Endocrine and metabolic impacts of warming aquatic habitats: differential responses between recently isolated populations of a eurythermal desert pupfish

Imperilled desert fishes occupying isolated habitats that limit dispersal are vulnerable to temperature alterations because of a changing climate. Here, we identify differences in temperature effects on thyroid hormone signalling, anaerobic metabolism and thyroid hormone-mediated metabolic gene expression between recently isolated populations of a desert pupfish.


Introduction
Inland aquatic habitats are warming more rapidly than either the oceans or the atmosphere under a changing global climate (Schneider et al., 2009;Layden et al., 2015;O'Reilly et al., 2015). Such rising temperatures are expected to alter freshwater fish habitat use, shift species' ranges and life histories, and induce a suite of physiological responses to high temperature (Wood and McDonald, 1997;Roessig et al., 2004;Ficke et al., 2007). Understanding how fish respond physiologically to elevated temperature-and how these responses vary among taxa or populations-will be crucial for predicting how fish communities will change in the face of warming habitats (Pörtner and Farrell, 2008), as well as for evaluating whether species persistence will be aided by plasticity in thermal tolerance (Hofmann and Todgham, 2010;Bozinovic and Pörtner, 2015).
Pupfishes (genus Cyprinodon) of the North American southwestern deserts provide a model set of taxa for examining the fundamental mechanisms for coping physiologically with high temperatures. Pupfish in this region evolved some of the highest critical thermal maxima and broadest thermal tolerances recorded in freshwater fishes (Beitinger et al., 2000;Motani and Wainwright, 2015), with several species of pupfishes surviving temperatures exceeding 42°C (Brown and Feldmeth, 1971;Otto and Gerking, 1973;Feldmeth, 1981). In this study, we examined the influences of temperature on the endocrine physiology and metabolism of Amargosa pupfish, Cyprinodon nevadensis amargosae, from the Death Valley region of California and Nevada, USA. This extreme desert region is home to a clade of three pupfish species partitioned into eight subspecies (Miller, 1948;Soltz and Naiman, 1978), which appear to have evolved within the last 5000-20 000 years (Martin et al., 2016). Similar to most other pupfishes through the North American desert southwest, Death Valley's pupfishes occupy restricted geographical ranges and are isolated in remote groundwaterfed springs, desert streams or marshes (Soltz and Naiman, 1978). Isolation in such small, discrete habitats has made pupfishes in this region vulnerable to human activities (Fagan et al., 2002). One pupfish from the Death Valley region is already extinct (Miller et al., 1989), and several other pupfishes from this clade have 'threatened' or 'endangered' legal status (Moyle et al., 2011).
Amargosa pupfish occur in several habitats within the Death Valley region, including~13 km of permanent fluvial habitat in the Amargosa River and a small artesian spring referred to as Tecopa Bore, where water emerges from the ground at 47.5°C and cools gradually as it flows through an associated marsh (Naiman, 1974). This Tecopa Bore habitat was created inadvertently in 1967 as a result of commercial mineral drilling, and C. n. amargosae pupfish from the Amargosa River dispersed into the habitat shortly thereafter. Pupfish are found in areas of the Tecopa Bore habitat at temperatures exceeding 36°C, although preferred temperatures may be closer to 30°C (McCauley and Thomson, 1988). In contrast, the Amargosa River is a desert stream with temperature conditions that vary from near freezing to >40°C seasonally and can change more than 25°C diurnally (Soltz and Naiman, 1978). Recent examination of these two populations of C. n. amargosae found that pupfish in the extreme thermal environment of Tecopa Bore are not only smaller in length and mass than conspecifics from the Amargosa River, but that Tecopa Bore fish also differ in body shape and frequently lack pelvic fins (Caldwell et al., 2015). A previous study found that pharmacological inhibition of endogenous thyroid hormone (TH) production during larval life induces C. n. amargosae to develop a proportionally larger head and eye, a reduced body depth, and no paired pelvic fins (Lema and Nevitt, 2006). Cyprinodon nevadensis amargosae in Tecopa Bore mirror this same morphology, with a proportionally large head and eye, reduced body depth, and lower occurrence of pelvic fins (Caldwell et al., 2015). Notably, larval pupfish exposed to elevated temperatures are also less likely to develop paired pelvic fins (Lema and Nevitt, 2006;Lema, 2008Lema, , 2014. In teleost fishes, pelvic fins develop during the transition from a larval life stage to a juvenile fish (Yamanoue et al., 2010), and this developmental change is mediated in part by THs (Brown, 1997). Given that evidence, it is possible that the observed morphological differences between the Amargosa River and Tecopa Bore populations might arise in part from altered TH physiology in Tecopa Bore pupfish because of the consistently high temperatures of their habitat.
Thyroid hormone signalling might also play an important role in metabolic acclimation to temperature variation in teleost fishes (Little et al., 2013), as it does in mammals (Silva, 1995;Yen, 2001;Mullur et al., 2014). Zebrafish (Danio rerio) acclimated to 18 or 28°C and treated with a pharmacological goitrogen to inhibit endogenous TH production subsequently exhibited dissimilar swimming performance, metabolic scope and metabolism-associated gene expression in patterns indicative of impaired TH-mediated metabolic regulation at warm temperature (Little et al., 2013). Several studies in fishes have also found that exposure  (Leatherland et al., 1977(Leatherland et al., , 1980Lee et al., 2014). For instance, rainbow trout (Oncorhynchus mykiss) acclimated to 11°C and transferred to 19°C exhibited reduced plasma thyroxine (T 4 ; the primary circulating TH in teleost fishes) compared with trout transferred to 5°C (Eales et al., 1982). This reduction in T 4 at 19°C was accompanied by a 4-fold increase in T 4 metabolic clearance and conversion to 3,5,3′-triiodothryonine (T 3 ), a more active structural form of TH (Eales et al., 1982). Thyroid hormone conversion occurs via the action of iodothyronine deiodinase enzymes, with type II deiodinase (Dio2) mediating outer-ring deiodination (ORD) activity to convert T 4 to T 3 , and type III deiodinase (Dio3) catalysing inner-ring deiodination (IRD) to convert T 4 to the less active THs reverse T 3 (rT 3 ) or 3,5-diiodothyronine (T 2 ; Köhrl, 2000;Orozco and Valverde, 2005). In fish, there is evidence that deiodinase activity may vary in patterns linked to temperature. For instance, in trout liver microsomes, rates of ORD of T 4 were 2.4-fold higher at 12-13°C and 2.8-fold higher at 18°C than at 4°C (Johnston and Eales, 1995).
In this study, we examined whether temperature experience alters TH physiology and metabolism in patterns that help to explain the observed phenotypic differences between the Amargosa River and Tecopa Bore populations of C. n. amargosae. First, we tested whether these two populations differ in TH physiology by quantifying plasma T 4 and hepatic gene transcript abundance for deiodinase enzymes type I (dio1), type II (dio2) and type III (dio3), as well as TH receptors (trαA, trαB and trβ) in the wild. Based on the morphological variation between the populations and our previous experimental findings of how food availability and temperature alter morphological development in pupfish (e.g. Lema and Nevitt, 2006;Caldwell et al., 2015), we hypothesized that C. n. amargosae in Tecopa Bore would exhibit characteristics resembling a hypothyroid state, including reduced plasma T 4 and higher hepatic dio2 mRNA levels. Second, we hypothesized that any observed population differences in TH physiology between the Amargosa River and Tecopa Bore populations would be caused predominantly by the differing temperature experience of these populations. We tested this idea by acclimating pupfish from both populations to 24 or 34°C in captivity for 88 days. At the end of this 88 day thermal acclimation period, we administered exogenous T 3 for 18-24 h to a subset of pupfish from both populations and temperature treatments to activate T 3 -regulated changes in gene expression and metabolism. Liver and skeletal muscle tissues were examined for differences in the relative abundance of gene transcripts encoding TH signalling-associated proteins, including deiodinase enzymes, TH receptors and the TH membrane transporters monocarboxylate transferase 8 (mct8) and organic anion-transporting polypeptide 1c1 (oatp1c1; Muzzio et al., 2014). Citrate synthase (CS) activity was also quantified as an indicator of aerobic metabolic capacity, and lactate dehydrogenase (LDH) activity was examined as an indicator of anaerobic metabolic capacity. The relative abundance of mRNAs encoding several metabolismlinked proteins was also quantified to evaluate T 3 induction of metabolic gene expression.

Study populations
Cyprinodon nevadensis amarogsae pupfish were studied from two allopatric populations, the Amargosa River (35°5 1.275′N, 116°13.833′W) and Tecopa Bore (35°53.140′N, 116°14.050′W), located in the Death Valley desert region of eastern California, USA. All procedures were approved by the Animal Care and Use Committee of California Polytechnic State University (protocol #1507).

Animal collection
On 9 May 2014, adult C. n. amargosae pupfish were collected from the Amargosa River and Tecopa Bore habitats using minnow traps. A subset of pupfish from the Amargosa River population (n = 9 fish of each sex) and from Tecopa Bore (n = 8 males and n = 11 females) were immediately euthanized (tricaine methansulfonate, MS222; Argent Chemicals, Inc.), and then length and mass were measured. Blood was collected and centrifuged, and plasma stored for subsequent T 4 quantification. Liver tissues were dissected and immersed in RNAlater ® (ThermoScientific) at 4°C for 48 h before being stored at −20°C.
Additional adult C. n. amargosae pupfish collected from the Amargosa River and Tecopa Bore on 9 May 2014 were transported to California Polytechnic State University, San Luis Obispo, CA, USA. After collection, fish were maintained in 208 litre holding tanks to allow for acclimation to captive conditions. Pupfish from the Amargosa River were maintained at 24°C, and Tecopa Bore pupfish at 34°C. For the duration of the study, all fish were fed a 1:1 mixture of commercial spirulina (Aquatic Eco-Systems, Inc., Apopka, FL, USA) and brine shrimp (San Francisco Bay Brand, Inc., Newark, CA, USA) flake feeds ad libitum twice daily and were maintained under a 14 h light-10 h dark photoperiod in synthetic 2.1 ppt water made with Instant Ocean ® salt (Unified Pet Group, Inc., Blacksburg, VA, USA) and deioinized water. After 10-15 days in the 208 litre holding tanks, pupfish were assigned to 38 litre experimental tanks (four replicate tanks per treatment), with four fish (two males and two females) per tank. Fish from each population were maintained separately in these 38 litre tanks for 88 days in two temperature conditions: 24 or 34°C. Temperature measurements recorded using HOBO ® U12 External Data Loggers (Onset Corp., Bourne, MA, USA) over the 88 days period confirmed the temperature treatments as 23.97 ± 0.40 and 34.18 ± 0.24°C (means ± SD). On the day immediately prior to sacrifice, four tanks from each population-temperature combination were treated with waterborne T 3 (dissolved in 0.01 M NaOH; Sigma-Aldrich) for an exposure dose of  Johnson and Lema, 2011;Jones et al., 2016). All other tanks received control vehicle only. At 18-24 h after hormone treatment, pupfish (n = 120) were netted from the experimental tanks, euthanized (MS222), and weighed and measured. Blood was centrifuged (850 × g for 10 min at 4°C) for plasma collection, and liver and skeletal muscle tissues were dissected, flash frozen in liquid N 2 and stored at −80°C. All fish were sampled between the hours of 13.00 and 16.00 h to reduce photoperiod effects.

Radioimmunoassays of T 4 and T 3
Plasma total T 4 and total T 3 were quantified using radioimmunoassay (RIA). Detailed methods for this RIA procedure are provided by Dickhoff et al. (1982). Anti-L-T 4 antiserum (1:3500) and anti-L-T 3 antiserum (1:10 000) were obtained from Fitzgerald Industries International (Acton, MA, USA). Given the small body size of C. n. amargosae [typically <40 mm standard length (SL)], plasma from some fish was pooled to obtain sufficient volumes for RIA. Samples were run in duplicate when allowed by plasma volume. The intra-assay percentage coefficient of variation for the T 3 RIA was 4.83%, and the inter-assay coefficient of variation was 6.21%. All samples for the T 4 RIA were run singly because of limited plasma volume.
Pupfish acclimated to 24 or 34°C for 88 days and then treated with exogenous T 3 were confirmed to have elevated plasma T 3 concentrations, with the magnitude of that elevation influenced by acclimation temperature (temperature × hormone: F 1,61 = 8.732, P = 0.0044; Table 1). Mean plasma T 3 ranged between 11.3 and 15.1 ng/ml for pupfish not treated with exogenous T 3 , and there was no apparent effect of population origin or temperature acclimation on T 3 concentrations in those fish. Plasma T 3 increased to 95.0 ng/ml (mean, both populations combined) in pupfish at 24°C after 18-24 h of exposure to waterborne T 3 (15 nM), but to 328.4 ng/ml in fish acclimated to 34°C, indicating that temperature strongly influenced the increase in circulating T 3 following exogenous T 3 treatment.

Citrate synthase and lactate dehydrogenase enzyme activity
Enzyme activity for both CS and LDH was quantified using methods described by Walsh and Henry (1990), with modifications to the CS assay as presented by Boyle et al. (2003) and Johnson et al. (2004). Liver and skeletal muscle tissues were homogenized (Fisher Scientific PowerGen Homogenizer) in 5-65 volumes of Tris-HCl buffer (50 mM Tris-HCl, 1 mM EDTA, 2 mM MgCl 2 and 2 mM dithiothreitol, pH 7.6) then centrifuged at 16 000g for 20 min at 4°C. Supernatants were removed and stored at −80°C until assayed.
For CS activity, 50 μl of supernatant was added to 1.9 ml of Tris-HCl buffer (50 mM Tris-HCl and 0.1 mM 5,5dithio-bis-(2-nitrobenzoic acid) [DTNB or Ellman's reagent], pH. 8.1) containing 0.1 mM acetyl CoA and allowed to rest for 5 min to achieve a stable baseline absorbance. The reaction was initiated by the addition of 50 μl of 20 mM oxaloacetate. We recorded the increase in absorbance at 412 nm (P300 Nanophotometer) over 5 min.
To quantify LDH activity, a 10 μl sample of 1:10 diluted supernatant (in 50 mM Tris-HCl buffer, pH 7.6) was added to 1.94 ml of Tris-HCl buffer (50 mM Tris-HCl, pH 7.6) containing 0.12 mM NADH. The reaction was initiated by the addition of 50 μl of 80 mM pyruvate, and the decrease in absorbance at 340 nm was recorded.
The CS and LDH activity (in micromoles per minute per gram of tissue) was calculated using the slope of the absorbance change immediately after addition of the oxaloacetate or pyruvate. We used a molar absorption coefficient of 14.15 ml μmol −1 cm −1 for reduced DTNB at 412 nm, and 6.22 ml μmol −1 cm −1 for NADH at 340 nm. All assays were performed in 1.5 ml methacrylate cuvettes at 25°C.

Quantitative real-time reverse transcriptase-polymerase chain reaction
Quantitative real-time RT-PCR (qRT-PCR) was performed in accordance with the guidelines of Bustin et al. (2009).
We also used qRT-PCR to quantify relative mRNA levels of several metabolism-associated genes, including the mitochondrial ATPase6/8 (atpase6/8) and cytochrome oxidase subunit-2 (cox2), and the nuclear genes citrate synthase (cs), L-lactate dehydrogenase A chain (ldhA), B chain (ldhB) and C chain (ldhC), cytochrome c oxidase subunit-5b (cox5b), nuclear respiratory factor 1 (nrf1) and peroxisome proliferator-activated receptors α (pparα) and δ (pparδ). In addition, we quantified heptic mRNA levels for the orphan nuclear receptors estrogen-related receptor α (esrra) and estrogen-related receptor γ (esrrg), which regulate mitochondrial gene expression and biogenesis, and gluconeogenesis, respectively (Villena et al., 2007;Kim et al., 2012). SYBR Green primers for atpase6/8 and cox2 were designed using the complete mitogenomes of the Death Valley pupfishes C. n. amargosae (KU883631) and C. diabolis (KX061747; Lema et al., 2016), and primers for transcripts encoding all other proteins were designed to coding region sequences obtained from the C. n. pectoralis genome (GCA_000776015). When possible, primers were designed to span an intron boundary. All primers were synthesized by Eurofins MWG Operon (Huntsville, AL, USA). Primer sequences and corresponding Genbank accession numbers are provided in Supplementary material Table S1. The specificity of each primer set was confirmed by cloning (TOPO ® TA Cloning, Life Technologies) and Sanger sequencing select PCR products (Molecular Cloning Laboratories, South San Francisco, CA, USA).
For all tissues, total RNA was extracted using Tri-Reagent (Molecular Research Center, Inc.) with bromochloropropane as the phase separation reagent. The resulting RNA was DNase I treated (TURBO DNA-free Kit; Ambion) and quantified using a P300 NanoPhotometer (Implen, Inc.; 260:280 ratios >1.96). For the evaluation of mRNA levels in wild pupfish, total RNA from the liver was reverse transcribed in 16 μl reaction volumes containing 8 μl of total RNA (15 ng/μl), 0.8 μl dNTPs (Promega), 0.8 μl random primers (500 μg/ml; Promega), 0.0625 μl recombinant RNasin ® ribonuclease inhibitor (40 u/μl; Promega), 0.1375 μl nuclease-free H 2 O, and 3.2 μl 5× buffer, 2.4 μl MgCl 2 and 0.6 μl GoScript™ reverse transcriptase (Promega) using a thermal profile of 25°C for 5 min and 42°C for 1 h, followed by 70°C for 15 min to inactivate the reverse transcriptase. For examination of relative mRNA levels in the liver and muscle of pupfish acclimated to conditions of 24 or 34°C, reverse transcription reactions were run with identical reagent composition as described above but scaled proportionally for a 45 μl reaction volume.
All qRT-PCR reactions were conducted at 16 μl, containing 8.0 μl iQ Universal SYBR Green Supermix (Bio-Rad, Hercules, CA, USA), 0.8 μl each of forward and reverse primers (10 μM), 4.9 μl nuclease-free water (Sigma, St Louis, MO, USA) and 1.5 μl of reverse-transcribed cDNA template. Assays were run on a 7300 Real-Time PCR System (Applied Biosystems, Inc.) using a thermal profile of 50°C for 2 min, 95°C for 10 min, and 42 cycles of 95°C for 15 s and 59°C for 1 min, followed by a melt curve analysis.
Standard curves were made for each tissue from a pool of RNA from samples representing all factor categories (e.g. population, temperature, hormone treatment) and sexes. Each standard was serially diluted and assayed in triplicate. DNA contamination was assessed by analysing RNA samples that were not reverse transcribed. Each qPCR run also included samples without cDNA as a further control. The PCR efficiencies for each gene were calculated as percentage efficiency = [10( 1/slope ) − 1] × 100, and are provided in Supplementary material Table S1. Correlation coefficients (r 2 ) were >0.98 for the standard curve for all genes. Relative expression levels of elongation factor 1α (EU906930; Lema et al., 2010) and 60S ribosomal protein L8 (rpl8; KJ719257; Lema et al., 2015) were quantified as internal control genes. In the liver of wild pupfish, only rpl8 mRNA levels were stable across populations and sexes, and data were normalized to rpl8 from the same tissue prior to being expressed as a relative level. For pupfish acclimated to 24 or 34°C in captivity, relative expression levels of each gene were normalized to the geometric mean of ef-1α and rpl8 mRNA levels, which did not vary with population, temperature and hormone treatment.

Statistical analyses
Relative gene transcript abundance data from the liver of wild pupfish were log 10 (x + 1) or sqrt(x + 0.5) transformed to conform to parametric assumptions. Population differences in plasma T 4 and mRNA abundance were assessed using Student's unpaired t-tests (α = 0.05).
Plasma T 3 concentrations in captive-held fish were log 10 transformed and compared using a three-factor ANOVA model with population origin, temperature condition and hormone treatment as factors. Tukey's HSD tests were then run to determine pairwise differences. Relative gene transcript abundance data from the liver and skeletal muscle were log 10 (x + 1) transformed to meet parametric assumptions when needed. Prior studies have found sex differences

5
in TH signalling-associated gene expression in the liver of fish (e.g. Lema et al., 2009;Johnson and Lema, 2011); for that reason, sex was included as a factor in the statistical analyses. Four-factor ANOVA models with population origin, temperature condition, hormone treatment and sex as factors and all interactions were used to examine sources of variation for the expression level of each gene transcript. For liver deiodinase mRNA levels, planned pairwise comparisons were calculated to test for temperature or T 3 exposure effects within the same sex and population only. As sex differences were rarely observed for all other TH signalling-and metabolismassociated transcripts, planned comparisons for those mRNAs were calculated using data from both sexes combined. To reduce type I errors, only statistical differences at α = 0.025 are reported from these pairwise comparisons, and categorical levels of statistical significance are provided to allow evaluation of the relative importance of each effector.
Metabolic enzyme activity data were likewise log 10 (x + 1) transformed and then analysed using ANCOVA models with population origin, temperature condition, hormone treatment and sex as main effect factors, and body size (SL, in millimetres) as a covariate. The ANCOVA models included all interaction terms. Outliers in the data were defined as values greater than four standard deviations from the mean of that individual's treatment group. Such outliers included two individuals for lactate dehydrogenase activity in the liver and a single individual for citrate synthase in the liver; those outliers were excluded from ANCOVAs. Only planned pairwise comparisons (t-tests) compliant to a significance value of α = 0.025 are reported. All tests were two tailed, and data are plotted as mean ± SEM values. All analyses were completed using JMP ® 11.2.0 software (SAS Institute, Inc.).

Results
Thyroid hormone signalling variation between wild populations Plasma T 4 concentrations were similar in adults from the two wild populations [P = 0.629; 19.7 ± 3.1 ng/ml (mean ± SEM) in Amargosa River pupfish and 22.9 ± 5.5 ng/ml in Tecopa Bore pupfish]. Replication after pooling plasma samples from individuals with low volumes was insufficient for sex-specific comparisons between populations.

Temperature and T 3 influence hepatic gene transcript abundance
The relative abundance of mRNAs encoding deiodinase enzymes in the liver was altered by exogenous T 3 , but the magnitude and direction of those effects varied with population origin and thermal experience. Transcripts encoding dio1 in the liver were lower in T 3 -treated fish ( Fig. 2a; hormone: F 1,104 = 19.228, P < 0.0001). Significant declines in hepatic dio2 transcript abundance were also observed in T 3treated males and females from both populations (Fig. 2b). In addition to these T 3 -induced changes, males from the Amargosa River population acclimated to both 24 and 34°C exhibited higher hepatic dio2 mRNA levels compared with males from Tecopa Bore (population × sex × hormone: F 1,104 = 5.524, P = 0.0207). This population difference in liver dio2 mRNAs was observed in males but not in females. Males from Tecopa Bore had higher hepatic dio3 transcripts compared with males from the Amargosa River, and T 3 increased hepatic dio3 mRNAs >13.5-fold in both populations at 34 but not at 24°C (Fig. 2c; population × temperature × hormone: F 1,104 = 5.592, P = 0.0199).

Effects on muscle gene transcript abundance
The relative abundance of dio2 mRNAs in skeletal muscle was higher in pupfish from Tecopa Bore than in conspecifics from the Amargosa River ( Fig. 4b; population: F 1,103 = 27.233, P < 0.0001). Transcript abundance for dio2 was not affected by either temperature or T 3 in either population. Muscle dio3 mRNA levels were higher in T 3 -treated fish ( Fig. 4c; hormone: F 1,103 = 38.709, P < 0.0001). Both populations also exhibited greater muscle dio3 transcript abundance at 34 compared with 24°C (temperature: F 1,103 = 4.838, P = 0.0301). Relative deiodinase dio1 transcript abundance was not affected by temperature, population origin or hormone treatment (Fig. 4a).
3,5,3′-Triiodothryonine increased the relative abundance of transcripts encoding trαA in both populations and sexes at 24 but not at 34°C (Fig. 4d; temperature × hormone: (a) Treatment of fish with 3,5,3′-triiodothryonine (T 3 ) reduced the relative level of mRNAs encoding dio1 in liver (ANOVA model, P < 0.0001), although this effect of T 3 was apparent only in females in pairwise comparisons (α = 0.025). (b) 3,5,3′-Triiodothryonine depressed liver dio2 mRNA abundance in males and females from both populations. Male pupfish from the Amargosa River also had higher dio2 mRNA levels than males from Tecopa Bore. (c) 3,5,3′-Triiodothryonine-induced elevation of hepatic dio3 transcripts was temperature dependent. In both sexes, T 3 treatment increased dio3 mRNAs at 34 but not at 24°C. Male pupfish from Tecopa Bore not treated with T 3 also had higher hepatic dio3 mRNA levels compared with Amargosa River males. Lines denote significant pairwise differences between groups (planned Student's unpaired t-tests: *P < 0.025, **P < 0.01, ***P < 0.001); n = 6-10 fish per category.  Muscle trαB mRNAs were also higher in Tecopa Bore pupfish compared with Amargosa River pupfish ( Fig. 4e; population: F 1,103 = 14.145, P = 0.0003) and were more abundant at 24 than at 34°C in both populations (temperature: F 1,103 = 12,154, P = 0.0007). Muscle trβ mRNAs in both populations and at both acclimation temperatures were elevated~45% by T 3 treatment ( Fig. 4f; hormone: F 1,103 = 31.972, P < 0.0001) and were also higher at both acclimation temperatures in Tecopa Bore pupfish compared with the Amargosa River population (population: F 1,103 = 13.787, P = 0.0003). Transcripts for mct8 and oatp1c1 were not quantified in skeletal muscle because prior studies in fish indicate very low mRNA abundance for these solute carriers in this tissue (Muzzio et al., 2014).

Metabolic enzyme activity in skeletal muscle
The activity of CS in muscle was reduced at 34 compared with 24°C ( Fig. 5a; population × temperature × SL interaction: F 1,101 = 5.850, P = 0.0174). Supplementation with T 3 had no influence on CS activity at either temperature.
Exogenous T 3 , however, depressed muscle LDH activity in pupfish from Tecopa Bore but not from the Amargosa River ( Fig. 5b; population × hormone: F 1,101 = 5.101, P = 0.0261). Hepatic CS and LDH activity is provided in the Supplementary material (Fig. S1).

Metabolic gene expression in skeletal muscle
Transcripts encoding the mitochondrial genes atpase6/8 and coxII were at higher relative abundance in the Amargosa River population compared with Tecopa Bore (Fig. 6a and b; population: atpase6/8, F 1,103 = 4.766, P = 0.0313; coxII, F 1,103 = 5.745, P = 0.0183). Both mitogenome transcripts were also increased by T 3 treatment (atpase6/8, F 1,103 = 9.647, P = 0.0025; coxII, F 1,103 = 14.095, P = 0.0003) and at reduced relative levels at 34 compared with 24°C (atpase6/8, F 1,103 = 6.473, P = 0.0124; coxII, F 1,103 = 21.145, P < 0.0001). Relative cs mRNA levels in muscle increased in response to T 3 ( Fig. 6c; hormone: F 1,103 = 15.658, P = 0.0001). Transcript abundance for cs was lower in Tecopa Bore fish at 34 compared with 24°C; however, this temperature effect was not observed in the Amargosa River population (population × temperature: Figure 3: Influences of population, temperature and T 3 on hepatic thyroid hormone (TH) receptors and membrane transporters. (a) 3,5,3′-Triiodothryonine reduced the abundance of liver mRNAs encoding TH receptors trαA at 24°C, but increased trαA mRNAs at 34°C. (b) Liver trβ mRNA levels were affected by temperature in opposing directions, depending on population origin, and were also reduced in fish treated with T 3 . Transcript abundance for the TH plasma membrane transporters mct8 (c) and oatp1c1 (d) was reduced in fish at 34 compared with 24°C. 3,5,3′-Triiodothryonine reduced liver mct8 mRNA levels in the Amargosa River population, but not the Tecopa Bore population, at 24 but not 34°C. Data from males and females are shown combined. Sample sizes are n = 14-17 fish per group. Lines indicate significant pairwise differences between groups (planned Student's unpaired. t-tests: *P < 0.025, **P < 0.01).

Discussion
Many native fishes of North America's southwestern deserts have protected 'endangered' status, and the small population sizes of these species continue to make them vulnerable to a variety of anthropogenic impacts Moyle et al., 2011). Climate models predict that increasingly warm and arid conditions in the North American southwest will intensify the region's already extreme temperatures (Seager et al., 2007;Seager and Vecchi, 2010). Such temperature changes are likely to impact the region's native fishes given that many of these taxa occur only in a few small, isolated habitats and have limited ability to shift their geographical range (e.g. Hausner et al., 2014). Assessing patterns of local population variation in thermal physiology will therefore be crucial for identifying populations most likely to persist under a changing climate (Chown et al., 2010;Hoffman and Sgrò, 2011).
Here, we provide evidence for variation in TH physiology in two wild, allopatric populations of C. n. amargosae pupfish. Specifically, we found that male C. n. amargosae pupfish sampled directly from the Amargosa River habitat exhibited 5-fold greater liver dio2 mRNA levels compared with males in Tecopa Bore, whereas liver dio3 mRNAs werẽ 12.5-fold greater in abundance in Tecopa Bore males. These differences in deiodinase mRNA levels were observed without any evidence for variation in circulating T 4 between the populations, but were paralleled by~2-fold higher hepatic mRNA levels for the TH receptor trβ in Amargosa River males. As temperature can influence TH pathways in fish (Leatherland et al., 1977(Leatherland et al., , 1980Eales et al., 1982), and the temperature of Tecopa Bore is, on average, higher than that of the Amargosa River, we acclimated pupfish from these populations to stable conditions of 24 or 34°C to test whether the observed variation in TH pathway-associated mRNA levels between the wild populations arises from dissimilar thermal experience. Although we did not detect any effects of acclimation temperature on endogenous T 3 concentrations in either population, males from these populations retained differences in liver dio2 and dio3 and, to a lesser extent, trβ gene transcript abundance even after nearly 3 months of similar thermal experience. Tecopa Bore pupfish also had higher mRNAs levels for dio2 and trβ in skeletal muscle than Amargosa River fish, and this population variation was independent of thermal acclimation.
Dio2 has ORD activity and removes iodine from the 5′ outer-ring site to convert T 4 to the more active form, T 3 , whereas Dio3 acts as an inner-ring deiodinase and removes iodine from the inner ring of T 4 and T 3 to convert these hormones to reverse triiodothyronine (rT 3 ) and diiodothyronine (T 2 ), respectively (Köhrl, 2000;Orozco and Valverde, 2005;St Germain et al., 2009;Little, 2016). The differences in hepatic and muscle dio2 and dio3 mRNA levels between the Amargosa River and Tecopa Bore populations therefore implies variation in TH conversion. Given the multiple post-transcriptional regulatory steps between gene transcription and the production of functional proteins, variation in transcript levels does not necessarily translate into differences in enzyme activity (e.g. Maier et al., 2009;Ghazalpour et al., 2011;Suarez and Moyes, 2012). As such, it is difficult to interpret definitively the functional implications of population-level variation in the dio mRNA levels observed between the Amargosa River and Tecopa Bore populations. Even so, recent studies point to TH action being strongly regulated by local variation in the expression of deiodinases and TH transporters at the target tissue (Schweizer et al., 2008;Darras et al., 2015). If the mRNA abundance variation observed here associates positively with Dio activity, Tecopa Bore pupfish may have increased TH inactivation by Dio3 in the liver, whereas thẽ 2-fold higher dio2 mRNA levels in skeletal muscle may indicate elevated conversion of T 4 to T 3 in that tissue. Figure 5: Citrate synthase (CS) and lactate dehydrogenase (LDH) activity in skeletal muscle. (a) Muscle aerobic CS activity was reduced in fish acclimated to 34°C and was unaffected by T 3 at either acclimation temperature. (b) Lactate dehydrogenase activity in muscle was higher in pupfish from the Tecopa Bore population compared with fish from the Amargosa River. 3,5,3′-Triiodothryonine reduced LDH activity in Tecopa Bore pupfish acclimated to both 24 and 34°C but had no effect in Amargosa River pupfish at either temperature. Sample sizes are n = 6-10 fish per group. Lines indicate significant differences (planned Student's unpaired t-tests: P < 0.025, **P < 0.01, ***P < 0.001).

10
Thyroid hormones mediate changes in gene expression via the action of nuclear TH receptors, which bind to thyroid response elements (TREs) to regulate transcription (Yen et al., 2006). A wide variety of genes have been found to be regulated by THs (e.g. Chatonnet et al., 2015;Gil-Ibañez et al., 2015), although perhaps the best-studied TH targets are genes encoding proteins involved in TH signalling pathways and metabolism. For instance, previous studies in fishes have demonstrated up-regulation of hepatic dio3 mRNAs and down-regulation of dio2 mRNAs in response to exogenous T 3 (García-G et al., 2004;Bres et al., 2006;Lema et al., 2009;Johnson and Lema, 2011). Our data here confirm these general patterns of T 3 induction of dio3 and downregulation of dio2 in the liver, suggesting that the changes in mRNA levels observed here may correlate with similar variation in Dio activity. Moreover, our data provide evidence for T 3 regulation of dio3, but not dio2, in skeletal muscle. Thyroid hormone receptor genes of mammals and amphibians contain TREs and are autoinduced by T 3 (Yen et al., 2006), and indications are that trαA (or trα) and trβ transcription may be autoinduced in teleost fishes as well (Kawakami et al., 2006;Nelson and Habibi, 2006;Johnson and Lema, 2011;Nelson et al., 2011). Our findings here provide further support for tr autoinduction by T 3 , with transcripts encoding trαA, trαB and trβ in muscle detected at higher levels in T 3 -treated pupfish.
To our knowledge, the present findings provide the first evidence in fish for temperature effects on expression of the TH transporters Mct8 and Oapt1c1. The Mct8 and Oatp1c1 solute carriers facilitate TH transport across the cellular plasma membrane (Arjona et al., 2011;Visser et al., 2011), and changes in expression of these carriers have the potential to influence the transport rates of T 4 and T 3 into target cells. 3,5,3′-Triiodothryonine has also been observed to regulate hepatic gene expression of mct8 and oatp1c1 in the fathead minnow, Pimepheles promelas (Noyes et al., 2013;Muzzio et al., 2014). In the present study, T 3 down-regulated liver mct8 mRNA levels while concurrently up-regulating oatp1c1 mRNAs in pupfish acclimated to 24°C. In zebrafish, Mct8 transport of THs is temperature dependent, transporting only T 3 at 26°C, but both T 3 and T 4 at 37°C (Arjona et al., 2011). Although no data on the transport selectivity of Mct8 or Oatp1c1 are available for other fishes, the temperaturedependent TH selectivity observed by Arjona et al. (2011) combined with our finding of temperature effects on mct8 and oatp1c1 mRNA levels in pupfish suggests that hepatic T 3 and T 4 uptake or release may vary with temperature. , ldhC (f), atpaseA (g), atpaseB (h), cox5b (i), nrf-1 (j), pparα (k) and pparδ (l). Note that scales of y-axis relative mRNA levels differ among transcripts. Sample sizes are n = 14-17 fish per group. Lines indicate pairwise differences between groups (planned Student's unpaired t-tests: *P < 0.025, **P < 0.01, ***P < 0.001).
Pupfish acclimated to 34°C likewise showed altered metabolic enzyme activity and metabolism-associated mRNA levels in skeletal muscle and liver (see Supplementary material) compared with fish at 24°C. Muscle aerobic CS activity and cs mRNA levels were lower in pupfish at 34 than at 24°C, suggesting decreased oxidative metabolic capacity at 34°C. In contrast, LDH activity in skeletal muscle was not affected by thermal acclimation but was greater in pupfish from the Tecopa Bore population compared with fish from the Amargosa River population. Interestingly, T 3 depressed muscle LDH activity in Tecopa Bore fish but not Amargosa River fish, suggesting population differences in the regulation of anaerobic metabolism by TH.
The differential pattern of LDH subunit mRNA expression observed in these populations suggests that these LDH activity differences may be related to variation in LDH isoform expression. Lactate dehydrogenase is a tetramer, with isoforms composed of combinations of three LDH subunits (A-chain, B-chain and C-chain), each of which is encoded by a separate gene (Quattro et al., 1993). Subunit composition generates variation in kinetic properties and thermal stability among LDH isoforms (e.g. Place and Powers, 1984;Henry and Ferguson, 1985), and evolutionary changes in LDH-and, in particular, A 4 -LDH, the muscle isoform of LDH in vertebrates-have been linked to taxonomic variation in anaerobic thermal scope in fishes (e.g. Fields et al., 2015). Here, we observed that T 3 -mediated mRNA regulation of ldhA in skeletal muscle differed between the populations, with T 3 up-regulating ldhA mRNAs in Tecopa Bore pupfish but not in Amargosa River pupfish. 3,5,3′-Triiodothryonine also up-regulated ldhB and ldhC transcripts, but in a similar pattern in both populations. Thus, the divergent responses of muscle LDH activity to T 3 that we observed in the Amargosa River and Tecopa Bore populations might be a result of population-level variation in T 3 regulation of ldhA transcription. Tecopa Bore has higher average temperatures and lower dissolved O 2 (~1.8 mg/l), so that pupfish occupying this habitat are likely to have a greater reliance on anaerobic metabolism. Little et al. (2013) observed T 3 -induction-as well as T 2 -inhibition-of muscle LDH activity in zebrafish pharmacologically made hypothyroid by propylthiouracil exposure, suggesting that TH regulation of anaerobic metabolic activity might occur broadly in fishes. At the same time, these same T 3 -treated zebrafish also showed increased mRNA abundance for several mitochondrial protein subunits, including atpase6/8, cox2, atpaseB and coxv2b, suggesting that T 3 might concomitantly enhance LDH activity and aerobic respiration.
The reduced muscle CS activity observed in both populations at 34°C also suggests a reduced reliance on aerobic metabolism by pupfish at elevated temperature. Given that rising temperatures lower water O 2 content, this decline in CS activity is likely to enable better survival in lower dissolved O 2 conditions. Why we did not observe a corresponding increase in LDH activity at 34°C is less clear, however, because fish exposed to elevated temperatures often increase anaerobic metabolic activity. Recent findings by Heuton et al. (2015) may point to a possible answer, because these authors proposed that pupfish (C. n. mionectes × C. diabolis hybrids) may not rely heavily on lactate production when acclimated to high temperature (33°C), but instead enter into a state of 'paradoxical anaerobism' wherein fish rely on ethanol production. In the face of observations by Heuton et al. (2015), future studies should examine how alcohol dehydrogenase activity varies in pupfish acclimated to varying temperature and dissolved O 2 conditions. The patterns of mRNA changes that we observed in the skeletal muscle provide additional evidence for T 3 induction of aerobic metabolic pathways. 3,5,3′-Triiodothryonine upregulated mRNAs encoding the mitochondrial genes atpase6/ 8 and cox2 in pupfish muscle, suggesting T 3 induction of mitogenome transcription and mitochrondrial activity in fish, as has been observed in mammals and birds (Pillar and Seitz, 1997;Psarra et al., 2006;Collin et al., 2009;Psarra and Sekeris, 2013). We also detected T 3 induction of mRNAs encoding cox5b, nrf-1, pparα and pparδ in muscle, implying a general up-regulation of aerobic metabolic gene expression. For instance, T 3 treatment increased the relative levels of muscle nrf-1 mRNAs in both pupfish populations, although this effect was more pronounced at 24 than at 34°C. In the rat liver, TH stimulation of mitochondrial biogenesis and respiratory chain activity is mediated, in part, via T 3 induction of nrf-1 and nrf-2 transcription (Rodríquez-Peña et al., 2002). Temperature-and T 3 -associated variation in expression of the peroxisome proliferator-activated receptors (PPARs) pparα and pparδ in pupfish muscle again points to TH-mediated changes in energy homeostasis. The PPARs are ligandactivated transcription factors, and the activation of PPARα and PPARδ has been linked to changes in mitochondrial and peroxisome fatty acid metabolism (Tyagi et al., 2011).
It is crucial to point out, however, that the temperatureassociated variation in T 3 -induced mRNA expression observed here might alternatively be related to differential circulating T 3 concentrations in T 3 -treated fish at 34 compared with 24°C, rather than to variation in transcriptional responses to T 3 at these two temperatures. We used a waterborne T 3 treatment method instead of an injection, implantation or oral dosing method to avoid effects of acute stress from netting and handling fish and to minimize dosing variation that would likely have occurred if T 3 was administered orally (e.g. via food) over the short 18-24 h period. The waterborne exposure, however, raised circulating T 3 several-fold higher than the endogenous concentrations observed in control pupfish, and this pharmacological dose might have induced effects different from what would have occurred with a physiological dose. Additionally, temperature affected the uptake rate of T 3 from the water, and comparisons of data from the T 3 -treated pupfish at 24 and 34°C should be interpreted cautiously. Although we are unaware of any study in fish that has examined temperature-dependent kinetics of TH uptake from water, temperature can influence rates of chemical uptake from water in fishes (McKim and Erikson, 1991). For instance, waterborne estrogen uptake across the gill is positively associated with oxygen consumption rates (Blewett et al., 2013a,b), which increase at elevated temperatures.
Even so, the variation in gene transcript levels observed between the Amargosa River and Tecopa Bore populations in the absence of T 3 treatment suggests that these populations vary in patterns of gene expression associated with TH pathways and energy metabolism. The Tecopa Bore population of C. n. amargosae was founded by individuals from the Amargosa River <50 years ago, so any underlying genetic bases to these TH pathway and metabolic differences could represent contemporary evolution resulting from either drift or selection. Rapid rates of evolution are also not unprecedented in pupfishes, and evolutionary divergence in thermal scope has already been documented in C. nevadensis populations (Hirshfield et al., 1980). Recent studies in pupfishes have also documented contemporary evolution of body shape (Collyer et al., 2007(Collyer et al., , 2011, and two adaptive radiations of pupfishes have been linked to exceptionally high rates of evolution in trophic morphology (Martin and Wainwright, 2011). In addition, population variation in TH physiology has been observed in other fishes and may be an important physiological mechanism underlying local population differentiation in energy metabolism or morphology (e.g. Kitano et al., 2010;Kitano and Lema, 2013;Levin and Bolotovskiy, 2015). Based on the differences in hepatic and muscle dio and tr expression observed here between the Amargosa River and Tecopa Bore C. n. amargosae populations, it is possible that variation in TH signaling, even if only at the level of some target tissues (e.g., Lema and Kitano, 2013), could contribute developmentally to the morphological variation between these populations.

Conclusions
Conservation and management strategies for native desert fishes commonly aim to protected the species' native habitat, which often consists of a single, small freshwater spring or marsh (Deacon and Minckley, 1991). Even for species where these habitats are currently protected, a changing climate may lead to altered temperatures and nutrient cycling in patterns that impact the energetics, growth and reproduction of native desert fishes (e.g. Hausner et al., 2014). For other desert fishes, habitat loss resulting from groundwater extraction or flow impediments, such as dams, has made it necessary to translocate fish to unoccupied habitats to establish additional populations or to breed fish in captivity for reintroduction into the wild (Johnson and Jensen, 1991;Minckley et al., 1991). Our finding here that endocrine and metabolic regulation by temperature differs in recently isolated (<50 years) populations of C. n. amargosae pupfish points to the importance of considering both contemporary population adaptation and thermal experience in management decisions about hatchery rearing conditions, reintroduction site selection and even fish translocation procedures in on-going efforts to conserve North America's native desert fishes.

Supplementary material
Supplementary material is available at Conservation Physiology online.