Hydrogen and oxygen isotope values in muscle are grouped by diet treatment ( Figures 1 , 2 ). Diet-tissue offsets (Δδ²H and Δδ¹⁸O) in muscle vary by diet in a consistent manner in all three species ( Figures 3 , 4 ; Table 2 ). Δδ²H is smallest for plant, mid-sized for insect, and largest for the meat diets. The difference between the diet groups is in the range of Δδ²H = 4−16 ‰ ( Table 3 ). Similarly, Δδ¹⁸O is smallest for the plant diet and a greater and overlapping difference for the insect and meat diets, for all three animal groups (excluding ¹⁸O-enriched treatment quail), with differences of ~1.2−2.0 ‰ ( Figure 4 , Table 4 ). The eight quail given ¹⁸O-enriched drinking water have higher δ¹⁸O in muscle than the quail consuming regular drinking water (12.9−13.4 ‰ vs. 8.1−9.7 ‰, respectively, Supplementary Table 2 ), consistent with the quail incorporating the drinking water and experimental new diets into the muscle tissue.

In dentine collagen, there is a large variation in the Δδ²H and Δδ¹⁸O values, but some isotopic patterning by diet broadly similar to the muscle results. In dentine collagen, Δδ²H is smaller in the meat-containing diet than the plant- or insect-based diets for the guinea pigs by ~20 ‰ (Tukey HSD p ≤ 0.0031) and ~12 ‰ for rats (non-significant differences); there is a larger and equal Δδ²H for the insect and plant diets ( Figure 3 ; Table 5 ). While the small number of samples precludes a strong comparison, Δδ¹⁸O is smallest on the plant diet and larger on the meat and insect diets for the guinea pig, in agreement with the pattern observed for Δδ¹⁸O in muscle ( Figure 4 ).

There are large and systematic isotopic differences by tissue between muscle and dentine collagen for both H and O in both rats and guinea pigs ( Table 2 ): higher δ²H and δ¹⁸O values in dentine collagen than muscle ( Figures 1 , 2 ); a big positive diet-tissue Δδ²H in dentine collagen (~ +10−40 ‰) vs. ~ -20 ‰ (opposite direction) Δδ²H in muscle ( Figure 3 ); and ~4 ‰ greater magnitude (negative) Δδ¹⁸O in dentine collagen than muscle ( Figure 4 ).

Inter-species isotopic differences are generally smaller than the differences between diets. However, there are δ²H differences between rat and guinea pig dentine collagen (by ~20 ‰ on average, though with a large range and SD of the mean of 7−11 ‰), and in muscle for plant-fed rats compared to plant-fed guinea pig and quail (7−8 ‰). Rat muscle δ¹⁸O is somewhat higher than in guinea pigs and quail on all three diets (by 0.7−1.1 ‰, Table 2 ).

The muscle δ²H and δ¹⁸O isotope values are tightly grouped by diet and species, and diet type affects the diet-tissue H and O isotope offset in muscle in all three species in the same way. In this and other controlled feeding experiments, tissue H and O isotope values are often tightly grouped within each diet or water treatment; the typical intra-group uncertainties are 2.4 ‰ for δ²H and 0.44 ‰ for δ¹⁸O (median values of 2 x the standard error of the mean, compiled in Figures 5 , 6 ; Hobson et al., 1999; Tuross et al., 2008; Podlesak et al., 2008; Kirsanow and Tuross, 2011; Wolf et al., 2012; Wolf et al., 2013; Newsome et al., 2017; Rodriguez Curras et al., 2018; Topalov et al., 2019). Our isotopic differences by diet greatly exceed this intra-group variability. The magnitude of the difference for Δδ²H of ~ 4–16 ‰ and Δδ¹⁸O up to ~2 ‰ is relatively small but not negligible ( Figure 4 ; Tables 3 , 4 ); greater than typical intra-group uncertainties, but also smaller than the variance found in some studies (discussed further in sections 4.4 and 4.5, e.g. δ²H can range up to 20–80 ‰ in bird keratin, Hobson et al., 2014). In dentine collagen in the guinea pig Δδ²H varies by ~20 ‰ and Δδ¹⁸O by ~1.5 ‰ between the diets (diet group mean, Figures 3 , 4 ).

In the same individual animal, dentine collagen and muscle Δδ²H and Δδ¹⁸O values are generally offset in parallel ( Figure 7 ), indicating that both tissues are reflecting the same isotopic dietary input, and further, that conclusions drawn from the muscle results are probably generalizable to collagen (dentine and bone) and other proteinaceous tissues such as feathers.

This study does not directly elucidate the mechanisms responsible for these observed diet-tissue H and O isotope differences, and further work would be required to address this. Digestion, absorption, biochemical transformation, and differential incorporation of diet macronutrients and amino acids into tissues impart isotopic change (Macko et al., 1986; Bowen et al., 2009; Vander Zanden et al., 2016; Newsome et al., 2017; Magozzi et al., 2019; Newsome et al., 2020). The diet-tissue offset is plausibly affected by the protein composition and/or lipid and carbohydrate amounts (e.g. for nitrogen isotopes, protein quality influences diet-tissue offsets, e.g. Robbins et al., 2005).

The diets in this present study are isonitrogenous (Leichliter et al., 2021), but they differ in their major component of lucerne, lamb, or soldier fly larval insect meal ( Table 1 ). We first consider the meat- and insect-based diets. These two diets are nearly the same in the non-meat/non-insect components ( Table 1 ), so that the only difference is between the main component, the larval meal or lamb meat. The δ²H values of the lipid-removed diets are the same (non-lipid, -77 ‰ vs. -75 ‰, Table 1 ), which means that the main ingredient in these diets must have the same δ²H values as each other. There is no carbohydrate in lamb (U.S. Department of Agriculture and Agricultural Research Service, 2019), so the dried and defatted lamb meat is principally protein. The larval meal contains chitin, a glucosamine-based polymeric carbohydrate. In black soldier fly meal chitin levels can be in the range of 5−9% (dry matter basis, variable defatting; Finke, 2013; Schiavone et al., 2017; Caligiani et al., 2018). Recent work has shown that some mammals and birds express chitinases and can therefore hydrolyze chitin to the glucosamine monomer, but the expression and activity of Chia (one of the chitinases) varies between species (Tabata et al., 2018). Mice, chickens, and pigs have higher Chia expression than dogs or cattle; and guinea pigs lack the Chia gene (Tabata et al., 2018). One might predict rats and quail to have some ability to hydrolyze chitin. Glucosamines are absorbed by the gastrointestinal tract and can be further metabolized (Setnikar and Rovati, 2001; Anderson et al., 2005).

Despite the main component having the same δ²H value, Δδ²H in muscle varies significantly between the meat and insect diets (6–8 ‰, Tukey HSD p ≤0.002, Table 3 ). This suggests a few possibilities: first, if the protein δ²H value is the same in the meat- and insect-main components, then the protein quality or type affects the protein H utilization and diet-tissue fractionation (Δδ²H). Second, if the chitin has a different δ²H value than the insect-meal protein, then the inferred different protein δ²H between meat- and insect-diets could be the cause of the Δδ²H difference. Third, any incorporation of chitin H into tissue could also affect δ²H on the insect-based diet. Given the chitin component is likely small (<10% w/w of the main component), it is perhaps more likely that differences in protein quality and amino acid composition are important in the resulting Δδ²H offsets and the differences between the meat-based and insect-based diets.

In contrast, Δδ¹⁸O presents a more mixed picture in comparing the insect and meat diets. The insect and meat diet muscle δ¹⁸O values are fairly close at ~0.6 ‰ apart, and the resulting Δδ¹⁸O is significantly different between those diets only for the rat (0.6 ‰, Tukey HSD p < 0.001, Table 4 ).

The plant diet results in very different and smaller Δδ²H and Δδ¹⁸O in muscle than the other two diets ( Figures 3 , 4 ). The difference of the plant diet Δδ²H to that of other diets is significant (Δδ²H = 12−16 ‰ plant-meat, 4−10 ‰ plant-insect, Tukey HSD p ≤0.013, Table 3 ). Similarly, Δδ¹⁸O differences between plant and other diets are up to 2.1 ‰ ( Table 4 ). The plant diet has a higher proportion of the main ingredient (lucerne), additional potato protein, and a lower proportion of the remainder of the diet components (to maintain an iso-nitrogenous condition, Table 1 ). The nutritional composition of the diets is similar, except the plant-based diet is somewhat lower than the other two diets in neutral detergent fibre and starch ( Table 1 ). The variable dietary macronutrients (proportion of each and their isotope values) in the plant diet and/or the protein composition itself are possible reasons for the observed diet-tissue offset differences from the other two diets in both H and O.

The model elaborated by Magozzi et al. (2019) includes routing from dietary protein to tissue protein (keratin), meaning the protein component plays a strong role in setting the tissue isotope value (e.g., 60% of dietary protein H to keratin). Routing from the dietary protein with a different isotopic composition is plausibly why the plant diet in particular has different Δδ²H and Δδ¹⁸O than the insect and meat diets; a second effect may be variation in the fraction of dietary protein H and O routed to tissue between these diet types. In controlled feeding experiments where the water input is changed, the estimated total food contribution to tissues ranges from 70−85 % H in different proteinaceous tissues such as collagen/hair/feather/nail (Kirsanow and Tuross, 2011; Hobson et al., 1999; Topalov et al., 2019) and 30−80 % H in muscle (Hobson et al., 1999; Wolf et al., 2012); this variability may be a real reflection of different routing of atoms from diet component to tissue in different diet conditions as well as differences in how the experiments were conducted (see Vander Zanden et al., 2016 for more data and discussion).

The results here agree with previous studies showing that isotopic values vary by tissue (e.g., Tuross et al., 2008; Kirsanow and Tuross, 2011; Wolf et al., 2012, for H and O). Collagen H and O isotope values are reported in only a few other controlled feeding studies (rats: Kirsanow and Tuross, 2011; pigs: Tuross et al., 2008; mice: Topalov et al., 2019), and only one reports both muscle and collagen (Tuross et al., 2008). The isotopic pattern seen here agrees with bone collagen results found in pigs (Tuross et al., 2008): there are higher δ²H values and lower δ¹⁸O values in collagen (here, in dentine collagen) than in muscle. We find a large isotopic difference between dentine collagen and muscle of δ²H = ~55 ‰ (range 40−67 ‰) and δ¹⁸O = -3−4 ‰ with the same diet, species, and water combinations ( Figures 1 , 2 ). These inter-tissue differences are important to bear in mind in ecological or migration studies.

There is slight δ¹⁸O patterning in muscle by species for all three diets, with δ¹⁸O values following quail < guinea pigs < rats (0.2−1.5 ‰, Figure 2 ; Table 6 ). Hydrogen isotopes also vary between species on some diets in muscle, and more strongly in dentine collagen for all three diets (13−24 ‰, Figure 1 ; Table 7 ). Inter-species variation is not unexpected given the effect of metabolic rate and water flux (particularly from drinking water) on resultant tissue oxygen and hydrogen isotopes (O: Bryant and Froehlich, 1995; Kohn, 1996; H and O: Magozzi et al., 2019). Many of the relevant H and O fluxes scale with body mass with an allometric coefficient less than one (i.e., flux ∝ mass^x, where x < 1); some of these include the metabolic rate (and thus O₂ and food consumption), total water flux, and water vapour loss. These fluxes have slightly different scaling relationships with body mass, so that the balance between them is important in determining the resultant body water and tissue isotopic compositions. Total water flux (and often thus the liquid water drinking rate) increases with body mass (Nagy and Peterson, 1988; Bryant and Froehlich, 1995; Kohn, 1996); albeit with different scaling relationships for captive and wild animals as well as between wild herbivorous mammals, wild omnivorous mammals, and wild birds (Nagy and Peterson, 1988).

In terms of isotopic effects, body water and tissue δ²H and δ¹⁸O values are predicted to decrease (body water in the direction of the drinking water isotopic value), with increasing liquid drinking water inputs and increasing total water flux. More isotopically fractionated water loss (as vapour), results in higher body water and thus higher tissue δ²H and δ¹⁸O values (Kohn, 1996). Given the complexity of the varying influences on H and O fluxes, of which only a few were mentioned above, it is difficult to predict simply the expected H and O isotope behaviour for these three species. In any case, these results highlight that it is important to bear this inter-specific isotopic variation in mind.

Typically, isotope-based animal migration studies use an accessible and inert tissue, such as hair, feathers or claws. Controlled feeding experiments have shown that feathers or hair and muscle are isotopically related to each other; i.e. broadly speaking δ²H in tissues all shift systematically in concert with changes in water and/or dietary inputs (Hobson et al., 1999; Wolf et al., 2012; Newsome et al., 2017). As such the variations in muscle Δδ²H and Δδ¹⁸O by diet we observe here are likely relevant to migration and ecological studies using other tissues such as feathers.

Long-distance animal migration studies often have measured tissues with large variances in δ²H values between individuals in a given group. This present experiment’s Δδ²H differences with diet in muscle of ~4−16 ‰ and in dentine collagen of ~13−23 ‰ are small/moderate relative to the long-range geospatial variation in precipitation δ²H values and resultant tissue δ²H values; e.g., feathers can range up to ~80 ‰ in one species from across North America captured in one over-wintering location, and can typically be ~20−30 ‰ (Hobson et al., 2014). Idaho-resident and non-resident American kestrel claws differ in δ²H by up to ~40 ‰ (Ranck et al., 2023). The isotopic variation with diet type is small enough relative to the large geospatially-related variation in precipitation δ²H but might be important in cases with less of a δ²H gradient in environmental water and thus potentially blur the geospatial assignment of individuals.

Furthermore, an observational study also agrees with our experimental result that diet can affect tissue δ²H values. In non-migrant wild birds from the same locale (caught by mist nets, mostly smaller birds, van Wijk et al., 2021), feather δ²H values across species show differences between five dietary guilds (nectarivores, frugivores, granivores, omnivores, insectivores) of 2−30 ‰ – all much larger than the differences in muscle in this present controlled feeding experiment but somewhat closer to the differences we observe in dentine collagen. This variation in δ²H in the wild birds may result from a large variation in the δ²H values of the diet components, as well as any inter-species physiological differences.

The Δδ¹⁸O variations by diet in muscle of up to ~1.2−2.0 ‰ ( Figure 4 ; Table 4 ) and ~1.5 ‰ in dentine collagen are relatively large ( Figure 4 ; Tables 2 , 4 ). While studies of oxygen isotopes in organic tissues are few thus far, feather δ¹⁸O reflects 60−80% of the variation in environmental water δ¹⁸O and thus is dampened in its response to changing water δ¹⁸O values, yielding a reduced range of tissue δ¹⁸O values (Hobson and Koehler, 2015; Magozzi et al., 2019). The greater δ¹⁸O variation in tissue (muscle, dentine collagen) relative to the reduced geospatially-related variation in δ¹⁸O (for feathers) means that in oxygen isotope studies for migration- and ecology-related questions, greater attention is needed to consider differences in diet type as a modulating factor.

However, this varied diet-tissue δ²H and δ¹⁸O offset by diet type is possibly mitigated by the fact that many species have rather consistent dietary habits. However, in the case of omnivores/generalists, humans, and species that may vary their diet type (e.g., seasonally or through ontogeny), the differences in diet-tissue isotopic offsets may add extra variability in tissue isotopic values.

The Δδ²H difference in dentine collagen with the meat-based diet has interesting implications for archaeological or palaeoecological samples where bone or dentine collagen is analyzed – that diet is an important factor to consider in resultant collagen δ²H values, particularly in omnivores such as humans. The effect of diet type is supported by studies on archaeological samples: in dentine collagen from human first molars δ²H varied significantly along incrementally sampled sections (up to 30 ‰ range), corresponding to changes over the first few years of life (Ryan et al., 2020). These variations were likely dietary in nature as juveniles changed their diets from mother’s milk to possibly plant-food gruel (cooked) and further towards an adult diet (as represented by adult bone isotopic values).

There are limited studies on δ¹⁸O in archaeological protein (collagen), but our results in dentine collagen suggest diet type will be an important consideration in interpreting δ¹⁸O values, among other parameters such as water flux, metabolism, and geographic locality. Data thus far are mixed: medieval bone collagen δ¹⁸O from the same site shows considerable differences between species (up to 3 ‰), including humans (Ryan et al., 2018); in contrast, little inter-species δ¹⁸O variation is seen in bone collagen in Bronze Age archaeological material but a large intra-species range is noted (Reynard et al., 2020). Magozzi et al. (2019) predict lowering of δ¹⁸O in tissue protein with trophic level in a strictly-modelled trophic system where the diet input is the protein from the previous trophic level.

There is no a priori expectation for the effect of protein type or amount on resultant tissue isotopic values. Generally the isotopic composition of macronutrients in a given plant rank from δ²H _carbohydrate > δ²H _protein > δ²H _fat (Estep and Hoering, 1980; da Silveira Lobo Sternberg, 1989), δ¹⁸O _carbohydrate > δ¹⁸O _protein and δ¹⁸O _carbohydrate > δ¹⁸O _fat (Silva et al., 2015), and in animal muscle δ²H _protein > δ²H _fat ( Supplementary Table 9 ) so that the result of any change in protein routing to tissue (and thus over- or under-representation) is not straightforward to predict.

Acknowledging the small number of dentine samples, our results suggest for both δ²H and δ¹⁸O in bone or dentine collagen that there are important inter-species differences to consider when interpreting archaeological material, as well as consideration of variation due to changes in diet. For example, inter-site or temporal comparisons should involve the same or very similar species (in diet and physiology). In species with varied diets, e.g. humans, care is needed in interpretations of small magnitude δ²H and δ¹⁸O variations, as diet may have played a role in the noted bone or tooth isotopic variability.