activity even among the same species collected at the same location on the same day (Wistbacka et al., 2002). Multiple researchers have noted rapid elevation of thiaminase activity in laboratory studies over short periods of time (Lepak et al., 2008; Wistbacka and Bylund, 2008), presumably as a stress response, although the reasons for this are unresolved. In bacteria, thiaminase is a thiamine salvage pathway (Sannino et al., 2018) but this has not been demonstrated in fishes. There is strong evidence of elevated thiaminase in fish exposed to microbial pathogens (Wistbacka et al., 2009), and seasonal variation has been documented for at least some species (Tillitt et al., 2005). The very high levels of thiaminase activity we measured in grass carp juveniles did not differ between indoor and outdoor cultures, suggesting thiaminase activity was not related to diet, stress, or temperature, similar to previous experimental work on clupeids (Lepak et al., 2013). For black carp and silver carp, we found a slight negative relationship between total length and thiaminase activity, opposite of field observations of a positive correlation between alewife size and thiaminase activity in summer and fall, but not spring (Tillitt et al., 2005). However, limiting our samples to only juveniles within a narrow range of sizes may have limited our ability to detect trends. Additionally, all our study animals had body condition > 1, indicating that they had no food stress. Future research could explore patterns in the field and assess how thiaminase may vary within grass and silver carp across seasons, ages, and body condition/starvation indices. Additionally, although fish can produce thiaminase de novo (Richter et al., 2023), it is possible that the very low thiaminase activity seen in bighead and black carp could represent the microbiota contribution of thiaminase (Honeyfield et al., 2002) compared to grass and silver carp that are producing their own thiaminase. The thiaminase I-producing bacterium P. thiaminolyticus has been observed as a small component of the gut microbiome of alewife in the Great Lakes (Richter et al., 2012), and P. thiaminolyticus and/or other thiaminase I-producing bacteria could be present in other fish as well. The relative contributions of the microbiome vs. de novo production is an open question.

We found juveniles of a size that gape-limited predators could consume had consistent thiaminase activity. Potential negative effects of consuming thiaminase-producing fishes such as invasive carp are dependent upon both the thiaminase activity at time of consumption, which can vary within a species (Lepak et al., 2013), and the proportion of the diet containing high thiaminase activity prey (Honeyfield et al., 2005). In the Great Lakes, there is strong evidence that alewife thiaminase is linked to lake trout declines (Fitzsimons et al., 2010; Ladago et al., 2020; Riley et al., 2011). Piscivores with diets transitioning to a larger reliance on grass carp or silver carp could be at risk for developing thiamine deficiency complex and the symptoms that follow (Harder et al., 2018).

Comparisons with previous work are complicated by differences between the radiometric assay that directly measures thiamine degradation and allows flexibility in pH (Tillitt et al., 2005; Zajicek et al., 2009) compared to the present study assay that uses 4-NTP as the nucleophile and restricts pH to 6.9 (Hanes et al., 2007). However, a recent advance in thiaminase activity assays found thiamine degradation and 4-NTP assays yielded similar results (Porter and Pinger, 2025), so we will compare with these caveats in mind. If we generously assume the current assay represents optimum conditions for carp thiaminase, the thiaminase activity of grass carp is some of the highest reported in Great Lakes food webs, and the activity could be even higher if the pH optimum of grass carp thiaminase differs significantly from 6.9 as it does for some species (Zajicek et al., 2009). Alewife activity measured using the radiometric assay found 4 nmol/g/min (Tillitt et al., 2005) is enough to induce thiamine deficiency, Wolfe et al. (2023) using the 4-NTP assay found a mean thiaminase activity in adult silver carp viscera of 9 nmol/g/min, which was lower than the mean whole-body thiaminase activity we measured in juveniles of approximately 15 nmol/g/min. Average grass carp thiaminase activity of 60 nmol/g/min in the present study is on par with northern anchovy (Engraulis mordax) thiaminase activity (avg = 48 nmol/g/min) linked to thiamine deficiency in Pacific Chinook salmon (Mantua et al., 2025). However, there is a lot of uncertainty surrounding which predators may feed on invasive carp if they become established (Robinson et al., 2021). Anderson et al. (2021) found bighead carp and silver carp common within the diet of native predatory fish in the Illinois River, though this changed throughout the season. Rapid growth rates of invasive carp may facilitate a shift in diet to other available forage fish species as invasive carp grow to exceed the mouth gape of predatory fish. Regardless, juvenile silver and grass carp have high enough thiaminase activity to cause thiamine deficiency complex if they are a large part of a predator’s diet. Alewife comprising 35% or more of the diet of lake trout was enough to cause thiamine-dependent fry mortality (Honeyfield et al., 2005).

Thiaminase activity has been documented in other invasive forage fish species within the Great Lakes, including alewife and rainbow smelt (Tillitt et al., 2005). Great Lakes salmonines consuming grass and silver carp may be at additional risk for developing thiamine deficiency if they switch to consuming grass or silver carp. The potential for developing thiamine deficiency by consuming thiaminase-containing invasive carp extends beyond predatory fishes to other piscine predators such as waterbirds and mammals. Thiamine deficiency in European common eider (Somateria mollissima) has been linked massive population declines in recent years (Mörner et al., 2017). Waterbirds of the Great Lakes region like double-crested cormorant (Nannopterum auritum), mergansers (Mergus spp.), bald eagles (Haliaeetus leucocephalus), osprey (Pandion haliaetus), or great blue heron (Ardea herodias) are all piscivorous (Hebert and Morrison, 2003) and may be susceptible to thiamine deficiency induced by high thiaminase prey. Bald eagles may also scavenge potentially larger-bodied invasive carp carcasses or directly predate on living carp. Mammalian predators, like the river otter (Lontra canadensis) may also consume carp as part of their diet (Raesly, 2001). The timing of predation and subsequent physiological effects of a high thiaminase diet would depend on prey size and predator- prey behaviors at the time of co-occurrence. Migratory avian predators for example may not co-occur with invasive carp among habitats at the same time of year as prey size facilitates predation.

Dietary thiaminase can affect piscivore health directly. Experimental dietary thiaminase has been associated with reduced cardiac performance (Baker et al., 2023), decreased body condition and lower swimming performance (Houde et al., 2015), and negative effects on growth, survival, and food conversion efficiency (Therrien et al., 2025) in salmonines. High-thiaminase rainbow smelt was correlated with heart irregularities such as increases in ventricular mass, reductions in ventricular fluid content, and changes to the allometry of myocardia (Adeli et al., 2025). It is interesting that some predators seem to be selecting for bighead carp (Anderson et al., 2021), a species with low thiaminase activity. More research is needed to understand which invasive carp are a likely forage base to help us understand whether the very high thiaminase activities observed in grass carp and silver carp are a risk to aquatic and avian predators. Additional data on whether predators may preferentially avoid species with high thiaminase could be useful.

Importantly, enzymes are denatured with heat and suboptimal pHs. Silver carp in previous work had no detectable thiaminase activity when cooked (Wolfe et al., 2023). There is no risk for human consumption of thiaminase positive fishes if the enzyme is denatured. Research suggests carps are nutritious and a sought-after food source elsewhere in the world (Li et al., 2021). The risk of consuming thiaminase positive fishes only applies to raw fish, so harvesting invasive carp for human consumption and species control (Varble and Secchi, 2013) presents no risk to consumers of cooked carp even in species with high thiaminase activity.

The Laurentian Great Lakes are fragile ecosystems that are vulnerable to aquatic invasive species (Sterner et al., 2017). Losses from commercial and sportfishing due to aquatic invasive species are estimated to cost several hundred million dollars per year (Rothlisberger et al., 2012). Among the most damaging invasive species in the Great Lakes, two thiaminase-positive species, alewife and grass carp, were ranked by Lower et al. (2024) among the top ten species for most significant negative environmental and socio-economic effects. The current study adds an axis for grass carp; their extremely high thiaminase activity has potential to negatively affect predator health and demography. Our data suggest that silver and grass carp consumed in large quantities may put piscivorous fishes, birds, and mammals at risk of developing thiamine deficiency.