Formation of colorectal liver metastases induces musculoskeletal and metabolic abnormalities consistent with exacerbated cachexia

Advanced colorectal cancer (CRC) is often accompanied by development of liver metastases (LMs) and skeletal muscle (SkM) wasting, i.e. cachexia. Despite plaguing the majority of CRC patients, cachexia remains unresolved. By using mice subcutaneously (C26) or intrasplenically injected with C26 tumor cells to mimic hepatic dissemination of cancer cells (mC26), here we aimed to further characterize functional, molecular and metabolic effects on SkM and examine whether LMs exacerbate CRC-induced cachexia. C26-derived LMs were associated with progressive loss of body weight, as well as with significant reductions in SkM size and strength, in line with reduced phosphorylation of markers of protein anabolism and enhanced protein catabolism. mC26 hosts showed prevalence of fibers with glycolytic metabolism and enhanced lipid accumulation, consistent with abnormalities of mitochondrial homeostasis and energy metabolism. In a comparison with mice bearing subcutaneous C26, cachexia appeared exacerbated in the mC26 hosts, as also supported by differentially expressed pathways within SkM. Overall, our model recapitulates the cachectic phenotype of metastatic CRC and reveals that formation of LMs resulting from CRC exacerbate cancer-induced SkM wasting by promoting differential gene expression signatures. Research Cell biology Muscle biology


Introduction
By the end of 2020 about 148,000 new cases of colorectal cancer (CRC) are expected to be diagnosed in the United States, and over 53,000 people will ultimately die from the disease, thus representing the second leading cause of death among all cancers (1). Notably, in 70% of CRC cases the most serious impediment is the development of liver metastases (LMs), frequently accompanied by the onset of skeletal muscle wasting, i.e. cachexia (2), a condition which cannot be rescued by conventional nutritional support (3)(4)(5). Along with loss of muscle mass, cachectic cancer patients experience loss of muscle strength, as well as cardiac and respiratory failure, altogether contributing to functional impairment and inability to withstand anticancer treatments (6)(7)(8)(9)(10). Despite its known debilitating influence on patient outcomes and survival, cachexia remains an understudied field, mainly because of minimal progress on the identification of new pathogenic mechanisms and treatments.
A useful tool to interrogate mechanisms of disease and test therapeutic interventions to combat diseases is the use of small animal models. Unfortunately with respect to cancer cachexia, only a handful of mouse models are currently in use, and their poor molecular characterization represents a major limitation (11)(12)(13). Indeed, data on genetic and metabolic profiling are generally missing, with the exception of the Colon-26 (C26) allograft mouse model (14), which remains the most widely used model to study CRC cachexia (15). However, clinical relevance and translational capacity of the current circulating models used for the study of cachexia has been recently questioned. In fact, though cachexia can be detected already during the early stages of tumor progression, it is extremely prominent and highly correlated with advanced metastatic cancers (15), thus corroborating the need for preclinical models of metastatic cancer-induced cachexia. Such need was also recently and elegantly discussed by Tomasin et al., highlighting the fact that the field is still lacking quality, well-characterized and clinically translational models of metastatic cancer cachexia (15). In hopes of meeting this need and advancing the field of cachexia, we and others have employed novel models of CRC-associated LMs, thus taking advantage of instrasplenic injections, which allow for rapid migration of tumor cells to the liver (16)(17)(18)(19)(20).
In the present study we sought to thoroughly characterize the skeletal muscle in a model of colorectal LMs using the well-characterized C26 murine cell line. We also assessed the systemic metabolic effects of tumor hosts bearing LMs and examined whether the development of LMs could exacerbate the cachectic phenotype compared to the traditional C26 allograft model. Overall, we aimed at further establishing the importance of developing and utilizing preclinical 4 models of metastatic cancer to better understand cachexia induced by advanced CRC. Here we demonstrated that metastatic CRC induces dramatic loss of skeletal muscle mass and perturbs energy metabolism. Further, we provided evidence that colorectal LMs exacerbate loss of skeletal muscle mass and strength compared to subcutaneous C26 tumors, thereby highlighting the importance of using preclinical metastatic CRC models to better understand cachexia induced by advanced CRC.

Formation of colorectal LMs leads to loss of body weight, muscle mass, and muscle strength
To assess the impact of colorectal LMs on skeletal muscle mass in vivo, CD2F1 male mice were intrasplenically injected with 2.5 x 10 5 C26 tumor cells (mC26). This procedure resulted in formation of LMs following liver dissemination via the portal circulation, without generating tumors in extrahepatic organs, such as spleen or lungs.
Tumor injected and sham-operated animals were monitored daily for body weight over the course of the experiment.
Following tumor cell injection, mC26 hosts experienced progressive weight loss, which resulted in a 16% reduction compared to sham-operated animals (p < 0.0001; Figure 1A-B). mC26 hosts saw a non-significant increase in liver size (+21%) compared to sham-operated animals, which can likely be attributed to the localization of C26 tumors within the liver ( Figure 1C-E). The loss of body weight was accompanied by wasting in several skeletal muscles including the gastrocnemius (-26%, p < 0.01), tibialis anterior (-29%, p < 0.01), and quadriceps (-33%, p < 0.01) ( Figure 2A). The loss of skeletal muscle mass in the mC26 hosts was paralleled by a 25% decline in whole body grip strength (p < 0.01; Figure 2B), as well as muscle atrophy, as indicated by reduced tibialis anterior CSA (-22%, p < 0.05) ( Figure 2C).

mC2disrupts skeletal muscle mitochondrial homeostasis
We have recently demonstrated that cachexia, as induced by either cancer or chemotherapy, is accompanied by reductions in various mitochondrial proteins required for fusion and biogenesis (23,24). Therefore, we sought to determine whether disruptions of mitochondrial homeostasis were detected in the mC26 model of cachexia. Here we demonstrate reductions of Mitofusin-2 (-27%, p < 0.01), PGC1α (-15%, p < 0.05) and PGC1β (-44%, p < 0.01) levels in the skeletal muscle of mC26 mice compared to Sham animals ( Figure 5A). Meanwhile, we did not witness alterations in protein levels of Cytochrome-C, OPA1, VDAC, Cox IV, or Fis1, or in gene expression levels of Pink1 and Parkin2 ( Figure 5A). In line with the reduction seen in mitochondrial proteins, we assessed and determined reduced enzymatic activity of both pyruvate dehydrogenase (PDH; -87%, p < 0.05) and succinate dehydrogenase (SDH; -81%, p < 0.05) ( Figure 5B) in mC26 skeletal muscle. This was further complimented by SDH staining of the tibialis anterior muscle, which revealed a shift from oxidative to glycolytic metabolism (+14% glycolytic fibers) in mC26 hosts ( Figure 5B). To further examine possible disruption of skeletal muscle energy metabolism we performed Oil Red O staining and found robust increases in intramuscular fat accumulation (Integrated density: +99%, p < 0.01; % Area: +127%, p < 0.001) of mC26 skeletal muscle ( Figure 5C).

Formation of colorectal LMs appears to exacerbate skeletal muscle wasting
Since we observed exacerbated bone loss in the mC26 hosts, not previously identified in allograft C26 tumor hosts, we sought to examine whether formation of colorectal LMs was also responsible for worsened skeletal muscle wasting.

mC26 and C26 hosts have differentially expressed signaling networks
As we observed an exacerbated cachectic phenotype in mC26 hosts, we investigated the divergence in gene expression signatures and its associated pathways within skeletal muscle of metastatic and non-metastatic CRC. Next-Generation RNA-seq analysis revealed a large population of commonly shared genes (> 60%), between mC26 and C26, when compared with their respective controls. However, 1227 and 1494 differentially expressed genes were found to be unique to C26 and mC26, respectively ( Figure 11A-B). The differentially expressed genes were then used to run pathway and upstream regulator analysis, which revealed both common and differentially altered signaling pathways within C26 and mC26 hosts, as well as distinct upstream regulators ( Figure 11 C-D). Interestingly, one of the only altered pathways similar between C26 and mC26 was calcium signaling, revealing that formation of LMs is distinctly altering signaling within skeletal muscle. Identified upstream regulators were also distinct between the two groups, with TLR2 and SOCS1 identified in C26 skeletal muscle and NOS1 and STAT1 identified in mC26 skeletal muscle.

Discussion
According to recent statistics, CRC represents the third most prevalent cancer in the United States and worldwide, and with an estimated 1 in 25 lifetime probability of developing CRC, it remains a major health concern (1). Cachexia, a devastating comorbidity in several types of cancer, including CRC (2), directly contributes to over 20% of cancerrelated deaths and is often worsened by anticancer drugs (8-10, 24, 29, 30). We and others have discussed the importance of maintaining lean body mass to improve treatment tolerance and survival outcomes in cancer patients (31). Despite spanning several decades, research efforts have yielded minimal progress towards a cure for cachexia, likely also due to the limited availability of pre-clinical animal models. Indeed, it was recently discussed that minimal change in the use of clinically relevant and translational animal models for the study of cachexia has occurred over the past 10 years (15). This is particularly true when translating current animal models to metastatic cancer cachexia, especially with respect to CRC, in which the C26-bearing mouse remains the most widely used, published and characterized model (15). Given that cachexia is typically observed in the more advanced CRC patients, usually burdened by LMs, it is clear that better and thoroughly characterized models of metastatic CRC cachexia are needed not only to better understand the disease at a mechanistic level, but also in hopes of counteracting the loss of lean mass in order to improve survival (2). In this paper we sought to narrow this pressing gap in the literature by characterizing a metastatic model of CRC cachexia.
In order to mimic the cachectic phenotype of metastatic CRC cachexia, we employed an intrasplenic injection approach to disseminate LMs in vivo by using murine C26 colorectal tumor cells. This particular approach to induce LMs is an accepted state of the art model commonly used in cancer biology and has emerged as a growing area of interest to study cachexia. Though this particular approach has been used previously to demonstrate muscle wasting associated with LMs (16,18), minimal genetic, metabolic, or molecular alterations of skeletal muscle associated with metastatic CRC have been investigated. In the present study we demonstrated that C26 LMs induce progressive loss of body weight (Figure 1), which was accompanied by severe atrophy of several skeletal muscles and a progressive decline in strength ( Figure 2). Moreover, we demonstrated that the skeletal muscle wasting was accompanied by disruptions in skeletal muscle anabolism, catabolism, mitochondrial homeostasis, and energy metabolism.
We and others have implicated STAT3 in cancer-induced muscle wasting in several tumor models, including the C26 allograft and Apc min/+ CRC models, as well as the Lewis Lung Carcinoma (LLC), B16 melanoma, and ES-2 ovarian cancer models (14, 21,23,[32][33][34][35][36][37][38]. In line with these findings, here we showed a significant increase in STAT3 phosphorylation in mC26 hosts (Figure 3), showing that STAT3 may be a critical prognosticator of cancer-induced muscle atrophy in the presence of C26 metastatic tumors. In line with elevated STAT3 signaling, we also demonstrated heightened protein catabolism, as indicated by increased protein ubiquitination, as well as by elevated expression of the E3 ubiquitin ligases Atrogin-1, MuRF-1, and Fbxo31 (Figure 4), which have been previously reported in association with cachectic muscle, including cachexia associated with LM (18,23,(39)(40)(41). Interestingly, we did not observe alterations in the activation of several other proteins previously described to play a role in cancer cachexia, including ERK (Figure 3), previously showed increased in the skeletal muscle of mice bearing C26 allografts (22).
On the other hand, p38 and AKT phosphorylation were found unchanged (22) partially sustain skeletal muscle mass in CRC (43,44). In the present study we showed that mC26 hosts have reduced levels of mitochondrial proteins, including Mitofusin-2, PGC1α and PGC1β ( Figure 5), in line with our previously published findings in mice bearing C26 allografts, whereas Cytochrome-C and OPA1 were unchanged (29). Consistent with such impaired mitochondrial homeostasis, we also observed marked reductions in SDH activity. This is also in line with previous findings from ours and other groups showing that cachectic animals bearing cancers or receiving chemotherapy had reduced muscle SDH activity (18,23,24).
Given the perturbed skeletal muscle mitochondria we wanted to assess whether mC26 hosts also experienced impaired energy metabolism, both within the skeletal muscle and systemically. We have recently shown disruptions in skeletal muscle and systemic energy metabolism in animals bearing CRC and chemotherapy (25, 45). Our present findings indicate an increased systemic demand in glucose metabolism as reflected by reduced plasma and skeletal muscle glucose in mC26 hosts (Figures 6 & 7). This is also in line with our previous observations showing reduced plasma glucose in the C26 allograft model (25). Also similar to our previous study, here we showed a reduction in circulating BCAAs, consistent with increased muscle catabolism and subsequent oxidation. Increases in BCAA oxidation have been observed in whole body and skeletal muscle in conditions including sepsis, trauma and after endotoxin or tumor necrosis factor treatment (46)(47)(48). Interestingly, in mC26 skeletal muscle we showed elevated isoleucine and valine, consistent with aggressive catabolism of skeletal muscle proteins. Interestingly, TCA intermediates succinate and fumarate are suppressed within mC26 skeletal muscle, implying reduced TCA flux. This is in line with reduced SDH and PDH enzyme activity ( Figure 6B), also suggesting impaired oxidative metabolism. The impaired oxidative environment of skeletal muscle is consistent with a shift to glycolysis as the dominant energy producing pathway ( Figure 6A).
As the metabolome of both skeletal muscle and plasma of mC26 hosts revealed drastic impairments, we wanted to assess whether the liver was showing similar changes, especially given its robust metabolic flexibility in times of energy stress. Indeed, here we showed that the high systemic demand for glucose demonstrated by plasma and skeletal muscle was also reflected by drastic reductions in liver glucose ( Figure 8) and glycogen ( Figure S1). These alterations mimic prior data showing reduced liver glucose and glycogen in C26 hosts (25). The liver of mC26 hosts presented significant reductions in the levels of alanine and lactate which is consistent with their use in gluconeogenesis ( Figure   8) (25). A dramatic increase in the ketone body 3-hydroxybutyrate (+714%) is consistent with an increase in gluconeogenesis. Under these conditions, the oxaloacetate typically used to condense with Acetyl-CoA to form citrate and feed the TCA cycle, is reduced to malate and continues on the gluconeogenic pathway. The consequent backup of acetyl-CoA from fatty acid β-oxidation thus leads to the formation of 3-hydroxybutyrate. A decrease of NAD+ was observed which is consistent with its consumption in β-oxidation. The TCA cycle intermediates malate and fumarate were markedly increased in the mC26 hosts along with the anapleurotic substrate glutamate suggesting an increase in TCA cycle activity. Significantly increased TCA flux was not clearly evident in the liver of C26 bearers, indicating that tumor infiltration of the liver may induce greater energy perturbations (25).
In line with the need for improved models of metastatic cancer cachexia to further our understanding of the disease, there is also a necessity for better characterization of multiple organs that are negatively affected by different types of tumors and chemotherapeutics (49). Indeed, examining this organ crosstalk within cachexia may provide important clues to teasing out the mechanisms that drive this morose disease. Our group and others have demonstrated that cancer-induced cachexia does not solely affect skeletal muscle, but that heart, fat, and bone are also impaired (23,26,28,50). In particular, recent observations have implicated that abnormal muscle-bone crosstalk may play a significant role in cancer cachexia. For example observations generated in several mouse models of CRC, including C26, HT-29, and Apc min/+ , revealed differential bone loss, despite consistent loss of skeletal muscle mass (28). Interestingly, cancellous bone in the femurs of C26 tumor hosts was generally maintained, whereas in the present study the use of C26 tumor cells to induce LMs was sufficient to drive both skeletal muscle and bone loss (Figure 9), further indicating an exacerbation in bone phenotype with LMs. Our data also suggest that, since LMs often occur in advanced CRC patients, examination of bone mineral density should be warranted, especially considering that heightened bone loss can further contribute to muscle loss and weakness (50).
Given the finding that bone loss occurred in mC26 hosts, compared to a normal bone phenotype in mice bearing subcutaneous C26 tumors, we sought to understand whether skeletal muscle loss was also exacerbated in metastatic tumor bearing hosts. In a follow-up experiment involving a direct comparison among mice bearing C26 allograft and mice carrying mC26 tumors, we demonstrated greater losses in body weight, skeletal muscle mass, cardiac size, as well as whole body grip strength in the presence of LMs, with the quadriceps muscle 20% smaller in mC26 compared C26 tumor hosts (Figure 10; S2). This exacerbated wasting occurred in the animals with CRC LMs, despite the fact that the mC26 hosts initially received fewer tumor cells compared to the C26 bearers (2.5 x 10 5 vs. 1.0 x 10 6 , respectively). This is important to note as it implicates the site of injection as a critical prognosticator for the development of cachexia, as also previously demonstrated in work by Chiba et al. (51). Nonetheless, that study did not take into exam the impact of LMs. To understand the pathways that may potentially lead to this exacerbation, we performed RNA sequencing on quadriceps muscles. We found that over 60% of genes altered in mC26 and C26 hosts were similar, including known prognosticators of muscle wasting such as STAT3, MurRF-1, Atrogin-1, FBXO31, and PDK4 (14, 21, 39-41, 52). However, the fold-change elevation of these genes tended to be generally greater in mC26 hosts (STAT3: 5.6 vs. 4.6; MuRF-1: 25 vs. 20; Atrogin-1: 14 vs. 13.9; FBX031: 6 vs. 6.2; PDK4: 8.7 vs. 5.7), which may in part explain the worsened muscle wasting. Follow-up pathway analysis performed using the differentially expressed genes identified several differentially regulated pathways as well as differential upstream regulators within mC26 and C26 tumor hosts ( Figure 11D-E,). Several of the modulated pathways including calcium, sirtuin, STAT3, PTEN, and oxidative phosphorylation, have been implicated in muscle wasting diseases (14, 21, 23, 25, 53-55). The identified C26 regulator TLR2 has previously shown to mediate myotube atrophy, although in the present context TLR2 was found to be down-regulated (56). We also identified STAT1 as an upstream regulator within mC26 skeletal muscle. Interestingly, STAT1 has recently been implicated as a regulator of autophagy, a degradation process known to be upregulated in cachectic cancer patients and animal models of cachexia (57)(58)(59). Future studies will interrogate these pathways to better understand the mechanisms by which formation of LMs differentially alter skeletal muscle signaling.
Overall, our study clearly demonstrates that formation of LMs induces, and even aggravates, muscle atrophy induced by CRC. Though we examined and identified differential signaling networks within skeletal muscle of tumor hosts bearing LMs, investigation into how CRC metastases alter the liver endocrine function and how this may ultimately influence skeletal muscle wasting was not explored in the current study. Additionally, tumor burden in the mC26 model was crudely assessed using only histological analysis, thus possibly representing a limitation of our approach and preventing us from performing a direct comparison with tumor size in the C26 hosts. Given the search for more translationally relevant models, another limitation of the present study is the shortened nature in the development of cachexia. Though the goal of the study to examine cachexia in a context of CRC LM was achieved, future studies may consider using lower doses of C26 tumor cells or perhaps other CRC cells that may allow the progression of cachexia to extend beyond 2 weeks. Moreover, in the present study we did not take into consideration whether the administration of chemotherapeutics further aggravate muscle wasting, especially considering that our lab and others have demonstrated that several anticancer compounds induce cachexia independent of their effects on tumor growth (24,26,27,29,30,60). Lastly, another limitation of the current study was the focus on male animals, leaving out possible sex differences in response to LMs. As sexual dimorphism has been identified in other models of CRC, future studies should examine how LMs influence skeletal muscle in males vs. females (33).
In conclusion, we have demonstrated that formation of C26 CRC LMs induces robust skeletal muscle atrophy. Skeletal muscle atrophy in mC26 hosts was accompanied by elevated protein catabolism, disrupted mitochondrial homeostasis, and perturbed skeletal muscle metabolism. Formation of C26 LMs also leads to systemic alterations in energy metabolism and impaired bone homeostasis. Moreover, formation of C26 LMs aggravates cachexia and induces 14 differential gene expression within skeletal muscle compared CRC cachexia induced by C26 subcutaneous tumors.
Overall, our study provides support for the use of in vivo metastatic models for the study of cancer cachexia.

Cell lines
Prior to conducting animal work, Colon-26 (C26) cells, provided by Donna McCarthy (Ohio State University), were cultured in DMEM medium supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, and 1% sodium pyruvate and maintained in a 5% CO2, 37°C humidified incubator. C26 cells were cultured, passaged, and trypsinized when sub-confluent to be prepared for animal injection in sterile saline.
Briefly, 12-week-old CD2F1 male mice were placed under anesthesia and a side subcostal incision was made to carefully expose the spleen. Animals were then intrasplenically injected with 100µl of saline containing 2.5 x 10 5 C26 tumor cells (mC26), or saline alone (Sham) over the period of 1 minute, followed by 2 minutes of hemostasis (n = 5/group). The intrasplenic injection approach allows for tumor cells to quickly enter the portal circulation, thus infiltrating the liver, without forming tumors in the spleen or other common metastatic sites of CRC, such as the lung.
In a separate experiment 8-week old CD2F1 male mice were either intrasplenically (mC26; 2.5 x 10 5 ) or subcutaneously (C26; 1.0 x 10 6 ) injected with C26 cells (n = 4-6/group) (62). Non-tumor bearing (Con) and shamoperated (Sham) mice were used as controls. Mice were weighed daily, then euthanized under light isoflurane anesthesia. At the time of sacrifice, skeletal muscle tissues were harvested, weighed, then snap frozen in liquid nitrogen and stored at −80°C for further studies. The tibialis anterior muscles were frozen in liquid nitrogen-cooled isopentane for histology, as previously described (23). All mouse carcasses, including a portion of the liver were fixed for two days in 10% neutral buffered formalin and then transferred into 70% ethanol.

Whole-body Grip Strength Assessment
Whole-body grip strength was assessed using a commercially available automatic grip strength meter (Columbus Instruments, Columbus, OH, USA) as previously shown (63). The absolute force (expressed in grams) was recorded over five measurements, with the top three measurements utilized for analysis. To further avoid habituation bias, animals were only tested once a week during the experimental period.

Hematoxylin and eosin (H&E) Staining
To examine the formation of liver metastases, fixed liver tissue was paraffin embedded and sectioned (10 μm) in preparation for H&E staining (23). Stained liver sections were then observed under an Axio Observer.Z1 motorized microscope (Zeiss, Oberchoken, Germany) and 5x images were recorded for tumor infiltration assessment. Using ImageJ 1.43 software, images were assessed for the tumor area relative to liver area (expressed as a percentage).

Muscle Cross-sectional Area (CSA)
To assess skeletal muscle atrophy, 10 μm-thick cryosections of tibialis anterior muscles, taken at the mid-belly, were processed for immunostaining as described previously (30). Briefly, sections were blocked for one hour at room temperature and incubated overnight at 4˚C with a dystrophin primary antibody (Developmental Studies Hybridoma Bank, Iowa City, IA, USA; #MANDRA1(7A10)), followed by a one-hour secondary antibody (AlexaFluor 594 # A-11032; Thermo Fisher Scientific, Waltham, MA, USA) incubation at room temperature. Entire dystrophin stained sections were analyzed for CSA using Lionheart LX automated microscope (BioTek Instruments, Winooski, VT, USA).

Western Blotting
Skeletal muscle protein extracts were obtained by homogenizing 50 mg of quadriceps muscle tissue in RIPA buffer

Real-Time Quantitative Polymerase Chain Reaction (qRT-PCR)
RNA from quadriceps muscle was isolated using the miRNeasy Mini kit (Qiagen, Valencia, CA, USA), following the protocol provided by the manufacturer. RNA was quantified using a Synergy H1 spectrophotometer (BioTek, Winooski, VT, USA). Total RNA was reverse transcribed to cDNA using the Verso cDNA kit (Thermo Fisher

Pyruvate Dehydrogenase (PDH) and Succinate Dehydrogenase (SDH) Enzymatic Activity
The enzymatic activities of PDH and SDH were measured using Colorimetric Assay Kits (MAK051 and MAK197, respectively) from MilliporeSigma based on the manufacturer's instructions. Briefly, 10 mg of quadriceps muscle was homogenized in 100 μl of ice-cold assay buffer followed by centrifugation. 10 μl of sample supernatant was added to 96-well plates. PDH and SDH reaction mixes were added to appropriate wells, resulting in a colorimetric (450nm for PDH and 600nm for SDH) product proportional to the enzymatic activity. The absorbance was recorded by incubating the plate (37˚C for PDH and 25˚C for SDH) and taking measurements (450 nm and 600nm) every 5 min for 30 min.

SDH Staining
Tibialis anterior muscles were cut into 10μm cross-sections on a cryostat and incubated for 30 min at 37°C with 0.5 mg/ml nitroblue tetrazoliumand, 50 mM Na-succinate, and 0.08 mM phenazine methosulfate in PBS. Sections were then rinsed 3 times in deionized water, mounted with PBS-glycerol, and photographed using an Axio Observer.Z1 motorized microscope (Carl Zeiss). Entire SDH-stained sections were quantified for integrated density, as well as total, oxidative, and glycolytic fiber number using ImageJ software.

Oil Red O (ORO) Staining
For ORO staining, tibialis anterior muscles were sectioned (10µm) and immediately fixed in ice cold formaldehyde (3.7%) for 1 hour. Sections were serially washed in Milli-Q water and stained in ORO working solution (prepared as previously described (64)) for 45 minutes at room temperature in the dark. Following ORO staining, sections were again serially washed in Milli-Q water then rinsed in running tap water for 10 minutes. Sections were mounted in 50% glycerol (in PBS) and photographed using an Axio Observer.Z1 motorized microscope (Zeiss, Oberchoken, Germany). Entire ORO stained sections were analyzed for signaling intensity and area of positive staining using ImageJ software.

Metabolomics Analysis by Nuclear Magnetic Resonance (NMR)
Plasma samples for NMR analysis were prepared by diluting 100 μL of plasma with 500 μL of a deuterated phosphate buffer solution (pH = 7.4) containing 2,2,-dimethyl-2-silapentane-5-sulfonate sodium salt (DSS) with a final concentration of 0.5 mM to be used as a chemical shift and quantitation reference. The solution was filtered through a 10 KDa molecular weight cut-off filter to remove large proteins. Samples were then placed in 5 mm NMR tubes for analysis. Muscle and liver tissues for NMR analysis were prepared according to the methanol/chloroform water procedure as previously performed (25). Tissue samples of ~100 mg were used for all samples, but actual weights were recorded to normalize the data. NMR data were acquired on a Bruker Avance III 700 MHz NMR spectrometer with a TXI triple resonance probe operating at 25°C. Spectra were collected with a 1D NOESY pulse sequence covering 12 ppm. The spectra were digitized with 32,768 points during a 3.9 s acquisition time. The mixing time was set to 100 ms, and the relaxation delay between scans was set to 2.0 s. All data were then processed using Advanced Chemistry Development Spectrus Processor (version 2016.1; Toronto, Canada). The spectra were zero filled to 65,536 points and apodized using a 0.3 Hz decaying exponential function and fast Fourier transformed. Automated phase correction and first-order baseline correction were applied to all samples. Metabolite concentrations were quantified using the Chenomx NMR Suite (version 8.2; Edmonton, Canada). The DSS-d6 was used as a chemical shift and quantification reference for all spectra and was set to a chemical shift of 0.00 and a concentration of 500 μM.
Quantitative fitting of each spectrum was carried out in batch mode, followed by manual adjustments to correct for errors arising from spectral overlap. For tissue samples, the final concentrations were normalized based on the weight of the tissue used to prepare each sample. The quantification of glycogen in liver tissue was carried out using a colorimetric Glycogen Assay Kit II (Abcam, Cambridge, USA, #ab16955), per manufacturer's instructions.

Microcomputed Tomography Analysis of Femur Bone Morphometry
MicroCT scanning was performed to measure morphological indices of metaphyseal regions of femurs. After euthanasia, the left femurs were wrapped in saline-soaked gauze and frozen at -20ºC until imaging. Bone samples were rotated around their long axes and images were acquired using a Bruker Skyscan 1176 (Bruker, Kontich, Belgium) with the following parameters: pixel size = 9 μm 3 ; peak tube potential = 50 kV; X-ray intensity = 500 μA; 0.3° rotation step. Calibration of the greyscale levels was performed using a hydroxyapatite phantom. Based on this calibration and the corresponding standard curve generated, the equivalent minimum calcium hydroxyapatite level is 0.42 g/cm3. Raw images were reconstructed using the SkyScan reconstruction software (NRecon; Bruker, Kontich, Belgium) to 3-dimensional cross-sectional image data sets using a 3-dimensional cone beam algorithm. Structural indices were calculated on reconstructed images using the Skyscan CT Analyzer software (CTAn; Bruker, Kontich, Belgium).

RNA Sequencing
Total RNA was first evaluated for its quantity, and quality, using Agilent Bioanalyzer 2100. For RNA quality, a RIN number of 7 or higher is desired. Fifty nanograms of total RNA was used. cDNA library preparation included mRNA purification/enrichment, RNA fragmentation, cDNA synthesis, ligation of index adaptors, and amplification, following the KAPA mRNA Hyper Prep Kit Technical Data Sheet, KR1352 -v4.17 (Roche Corporate). Each resulting indexed library was quantified and its quality accessed by Qubit and Agilent Bioanalyzer, and multiple libraries pooled in equal molarity. Five microliters of 2 nM pooled libraries per lane were denatured, neutralized and applied to the cBot for flow cell deposition and cluster amplification, before loading to HiSeq 4000 for 75b paired-end sequencing (Illumina, Inc.). Approximately 30M reads per library were generated. A Phred quality score (Q score) was used to measure the quality of sequencing. More than 90% of the sequencing reads reached Q30 (99.9% base call accuracy).
The sequencing data were first assessed using FastQC (Babraham Bioinformatics, Cambridge, UK) for quality control.
Then all sequenced libraries were mapped to the mouse genome (mm10) using STAR RNA-seq aligner (65) with the following parameter: "--outSAMmapqUnique 60". The reads distribution across the genome was assessed using bamutils (from ngsutils) (66). Uniquely mapped sequencing reads were assigned to mm10 refGene genes using featureCounts (from subread) (67) with the following parameters: "-s 2 -p -Q 10". Quality control of sequencing and mapping results was summarized using MultiQC (68). Genes with read count per million (CPM) < 0.5 in more than 4 of the samples were removed. The data was normalized using TMM (trimmed mean of M values) method. Differential expression analysis was performed using edgeR (69). False discovery rate (FDR) was computed from p-values using the Benjamini-Hochberg procedure. The heatmap was generated using Partek Flow genomic analysis software (St. Louis, Missouri, USA). Ingenuity pathway analysis (IPA) software was used to identify the canonical pathways and upstream regulators. Pathways with p<0.05 were considered significant. We only considered the upstream molecules that were differentially expressed in our dataset. We considered the pathways which had any z-score. The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE142455 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE142455) (70).

Statistical Analysis
T-tests were used to determine differences between Sham and mC26 groups for figures 1-9. One-way analysis of variance (ANOVA) tests were performed to determine differences between Con, C26, Sham, and mC26. Post-hoc comparisons were accomplished via a Tukey's test, with statistical significance set a priori at p ≤ 0.05. All statistics were performed using GraphPad Prism 7.04 and all data are presented as means ± SD.

Study approval
All studies were in compliance with the National Institutes of Health Guidelines for the use and care of Laboratory   AKT, phospho-mTOR, mTOR, phospho-4EBP1, 4EBP1 (blot 1) and for phospho-p70S6K, p70S6K (blot 2) in the muscle of CD2F1 male mice (12-week old) intrasplenically injected with C26 tumor cells (250,000 cells/mouse in sterile PBS: mC26) or an equal volume of vehicle (Sham) (n=5). Tubulin was used as loading control in both blots.