Measuring complex phenotypes: A flexible high-throughput design for micro-respirometry

Variation in tissue-specific metabolism between species and among individuals is thought to be adaptively important; however, understanding this evolutionary relationship requires reliably measuring this trait in many individuals. In most higher organisms, tissue specificity is important because different organs (heart, brain, liver, muscle) have unique ecologically adaptive roles. Current technology and methodology for measuring tissue-specific metabolism is costly and limited by throughput capacity and efficiency. Presented here is the design for a flexible and cost-effective high-throughput micro-respirometer (HTMR) optimized to measure small biological samples. To verify precision and accuracy, substrate specific metabolism was measured in heart ventricles isolated from a small teleost, Fundulus heteroclitus, and in yeast (Saccharomyces cerevisiae). Within the system, results were reproducible between chambers and over time with both teleost hearts and yeast. Additionally, metabolic rates and allometric scaling relationships in Fundulus agree with previously published data measured with lower-throughput equipment. This design reduces cost, but still provides an accurate measure of metabolism in small biological samples. This will allow for high-throughput measurement of tissue metabolism that can enhance understanding of the adaptive importance of complex metabolic traits.


34
Understanding evolution and ecological adaptation can be enhanced by combining 35 genomics with quantitative analyses of complex phenotypic traits [1]. This integrative approach 36 requires sufficient sample size (i.e. 100s to 1000s) with precise measure of phenotypes, however 37 it can be challenging to obtain economical equipment for such high-throughput quantification. 38 To address this challenge, we present an inexpensive custom design to measure metabolism in 39 small biological samples such as cell suspensions, individual tissues or possibly small organisms. 40 Metabolism is a complex trait intricated in most physiological processes and is important to 41 organismal success. Thus, metabolism is ecologically and evolutionarily important [2][3][4][5][6][7][8][9][10][11]. The 42 effect of the environment on metabolism as well as tissue-specific variation can vary 43 considerably among individuals and populations [12][13][14][15][16][17][18]. These and other data suggest that 44 measuring metabolism can provide insights into the ecology and evolution of organisms [5]. 45 Metabolism is typically quantified via oxygen consumption rates (MO2). Numerous 46 systems to measure MO2 are available from companies including Unisense, PreSens, and Loligo, 47 but each has limitations with respect to technical design, throughput capacity, and cost. For small 48 biological samples, systems often have limited capacity (e.g. Oxygraph 2-K, OROBOS 49 INSTRUMENTS, Innsbruck, Austria) or require expensive reagents and disposables (e.g., 50 Seahorse XF Analyzers, Agilent, Santa Clara, CA). Therefore, these systems are not ideal for 51 high-throughput experimental designs as it becomes time consuming and expensive to measure 52 many samples. There is a need in the field for a simple design that can measure multiple sample 53 simultaneously at a reduced cost. Here we present a design for a high-throughput micro-54 respirometer (HTMR) that increases throughput of tissue-specific metabolism while minimizing 55 costs maintaining and maintaining efficacy. We validate the precision and accuracy of this 56 4 system by measuring both Saccharomyces cerevisiae and substrate specific metabolism in 57 Fundulus heteroclitus heart ventricles. 58

60
The HTMR consists of a custom external plexiglass water bath designed to enclose 1-ml 61 micro-respiration chambers (Unisense) (Fig 1A). The water bath is connected to a temperature-62 controlled, re-circulating system, and placed on a multi-place stir plate. Each chamber contains a 63 stir bar and nylon mesh screen for mixing media while keeping tissues suspended ( Fig 1B). 64 Exact chamber volumes were determined by measuring the mass (to 0.001 g) of water that 65 completely filled individual chambers with the mesh screen and stir bar. A fluorometric oxygen 66 sensor spot (PreSens) is adhered to the internal side of the chamber lid with a polymer optical 67 fiber cable affixed to each chamber lid for contactless oxygen measurement through the sensor 68 spot. All cables are connected to a 10-channel microfiber-optic oxygen meter (PreSens), which 69 uses PreSens Measurement Studio 2 software to collect oxygen data at a sampling rate of 20 70 measurements per minute. Sensors were calibrated at 0% (using 0.05 g sodium dithionite per 1 71 ml of media) and 100% air saturation (fully oxygenated media). Validation of the HTMR was 72 carried out using a four-chamber system; however, it can easily be extended to a 10-chamber 73 system as in Fig 1A. 74

81
Each ventricle is measured in each substrate for six minutes and then placed in the following substrate while 82 exchanging chamber media. Substrate conditions are measured as follows: 1) 5 mM glucose; 2) 1 mM palmitic acid;

86
Metabolic rate determinations 87 The precision and accuracy of the HTMR was validated by measuring MO2 in both yeast 88 (Saccharomyces cerevisiae) and teleost (F. heteroclitus) heart ventricles. MO2 is measured in the 89 sealed chambers by measuring oxygen concentration at a rate of 20 measurements per minute, 90 over a six-minute period. During each daily measurement, a minimum of three blank 91 measurements, during which only media was in the chamber, were run to determine any 92 background flux. For each six-minute measurement, the last three minutes (60 datapoints) were 93 used for calculating metabolic rate. To do so, oxygen concentration was regressed against time to 94 determine the raw oxygen consumption rate (pmol*µl-1 *min -1 ). Slopes were also calculated for 95 each blank measurement and averaged by chamber to quantify background flux, then subtracted 96 from each slope. Metabolic rate was measured as MO2 = (Msample -Mblank) * Vchamber * 1/60, 97 where MO2 is the final metabolic rate in pmol*s -1 , M is the slope of oxygen consumption per 98 sample in pmol*µl -1 *min -1 , and V is the volume of each chamber in µl. 99 PreSens datafiles provide data per sensor with oxygen concentration (µmol*L -1 ) at each 100 time point (minutes). An R markdown file detailing this analysis of the raw PreSens data files 101 can be found at https://github.com/ADeLiberto/FundulusGenomics.git. Desert Island, ME, and Deer Isle, ME. Individuals were trapped on public land, and no permit 105 was needed to catch these marine minnows for non-commercial purposes. All fish were common 106 gardened at 20°C and salinity of 15 ppt in re-circulating aquaria for at least five months and then 107 acclimated to 12°C or 28°C for at least two months prior to metabolic measurements. Fish were 108 randomly selected, weighed, and then sacrificed by cervical dislocation. Heart ventricles were 109 isolated and immediately placed in Ringer's media (1.5mM CaCl2, 10 mM Tris-HCl pH 7.5, 150 110 mM NaCl, 5mM KCl, 1.5mM MgSO4) supplemented with 5 mM glucose and 10 U/ml heparin to 111 expel blood. Media was incubated at the measurement temperature prior to use. Ventricles were 112 then splayed following precedent of previous cardiac metabolism measurements in F. 113 heteroclitus [19]. Splaying the hearts decreases variation and increases overall oxygen 114 consumption rates, as greater internal surface area is exposed to the substrate media [20]. After 115 splaying, hearts were not further stimulated, as mechanical disruption or homogenization can 116 increase variability in oxygen consumption rates [15]. All animal husbandry and experimental 117 procedures were approved through the University of Miami Institutional Animal Care and Use 118 Committee (Protocol # 19-045). 119

Methodological validation 120
In order to validate the HTMR performance, several parameters were tested: 1) net flux at 121 multiple oxygen concentrations, 2) between-chamber variability in MO2, and 3) consistency of 122 MO2 over time. To quantify net flux and confirm equal rates between chambers, flux was 123 measured at multiple oxygen concentrations in each chamber. Here we define net flux as both 7 background oxygen consumption and oxygen diffusion into the system. Flux at 100% air 125 saturation was measured with fully oxygenated Ringer's media. To measure net flux at lower 126 oxygen saturations, Ringer's media was deoxygenated to the desired level with nitrogen gas. 127 85% air saturation was chosen because cardiac MO2 measurements over the six minutes typically 128 deplete oxygen to approximately 92% of air saturation but do not exceed 85%. To determine net 129 flux, oxygen concentration was measured in each chamber for 10 minutes and repeated in 130 triplicate. 131 Biological repeatability between chambers was tested with yeast at 28°C. A cell 132 suspension was prepared using 1 g of yeast per 10 ml of Ringer's media supplemented with 5 133 mM glucose. In each chamber, 100 µl of the suspension was injected to account for variation in 134 chamber volume. Oxygen consumption was measured for 10 minutes in triplicate. MO2 was 135 calculated as above to confirm there were no differences among chambers. 136 In order to assess metabolic consistency over the time-course of the experiment as well as 137 chamber repeatability, hearts from four fish were isolated, and glucose metabolism was assayed 138 in each chamber at 28°C. Hearts were randomly assigned to one of the four chambers and cycled 139 through each of them, with media exchange between each measurement. Three blank 140 measurements were run at the conclusion of the experiment. MO2 was calculated as above and 141 then regressed against relative time of initial oxygen measurement per cycle to determine 142 metabolic rate consistency of heart tissue over time.

169
The HTMR is a simple custom design composed of a plexiglass water bath enclosing 170 micro-respiration chambers connected to a multi-channel oxygen meter (Fig 1A). For a 10-171 chamber system, the approximate cost per chamber is $1870, including the cost of the oxygen 172 meter and stir-plate. The full cost of the system is broken down in Table 1. The oxygen meter 173 itself represents the highest cost (~$14,000 for ten inputs). followed by 1 ml glass chambers with 174 lid containing two injection ports (~$300 each). Optical-fiber cables and sensor spots combined 175 are approximately $95 each. A multi-place stir-plate is also necessary (~$900). 176

193
To test biological repeatability among the chambers, yeast metabolism per chamber was 194 measured. Average MO2 was 12.502 ± 1.907 pmol*s -1 , and there were no significant differences 195 in metabolism between each of the chambers (ANOVA, p = 0.538; Fig 2C). This MO2 for yeast 196 is approximately 40-fold higher than the net flux at 85% air saturation. In addition to yeast 197 measurements, heart ventricles were measured across all four chambers over a 45-minute time 198 period to validate both repeatability among chambers and that ventricles can maintain consistent 199 metabolic activity over time. Among the four replicates, there was no significant difference in 200 metabolic rate when regressed against time (linear model, p = 0.657; Fig 3A). Additionally, there 201 were no significant differences in metabolic activity among the four chambers for each heart 202 (ANOVA, p = 0.363; Fig 3B). 203  (Fig 4). For substrate specific metabolism, data was analyzed separately for 212 individuals measured at 12°C or 28°C. There is a large inter-individual variation in substrate 213 specific metabolism, which reduces the statistical power to reject the null hypothesis of no 214 difference among substrates. To avoid this type II error, we apply paired t-tests that compare 215 substrates within each individual and use Bonferroni's test to correct for multiple tests [23]. 216 Glucose, FA, and LKA metabolism were significantly greater than endogenous (p=0.001, 217 Bonferroni's corrected p=0.006); except FA at 12°C (p=0.03; Bonferroni's corrected p=0.18; Fig  218   4A). Glucose metabolism was significantly greater than FA and LKA (Bonferroni's corrected 219 p=0.006) at both 12°C and 28°C. FA metabolism was significantly greater than LKA metabolism 220 at 28°C (Bonferroni's corrected p=0.006; Fig 4B) but not at 12°C (p=0.5; Fig 4A). 221

12
The log10 substrate specific MO2 from Maine and Massachusetts acclimated to 12°C and 228 28°C determined here can be compared to MO2 measured in Oleksiak et. al (2005) for Maine 229 individuals acclimated to 20°C using an ANCOVA with log10 body mass and temperature as 230 linear covariates. There were no significant differences (p = 0.55, 0.15, and 0.85 for glucose, FA 231 and LKA respectively) and the least squares fall within 5% of one another. 232

Allometric scaling of metabolism 233
Both body mass and heart ventricle mass were measured of each F. heteroclitus 234 individual measured at each temperature. The mean body mass of individuals measured at 12°C 235 and 28°C was 9.11 ± 2.87 g and 9.32 ± 2.90 g, respectively, and was not significantly different 236 between temperatures. Additionally, average ventricle masses were 0.013 ± 0.005 and 0.010 ± 237 0.004 for 12°C and 28°C, respectively and did not significantly differ between acclimation 238 temperatures. In F. heteroclitus body mass and heart mass are highly correlated (linear 239 regression at 12°C R 2 = 0.74, p<0.0001; for 28°C, R 2 = 0.66, p < 0.001) thus, body mass was used 240 to correct for variation due to mass between individuals, as done previously [24]. Body mass 241 explained a significant amount of the variation (30-70%) in metabolism among individuals for 242 all conditions (Fig 5). Variance explained by body mass (R 2 ), was higher at 12°C than at 28°C 243 (Fig 5, S1Table). For glucose MO2, allometric scaling was identical (to the 2 nd significant digit) 244 to previous determinations and nearly the same as in Jayasundara et al. (2015). Examining the 245 effect of temperature and substrates, allometric scaling coefficients (S1 Table), were between 246 0.65 to 1.29. While body mass contributed significantly to the variation between individuals, 247 there was no effect of sex on cardiac metabolism by linear regression at each substrate-248 temperature combination. A three-way ANOVA including substrate, body mass and sex showed 249 13 no significant differences between males and females in cardiac metabolism at 12°C or 28°C (p 250 = 0.0963 and p= 0.4143, respectively). 251 and at 28°C n=95. For full data on regression slopes, see S1 Table. 255

257
The HTMR provides a simple custom design for measuring small biological samples that 258 allows higher throughput measurements at lower costs. While cost is still not negligible, the probes were tested; however, they were more fragile and cumbersome to use. Temperature 281 control is also essential for consistent and repeatable measurements. Temperature has a large 282 impact on oxygen solubility; thus, precise temperature control is necessary and was closely 283 regulated and monitored during measurements. 284 Chamber mixing was another important factor to control. Without thorough mixing from 285 the stir-bars, oxygen measurements are inconsistent and inaccurate. This is an advantage of this 286 system design over other multi-well plate style oxygen readers in which the media is unstirred 287 during measurement, but instead rely on mixing prior to measurement. Continuous stirring 288 allows for longer measurement periods. The size of the mesh that holds the tissue above the stir 289 bar was also optimized: very small mesh inhibits mixing, but too large mesh would not separate 290 the tissue from the stir bar in the bottom of the chamber. Additionally, nylon mesh was used over 291 steel mesh, as it did not as readily retain air bubbles. 292 Finally, leak was tested extensively. At 85% air saturation, background flux (O2 use not 293 associated with biological sample) was small (< 1 pmol*s -1 ) compared to heart ventricle and 294 yeast MO2. To account for any amount of flux, blank measurements were taken throughout runs 295 and corrected for in each chamber, with no significant differences in leak among chambers. 296 Initially, the Unisense microinjection lids were chosen for flexibility; however, after completing 297 tests and measurements, the manufacturers released information that this particular model was 298 less airtight than other models. For future design construction, we recommend that researchers 299 use single-port lids with a sufficient path length to further minimize leak through diffusion. 300

301
The HTMR is sensitive to both substrates used to fuel heart ventricle metabolism and body 302 mass. HTMR determinations were very similar to previously published data (within 5%) [19]. 303 Additionally, substrate specific patterns, with highest rates supported by glucose, agree with 304 previous measurements in F. heteroclitus [19]. Metabolism was unaffected by the time course 305 for measuring the four substrate conditions. Ventricles continue to contract over the duration of 306 the experiment and show no significant decline in metabolic activity (Fig 2A.). Importantly, 307 body mass accounted for a significant amount of variation in these individuals following an 308 allometric scaling pattern, and the log mass against logMO2 linear regression has nearly identical 309 slopes to those determined by others [15,19]. These data suggest that the system is both precise 310 because the variation among samples did not obscure substrate or body mass effects and accurate 311 in that substrate specific metabolic rates are similar to previous measures [19] and have 312 allometric scaling coefficients very similar to published data [15]. 313

314
The HTMR was designed to measure metabolism in many individuals, because of the large 315 individual variation and adaptive significance of the trait. We were specifically interested in 316 cardiac metabolism in F. heteroclitus, as this species shows large inter-individual variance in 317 cardiac metabolism and the mRNAs associated with this variance [19], and these patterns may 318 hold true for many species. To better understand the physiological and evolutionary importance 319 of this variance requires many individuals which the HTMR allows. For example, in this study, 320 metabolism was quantified in approximately 200 ventricles in only 10 days. Although here the 321 system is tested only with F. heteroclitus, this system could easily be extended to study other 322 types of tissue-specific metabolism in many individuals. The decreased cost and efficiency of the 323 design can have countless applications, allowing for high-throughput measurement of tissue 324 metabolism that can enhance our understanding of the adaptive importance of these traits. 325