Lipoprotein Lipase Overexpression in Skeletal Muscle Attenuates Weight Regain by Potentiating Energy Expenditure

Currently, obesity afflicts 40% of adults and 19% of youth in the U.S. (1). Obesity is associated with numerous diseases, such as diabetes (2), cardiovascular disease (3), and cancer (4), and obesity-associated disease is estimated to cost ∼$200 billion a year in the U.S. (5,6). Weight loss reduces the disease burden associated with obesity in a dose-dependent manner (7); however, a majority of those with obesity are unable to maintain weight loss for over a year (8). Weight regain is partially driven by biological adaptations established during weight loss maintenance (WLM) that increases appetite and suppresses energy expenditure (9,10). Therefore, in order to successfully treat those afflicted with obesity, strategies need to be developed that maintain a reduced energy intake and/or increase energy expenditure.

In preclinical models, hyperphagia promotes rapid weight regain (11,12). However, even when limiting caloric intake to that of non–weight-reduced controls, weight gain is more rapid in rats that were subjected to caloric restriction, indicating that a suppressed energy expenditure contributes to weight regain (13). Prior research implicates a reduced energetic cost of storing nutrients as a culprit for suppressing energy expenditure and accelerating weight regain (9,14–18). During early weight regain, residual adaptations established during WLM may combine with the acute effects of overfeeding to suppress fat oxidation and increase carbohydrate oxidation, facilitating the efficient storage of lipids into adipose depots (13,14,18–20). However, a direct role for nutrient partitioning in thermogenesis during weight regain has yet to be established.

Lipoprotein lipase (LPL) controls the rate of triglyceride-derived fatty acid uptake by peripheral tissues, like adipose and skeletal muscle (21). LPL activity likely influences whole-body metabolism in a tissue-specific manner, as greater skeletal muscle LPL activity is associated with greater whole-body fat oxidation (22) and resistance to weight gain (23,24), and, in contrast, greater adipose LPL activity is associated with greater weight gain (23,24). WLM decreases LPL activity in skeletal muscle (25) and increases LPL activity adipose tissue (26–28), which may promote efficient weight regain by trafficking lipid away from oxidation and toward storage instead. Therefore, to determine the role of skeletal muscle lipid uptake and oxidation on efficiency of weight regain, we assessed nutrient partitioning and energy expenditure in diet-induced obese mice that overexpress LPL in skeletal muscle during WLM and weight regain.

Tissues came from a previous study (11), in which rats were acquired from Charles River Laboratories (Charles River Laboratories, Wilmington, MA) and individually housed at 22–24°C with a 12:12-h light/dark cycle. After rats were acclimatized to the facility for 2 weeks on a low-fat diet (LFD) (12% kcal fat; #11724; Research Diets, New Brunswick, NJ), rats were provided ad libitum access to high-fat diet (HFD) (46% kcal fat) for 1 week (#12344; Research Diets) and ranked into tertiles based on their rate of weight gain (11,29). The top tertile was designated as obesity prone and selected for the study. Obesity-prone rats were then split into obese, WLM, and regain groups. Obese rats were switched back to an ad libitum LFD for 42 weeks. WLM and regain rats were provided ad libitum HFD for 17 weeks and subsequently calorically restricted on an LFD to induce and maintain ∼10% weight loss for 17 weeks. Food for WLM was provided once per day, ∼0.5–1.0 h before the dark cycle. At this time, a subset of WLM rats was euthanized to provide tissues for the WLM group; the other subset of WLM rats was allowed ad libitum access to LFD for 8 weeks to promote weight regain and were then euthanized and their tissues harvested (Fig. 1A).

Mice that overexpress LPL specifically in skeletal muscle (mCK-hLPL) were rederived from a previously established line (30). Wild-type (WT) littermates were used as controls. Ten-week-old mCK-hLPL and WT mice were individually housed at 27°C and kept on a 14:10-h light/dark cycle. To induce obesity, mice were provided a high-fat high-sugar diet (#D15031601; Research Diets) (40% kcal from fat) for 8 weeks. Mice were then calorically restricted to ∼50% of ad libitum energy intake with a medium-fat diet (MFD) (#D07091301; Research Diets) (24.9% kcal from fat) to induce ∼20% weight loss, which was then maintained for 2 weeks by titrating feeding. Food for WLM was provided once per day, ∼0.5–1.0 h before the dark cycle. At the end of the 2-week WLM phase, mice were provided ad libitum access to food. On the final day of study, mice were anesthetized using isoflurane and skeletal muscle, adipose, and liver tissues were harvested and flash frozen. Mice were fasted for 2 h prior to sacrifice. All procedures were approved by the Institutional Animal Care and Use Committee at University of Colorado Anschutz Medical Campus.

Indirect calorimetry was performed using a metabolic monitoring system (Oxymax CLAMS-8M; Columbus Instruments, Columbus, OH). Measurements were acquired on the final day of the obese and WLM phase and throughout the relapse phase. Activity was determined by infrared beam breaks. Resting energy expenditure was calculated by averaging the lowest three rates of energy expenditure in a day when beam breaks for ambulatory activity were less than one. This measure of resting energy expenditure may be influenced by the thermic effect of food during the obese phase and throughout relapse, especially during the 1st day of relapse when the continuous measures of respiratory exchange ratios (RERs) suggest continuous eating throughout the day (Supplementary Fig. 3I and Supplementary Fig. 4G).

Total lean mass and fat mass were determined using quantitative magnetic resonance (EchoMR; Echo Medical Systems, Houston, TX).

Gastrocnemius skeletal muscle was excised from the hindlimb of obesity-prone male Wistar rats that had either been allowed ad libitum access to diet or calorically restricted to induce WLM. Mitochondrial isolations were performed at 0–4°C as previously described (22). Resting (state 4) and maximal coupled (state 3) mitochondrial oxygen consumption were measured in a respiration chamber maintained at 37°C (Strathkelvin Instruments, North Lanarkshire, U.K.). Incubations were carried out at 37°C in a 0.5-mL final volume containing 100 mmol/L KCl, 50 mmol/L MOPS, 10 mmol/L K₂PO₄, 10 mmol/L MgCl₂, 0.5 mmol/L EGTA, 20 mmol/L glucose, and 0.2% BSA, pH 7.4. Mitochondrial respiration was monitored at the following concentrations: 1 mmol/L malate, 10 mmol/L glutamate, 1 mmol/L pyruvate, 10 μmol/L L-palmitoylcarnitine, and 10 mmol/L succinate. Mitochondria and substrates were added, and the coupled maximal respiration rate was initiated with the addition of ADP (100 μmol/L). Respiration was normalized to protein concentration of mitochondrial isolates using the bicinchoninic acid assay.

Sensitivity to metabolic signals was assessed by adapting the creatine kinase energy clamp (31) in a cohort of weight-stable chow-fed mice (12 weeks old). Medial gastrocnemius fibers were extracted and permeabilized in buffer X (Supplementary Table 1) with 30 μg/mL of saponin. Buffer Z (Supplementary Table 2) with 5 mmol/L of creatine monohydrate, 20 units/mL of creatine phosphokinase, 5 mmol/L of ATP, and 0.01 mmol/L of blebbistatin was used as respiration media. The 5 mmol/L pyruvate plus 0.5 mmol/L malate was used for pyruvate assays; 25 μmol/L of palmitoyl-L-carnitine plus 0.5 mmol/L malate was used for palmitate assays. Tris-PCr was titrated in to control ADP/ATP ratios, and free energy (ΔG_ATP) was calculated as previously described (31). High-resolution O₂ consumption measurements were conducted in a 2-mL well at 37°C using the Oroboros Oxygraph-2K (Oroboros Instruments, Innsbruck, Austria). Respiration was normalized to dry tissue weight.

To trace dietary fatty acid trafficking and oxidation, 1-[¹⁴C]palmitate and 1-[¹⁴C]oleate tracers were blended into the diet (MFD) at a 1:3 ratio, which reflected their ratio in the diet, at a specific activity of 0.94 μCi/g of food. To measure oxidation of dietary fat, flow rate was individually measured for each metabolic cage, and effluent CO₂ from each cage was collected in 3.0-mL aliquots of a 22:1 mixture of methanol to methylbenzethonium hydroxide (#B2156; Sigma Chemical). The ¹⁴C content of these samples was then measured with a Beckman Coulter LS6500 scintillation counter. Lipids were extracted from tissues, and ¹⁴C content, which reflects dietary fat content, was determined as previously described (14).

RNA from rat gastrocnemius was isolated using the E-Z nucleic acid kit according to the manufacturer’s instructions (Omega Bio-Tek, Norcross, GA). RNA samples were hybridized to an Affymetrix GeneChip Rat Genome 2.0 array (Affymetrix, Santa Clara, CA), washed, stained, and imaged with the Affymetrix GeneAtlas Personal Microarray System. Raw data were normalized using the MAS5 algorithm.

Pulverized mouse gastrocnemius was homogenized in QIAzol, and RNA was subsequently isolated using the RNeasy Mini Kit according to the manufacturer’s instructions (Qiagen, Germantown, MD). Total RNA quality was determined using Agilent 4200 TapeStation (Agilent Technologies, Santa Clara, CA). A total of 100 ng of total RNA was used as input to construct mRNA libraries from the NuGEN Universal Plus mRNA-Seq protocol part number 9133 (NuGEN, Redwood City, CA). Sequencing was done on an Illumina NovaSeq 6000 instrument using an S4 flow cell and 2 × 150 paired end sequencing (Illumina, San Diego, CA). Generated reads were mapped to genes using Salmon (Mouse Release M23) (32). Genes were removed if their counts were <10.

All data were analyzed either in the R programming environment (version 4.0.3) (33) or using Prism 9 (GraphPad Software, La Jolla, CA). Differentially regulated pathways were assessed using the Generally Applicable Gene-set Enrichment (GAGE) package (34). Microarray P values were determined using the Linear Models for Microarrays package (35). Differential expression was determined using the DESeq2 package in R. When comparing two groups, P values were assessed using two-tailed t tests. When comparing more than two groups, P values were assessed using two-way ANOVA with planned comparisons using Fisher least significant difference test. Planned comparisons were between mCK-hLPL and WT mice. Significance was set at P < 0.05. When applicable, data are expressed as means ± SE.

RNA-sequencing data are stored in the Gene Expression Omnibus database, accession number GSE154943.

Obesity-prone male Wistar rats were subjected to WLM and weight regain (Fig. 1A). Pathway analysis from microarrays performed on skeletal muscle revealed that WLM reduced genes in the Kyoto Encyclopedia of Genes and Genomes (KEGG) fatty acid metabolism and peroxisome proliferator–activated receptor (PPAR) signaling pathways compared with the obese condition (Fig. 1B and C and Supplementary Fig. 1A). Moreover, isolated skeletal muscle mitochondria from rats subjected to WLM had reduced capacity to oxidize lipids compared with obese rats (Fig. 1D). Interestingly, after weight was regained, transcriptome analysis on skeletal muscle revealed a downregulation of genes in the KEGG fatty acid metabolism and PPAR signaling pathway compared with the obese condition (Fig. 1E and F and Supplementary Fig. 1B). Altogether, these results indicate that pathways involved in skeletal muscle fatty acid metabolism are reduced during WLM and after weight regain.

Since genes involved in fatty acid metabolism and PPAR signaling were reduced in rats during WLM and after weight regain, and prior research suggests that LPL can regulate genes involved in fatty acid metabolism in skeletal muscle by tuning PPAR activation (36,37), we examined the effect of skeletal muscle LPL overexpression (mCK-hLPL) on metabolic outcomes in mice during WLM and weight regain. Before performing WLM and weight regain studies, we first assessed the effect of mCK-hLPL on carbohydrate and fat metabolism. We found that skeletal muscle from mCK-hLPL mice oxidized pyruvate to a similar degree as controls (Fig. 2A and C) but oxidized palmitate to a greater degree as cellular energetic demand increased (Fig. 2B and C). mCK-hLPL also prolonged time to fatigue with an endurance exercise challenge (Supplementary Fig. 2).

We then calorically restricted diet-induced obese WT and mCK-hLPL mice, maintained weight loss (WLM), and allowed ad libitum access to food to promote relapse to the obese state (relapse group in Fig. 3A). Skeletal muscle LPL overexpression attenuated weight regain, which was primarily due to less fat mass accumulation, as lean mass accrual did not differ (Fig. 3B–D). Absolute whole body, lean, and fat mass were similar between mCK-hLPL and WT mice throughout all phases (Supplementary Fig. 3A–C). Compared with WT, mCK-hLPL mice had greater daily energy expenditures on days 2 and 3 of relapse, which were primarily due to increases in metabolic rate during the light cycle and in resting energy expenditure (Fig. 3E–G). Throughout a relapse, average daily energy expenditure tended to be higher in mCK-hLPL mice, while average resting energy expenditures were significantly higher (Supplementary Fig. 3D–F). Levels of activity between WT and mCK-hLPL mice were similar throughout all phases (Fig. 3H and I and Supplementary Fig. 3G and H). mCK-hLPL mice had lower RER on day 2 of relapse, lower RERs during the light and dark cycle on day 2 and the light cycle on day 7, and, on average, lower RERs throughout relapse (Fig. 3J and K and Supplementary Fig. 3I and J). Energy intake did not differ between groups during relapse (Fig. 3L and Supplementary Fig. 3K), yet mCK-hLPL mice exhibited lower energy balance on days 2 and 7 and, on average, a lower energy balance throughout the relapse (Fig. 3M and Supplementary Fig. 3L). In all, these results indicate that skeletal muscle LPL overexpression attenuates weight regain by increasing energy expenditure.

Since mCK-hLPL potentiated energy expenditure the most during early relapse—day 1 to day 3 (Fig. 3E)—we examined dietary fat trafficking during the 1st day of relapse in a separate cohort of mice. In this cohort, mCK-hLPL mice had greater energy expenditure, metabolic rate during the light cycle, and resting energy expenditure on the 1st day of relapse (Fig. 4A and B and Supplementary Fig. 4A and B). Food intake and levels of activity were similar between mCK-hLPL and WT mice (Fig. 4C and Supplementary Fig. 4C–F). mCK-hLPL failed to influence RER during WLM and early relapse (Supplementary Fig. 4G and H), yet mCK-hLPL affected rates of dietary fat oxidation throughout the day. During WLM, mCK-hLPL had higher rates of dietary fat oxidation early during the dark cycle and lower rates of dietary fat oxidation early during the light cycle; during relapse, mCK-hLPL had higher rates of dietary fat oxidation late during the light cycle (Fig. 4D). During relapse, mCK-hLPL tended to increase total daily dietary fat oxidized (Fig. 4E) and, similar to metabolic rate (Fig. 4B), significantly increased rate of dietary fat oxidation during the light cycle (Fig. 4F). Interestingly, two-way ANOVA revealed that relapse reduced absolute dietary fat oxidation during the dark cycle and relative dietary fat oxidation during both the dark and light cycle, irrespective of genotype (P < 0.05) (Fig. 4E and Supplementary Fig. 4I), providing evidence for a reduced capacity to oxidize fat at the initiation of relapse.

We also examined retention of dietary fat during relapse and found that mCK-hLPL tended to decrease storage of dietary fat in retroperitoneal (RP) adipose tissue and significantly decreased storage of dietary fats in visceral and subcutaneous adipose depots (RP and inguinal white adipose tissue combined) (Fig. 4G). Body weights and composition, tissue weights, and serum profiles were similar between WT and mCK-hLPL mice (Supplementary Fig. 4J–R). In summary, these results indicate that dietary fat oxidation is suppressed during early relapse, and mCK-hLPL facilitates the trafficking of dietary fat toward oxidation rather than storage.

To better understand the genetic regulation of nutrient trafficking during relapse, we performed transcriptome analysis on skeletal muscle. Principal component analysis revealed a major effect of relapse on the transcriptome compared with WLM (Supplementary Fig. 5A). Pathway analysis comparing mCK-hLPL to WT mice during WLM and relapse revealed that mCK-hLPL differentially regulates 19 KEGG pathways during WLM and 16 KEGG pathways during relapse; 14 of these pathways were commonly regulated by mCK-hLPL during both WLM and relapse (Fig. 5A). Interestingly, the PPAR signaling pathway, which was reduced during WLM and weight regain in rats (Fig. 1B and E), was regulated by mCK-hLPL during WLM and relapse (Fig. 5A) and by the transition from WLM to relapse (Supplementary Fig. 5B). Since PPARδ has been shown to regulate genes involved in carbohydrate and fat metabolism in skeletal muscle (38), we mined the RNA-sequencing data set for genes in these metabolic pathways. Comparing WLM and relapsing WT mice to determine the effect of relapse on the skeletal muscle transcriptome independent of genotype, we found that relapse generally reduced genes involved in fatty acid uptake, breakdown, and oxidation and upregulated genes involved in lipid storage and glucose uptake. Comparing WT and mCK-hLPL mice during relapse, we found that skeletal muscle from mice that overexpress LPL generally has greater expression of genes involved in fatty acid uptake and oxidation and lower expression of genes involved in glucose uptake. In summary, these data indicate that relapse induces robust changes in the skeletal muscle transcriptome, which may favor carbohydrate uptake and oxidation and suppress fat uptake and oxidation. Conversely, compared with WT mice during relapse, mCK-hLPL mice have a transcriptional profile that favors fat uptake and oxidation over glucose uptake.

The major findings from these studies are that skeletal muscle fat metabolism is reduced during both WLM and early weight regain and skeletal muscle LPL overexpression attenuates weight regain by potentiating energy expenditure. In our initial rat studies, we found that WLM reduced skeletal muscle’s ability to oxidize lipids, and, even when weight was regained, genes required for fat oxidation were reduced in skeletal muscle. We then followed up on these findings using mice with skeletal muscle–specific LPL overexpression to improve skeletal muscle fatty acid metabolism and found that LPL overexpression increases whole-body fat oxidation during relapse and attenuates weight regain by increasing energy expenditure. On the 1st day of relapse, both genotypes reduced their total and relative dietary fat oxidized as compared with the WLM phase; however, relapsing mCK-hLPL mice trafficked more dietary fat toward oxidative pathways and less toward storage than relapsing WT mice. Moreover, compared with WLM WT mice, skeletal muscle from relapsing WT mice exhibited a general reduction in genes involved in fat metabolism and an increase in genes involved in carbohydrate metabolism. Conversely, during relapse and compared with WT mice, mCK-hLPL mice generally exhibited greater expression of genes involved in fat uptake and utilization and lower expression of genes involved in glucose uptake. In all, these results support a role for skeletal muscle lipid metabolism in regulating energy expenditure during weight regain.

Prior research indicates that whole-body fat oxidation is reduced during WLM and weight regain (39–43), yet the mechanisms driving this response are not well-characterized. At the cellular level, glucose and lipids compete for oxidation (44), and this regulation is orchestrated by hormones, degree of substrate uptake (i.e., substrate availability), and cytosolic and mitochondrial enzymatic activities (45). WLM has been shown to reduce fat uptake (39), and our results support a role for reduced fat uptake mediating the reduction in fat oxidation. In rats, we found that WLM reduced skeletal muscle mitochondrial fat oxidation compared with the obese condition, and, even when weight was fully regained, genes required for fatty acid metabolism were reduced. Increasing fat uptake into skeletal muscle via LPL overexpression increased expression of genes involved in fat uptake and oxidation during WLM and relapse, which coincided with an increase in postprandial dietary fat oxidation during WLM and an increase in fat oxidation throughout relapse. In skeletal muscle, lipids can act as ligands for PPARs to induce transcription of enzymes that promote fat oxidation (46). In rats, we found that the PPAR signaling pathway was reduced during WLM and after weight regain; meanwhile, in mice, increasing lipid flux into skeletal muscle via LPL overexpression increased genes in the PPAR signaling pathway during WLM and on the 1st day of relapse. In all, these results support a role for fat uptake, and possibly subsequent PPAR activation, in regulating fuel utilization during WLM and relapse.

Although we were unable to determine how an increase in skeletal muscle lipid metabolism promotes energy expenditure, it is plausible that the act of oxidizing fat itself is responsible. Fat oxidation requires more oxygen than carbohydrate to synthesize ATP (i.e., a lower P-to-O ratio) (47); therefore, at least compared with glucose oxidation, oxidation of fat expends more energy to maintain cellular ATP levels. Moreover, fat oxidation in skeletal muscle has been found to induce uncoupling (48,49), which would decrease efficiency of ATP synthesis to further increase energy expenditure. In addition to the possible role of fat oxidation in potentiating energy expenditure during weight regain, skeletal muscle has been found to contribute to thermogenesis via other mechanisms. For example, skeletal muscle can regulate energy expenditure via contraction efficiency (50) or substrate- and calcium-based futile cycles (51–53). Moreover, skeletal muscle may influence other organs to dissipate energy, as factors derived from skeletal muscle can promote a thermogenic profile in adipose tissue and increase energy expenditure (54). Future work is needed to determine how skeletal muscle LPL overexpression increases energy expenditure during weight regain.

The effect of skeletal muscle fat oxidation on weight gain during a moderate positive energy imbalance, as seen with ad libitum HFD feeding, is not well understood. When challenged with an HFD, improved skeletal muscle fat metabolism via skeletal muscle–specific PPARδ overexpression has been shown to attenuate weight gain (55); however, improving skeletal muscle metabolism with skeletal muscle–specific PPARγ coactivator-1α overexpression fails to influence body weight (56). Improving whole-body fat metabolism with global acetyl CoA carboxylase-2 knockdown also leads to conflicting results, with some groups finding an attenuation of weight gain (57,58) and others finding little to no effect (59,60). When provided an HFD, we found that mCK-hLPL mice gain a similar amount of weight as WT mice, which challenges the theory that improved skeletal muscle lipid metabolism prevents diet-induced obesity. Rather, skeletal muscle LPL overexpression uniquely attenuated weight gain during a relapse-induced positive energy imbalance. Of note, the attenuation in weight regain was due in part to an increase in energy expenditure during the first 3 days of a relapse, when mice were in the greatest positive energy imbalance. Therefore, the effect of fat oxidation on energy expenditure may be more noticeable during periods of extreme overfeeding, like early relapse or spontaneous overfeeding, which are characterized by increased carbohydrate oxidation and suppressed fat oxidation (14,61). These data may be important for obesity progression in humans, in whom it has been hypothesized that progression to obesity occurs in spurts, often around the holidays (62), which are characterized by bouts of extreme overfeeding (63). Thus, an improved capacity to oxidize fat in skeletal muscle may help prevent progression and/or relapse back to obesity by attenuating weight gain during periods of extreme overfeeding.

Increasing lipid influx or oxidation in skeletal muscle induces insulin resistance via accumulation of fatty acid derivatives (64), inhibition of glycolysis from metabolites (65), or increased production of reactive oxidative species (66). Although mCK-hLPL failed to have an effect on serum glucose or insulin during WLM or early regain, LPL overexpression in skeletal muscle has previously been shown to induce insulin resistance in other metabolic states (67,68), with prior research implicating fatty acid derivatives in LPL overexpression–induced insulin resistance (67). In this study, we find that LPL overexpression reduces transcription of genes that promote glucose uptake, suggesting an alternative process by which LPL overexpression induces insulin resistance. In support of this, recently it was shown that activation of PPARδ—the skeletal muscle–specific PPAR isoform—reduces transcription of genes involved in carbohydrate metabolism (38). Similar to providing a PPARδ agonist, we found that skeletal muscle LPL overexpression upregulated genes that promote fat metabolism and downregulated genes involved in carbohydrate metabolism. Therefore, lipid-induced insulin resistance may be transcriptionally regulated as well.

In conclusion, we found that skeletal muscle fatty acid metabolism is reduced during WLM and regain, which may promote efficient storage of fat into adipose depots during weight regain. LPL overexpression improves skeletal muscle fatty acid metabolism, traffics dietary fat away from storage and toward oxidation, and attenuates weight regain by potentiating energy expenditure. Therefore, therapies that improve skeletal muscle lipid metabolism may support WLM by attenuating weight regain when lapses in dieting occur.

Funding. This work was supported by National Institutes of Health National Institute on Aging grant U54 AG062319 and Eunice Kennedy Shriver National Institute of Child Health and Human Development grant P50 HD073063 (to P.S.M.), National Institutes of Health National Institute of Diabetes and Digestive and Kidney Diseases training grant T32 DK007260 (to D.M.P.), National Center for Advancing Translational Sciences fellowships TR001081 TL1 (to D.M.P.) and KL2 CDA (to V.D.S.), and National Institute of Diabetes and Digestive and Kidney Diseases grant K01 DK109079 (to M.C.R.). Support was also provided by the Colorado Nutrition Obesity Research Center (National Institute of Diabetes and Digestive and Kidney Diseases grant P30 DK48520).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. D.M.P. designed and performed experiments, J.A.Hi. designed and performed experiments and edited the manuscript. analyzed the data, and wrote and edited the manuscript. M.C.R. designed and performed experiments, analyzed the data, and edited the manuscript. V.D.S. designed and performed experiments and edited the manuscript. M.R.J. designed and performed experiments and edited the manuscript. R.M.F. designed and performed experiments and edited the manuscript. K.L.J. analyzed the data. J.A.Ho. performed experiments. G.C.J. performed experiments. P.D.N. designed experiments and edited the manuscripts. R.H.E. designed experiments and edited the manuscript. P.S.M. designed and performed experiments, analyzed the data, and wrote and edited the manuscript. P.S.M. and D.M.P. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.