Janus kinase inhibitors suppress cancer cachexia‐associated anorexia and adipose wasting in mice

Cachexia, a syndrome of muscle atrophy, adipose loss, and anorexia, is associated with reduced survival in cancer patients. The colon adenocarcinoma C26c20 cell line secretes the cytokine leukaemia inhibitory factor (LIF), which induces cachexia. We characterized how LIF promotes cachexia‐associated weight loss and anorexia in mice through Janus kinase (JAK)‐dependent changes in adipose and hypothalamic tissues.


Introduction
Cachexia represents a wasting syndrome consisting of muscle atrophy, adipose loss, and anorexia. It is observed in up to 50% of patients with solid tumors. 1 Cancer patients with cachexia have a >50% reduction in overall survival when compared with stage-matched patients without cachexia. 1 Even when cachexia is identified in early stage cancer patients, survival is not improved because of a potential lack of effective therapeutic interventions. 2 Although cachexia has been recognized for more than half a century, most preclinical studies and clinical trials targeting the immune system [tumour necrosis factor alpha (TNF-α) and interleukin (IL)-6], appetite stimulation, and muscle regeneration have failed to durably improve the syndrome. 3 The majority of studies on cachexia have focused on the sarcopenia of this wasting syndrome, with less emphasis on adipose loss or anorexia. Recent evidence suggests that blocking adipocyte lipolysis using global lipase null mice limited not only the adipose wasting but also the sarcopenia observed in murine models of cancer cachexia. 4 Additionally, Zimmers and colleagues demonstrated that pancreatic cancer cachexia patients can have adipose wasting without sarcopenia and an associated decrease in survival. 5 Therefore, identification and complete characterization of cachexia factors and the common mechanisms that they utilize to induce adipose wasting and anorexia may lead to an effective treatment for patients with cancer cachexia.
When transplanted into syngeneic mice, the murine colon adenocarcinoma cell line C26c20 promotes cachexia-associated adipose wasting and anorexia. 6 Previously, we created an in vitro cachexia screen to identify tumour-secreted molecules that can contribute to cachexia-associated adipose wasting. This screen identified leukaemia inhibitory factor (LIF) as a cachexia-inducing molecule secreted from the C26c20 colon adenocarcinoma line. 6 In an analysis of >30 cancers in The Cancer Genome Atlas database, LIF was most highly expressed in gastrointestinal, thoracic, and genitourinary cancers, 6 which are all associated with cachexia. Two recent papers suggested that LIF is critical to pancreatic cancer development, 7,8 further illustrating the importance of this molecule to cachexia-associated cancers. LIF is a 21 kDa protein in the IL-6 family of cytokines that binds to its receptor, LIF receptor-α, and the co-receptor gp130 inducing JAK-signal transducer and activator of transcription (STAT) signalling. 9 Recombinant LIF (rLIF) administered to wild-type mice causes adipose and body weight loss reproducing the cachexia phenotype. 6 LIF induction of cachexia through adipose wasting and anorexia is associated with JAK/STAT signalling peripherally in the adipose tissue and centrally in the hypothalamus. 6 As rLIF induces the cachexia-associated adipose loss, there is a corresponding decrease in serum levels of leptin. Leptin is an adipokine, a cytokine-like molecule produced by adipose tissue, which regulates appetite through JAK/STAT signalling of the hypothalamus in the setting of changes in adipose levels. 10,11 As rLIF induces cachexia-associated adipose loss, the reduction of adipose leptin compensates for rLIF's anorexic effect.
In the present study, we show that the colon adenocarcinoma C26c20 and rLIF-driven cachexia mouse models have increased serum LIF and IL-6 with decreased leptin, defining a cachexia signature in mice. Consistent with these serum findings, adipose messenger RNA (mRNA) levels of LIF, IL-6, and leptin from both cachexia models were similarly altered. By showing that LIF was still able to induce cachexia in IL-6 À/ À mice, we demonstrated that both LIF and IL-6 could independently promote anorexia and adipose loss. We therefore hypothesized that inhibition of the JAK signalling pathway would suppress the cachexia phenotype, because the molecules that are altered in the cachexia cytokine/adipokine signature (LIF, IL-6, and leptin) all signal through this pathway. 12 A screen of candidate JAK inhibitors in our in vitro cachexia adipocyte lipolysis assay led us to test tofacitinib and ruxolitinib in vivo. These JAK inhibitors are effective in the treatment of ulcerative colitis, rheumatoid arthritis, and myelofibrosis. [13][14][15] The independent administration of either JAK inhibitor to rLIF-driven or tumour-driven models of cancer cachexia led to decreased STAT3 phosphorylation in adipose and hypothalamic tissues with a concomitant suppression of cachexia-associated changes to adipose mRNA expression and serum levels of IL-6 and leptin. The net effect of these changes resulted in a mitigation of rLIF-induced anorexia and adipose/body weight loss and additionally led to an improvement of survival in the heterotopic allotransplanted C26c20 cancer cachexia model. These studies suggest that (i) tumour-secreted LIF induces cachexia through JAK-STAT signalling in multiple tissues altering levels of IL-6 and leptin and (ii) JAK inhibition of these signalling events suppresses cachexia-associated adipose loss and anorexia long enough to lead to an improvement in cancer cachexia survival.

Materials and methods
See Supporting Information for a more detailed Materials and Methods.

Mouse studies
Male wild-type Balb/c mice and C57BL/6J were obtained from Charles River Laboratories or Jackson Laboratories at~8 weeks of age. IL-6 À/À mice (B6.129S2-IL6 tm1Kopf /J, 002650) were obtained from Jackson Laboratories at~7 weeks of age. All mice were allowed to acclimate in UT Southwestern animal facilities before experimentation for at least 1 week. Animals were kept in a temperature-controlled facility with a 12 h light/dark cycle and were fed normal chow diet and provided water ad libitum. Approximately 100 g of standard chow diet was placed in each cage. When food reached~50 g per cage, it was replenished to~100 g. Food was weighed at the same time daily and compared with the previous day's weight to calculate the 24 h food intake per cage. Body weight was measured using a standard balance (digital Soehnle scale). Adipose tissue mass and lean tissue mass were measured longitudinally using ECHO MRI (ECHO Medical Systems) at 9 a.m. at the indicated time points. Whole blood was drawn from tail vein bleed longitudinally or by cardiac puncture at terminal time points. Serum was obtained by subjecting the whole blood to centrifugation at ×960 g at 4°C for 10 min. Supernatant was removed followed by protein concentration quantification using a bicinchoninic acid kit (Pierce). For analysis of serum cytokine changes, 25 μL of mouse serum was diluted with 25 μL of phosphate-buffered saline (PBS) and sent to Eve Technologies (Calgary, Canada) for enzyme-linked immunosorbent assay (ELISA) multiplex cytokine analysis of multiple cytokines. For analysis of serum leptin or IL-6, serum dilution and enzyme-linked immunosorbent assay (ELISA) analysis were performed as per kit directions from Crystal Chem and R&D Systems, respectively. The rest of the serum was stored at À80°C for future blood analysis. For tumour studies, 0.75-1 × 10 7 C26c20 cells in 100 μL of PBS were injected into the right hind flank of mice at Day 0, and tumour volume was calculated by taking half of the product of the calliper (VWR) measurements of length, widh, and breadth at the indicated time points. At the end of the experiments, mice were euthanized at the indicated time point in non-tumour experiments or within 12 h of expected death in tumour studies as recommended by the Institutional Animal Care and Use Committee by using a CO 2 chamber, and organs were collected and snap frozen.

Adipocyte lipolysis assay
Media glycerol concentration from differentiated adipocytes was measured for each condition in triplicate as previously described. 6 Pre-adipocytes from the stromal vascular fraction were differentiated into adipocytes using the established and characterized protocol used to characterize adipocyte physiology. 16,17 Statistical analysis Data are presented as mean ± SEM, dot plots ± SEM, or dot plots with bars representing mean ± SEM. A Student's t-test was used to determine differences between groups at distinct time points. A generalized estimating equation approach was used to determine differences between groups over time. A Kaplan-Meier analysis was conducted for survival with statistical evaluation using the Gehan-Breslow-Wilcoxon approach. For some animal studies, the ROUT method was used to remove outliers followed by analysis of variance with a multiple comparison post-test (Dunnett's or Tukey's). Significance was considered if P < 0.05.

Study approval
All animal studies were conducted under an Institutional Animal Care and Use Committee approved protocol at UT Southwestern Medical Center (Dallas, Texas).

Results
Serum and adipose messenger RNA cytokine/adipokine levels in the C26c20 cachexia mouse model To verify that the increased LIF and decreased leptin serum levels observed previously in the rLIF-induced cachexia model 6 were also altered similarly in the C26c20 cancer cachexia model, we evaluated the cytokine and leptin levels in serum from the C26c20 mouse model. The C26c20 cell line is a subclone of the parental C26 colon adenocarcinoma line and induces a cachexia phenotype when administered in vivo. 18 Serum was collected from cancer-bearing mice for cytokine/adipokine ELISA analysis after each animal lost 30-40% of its adipose mass. These mice had a significant serum increase in LIF ( Figure 1A) and a decrease in leptin ( Figure  1B), similar to findings from the rLIF-induced cachexia model. 6 In addition to the expected changes in LIF and leptin levels in the C26c20 cancer cachexia model, we also observed a 10-fold increase in serum IL-6 levels. Other groups have demonstrated an association between altered serum IL-6 levels and cachexia progression. [18][19][20] Furthermore, we generalized these findings of increased serum IL-6 and LIF across multiple syngeneic murine cancer cachexia models, including those created with LLC and 4T1 tumour cells (data not shown).
Considering that adipose is the primary source of leptin, 21 we next tested whether the cancer-induced changes in serum cytokines/adipokines paralleled the changes in adipose mRNA expression levels. As expected, mice bearing C26c20 tumours demonstrated a decrease in mRNA expression of leptin in their white adipose tissue (WAT) as compared with mice receiving PBS alone ( Figure 1C). Although WAT from tumour-bearing mice did not exhibit an increase in their LIF mRNA expression ( Figure 1D), there was an~10-fold increase in the mRNA expression of IL-6 in WAT from tumour-bearing mice compared with vehicle-treated mice ( Figure 1E). These data suggest that cancer secreted factors, such as LIF, not only cause adipose wasting/lipolysis but also change the expression profile of other cytokines/adipokines, IL-6 and leptin, which can also contribute to the adipose wasting and anorexia observed in cachexia.
Serum and adipose messenger RNA leukaemia inhibitory factor, interleukin-6, and leptin levels in the recombinant leukaemia inhibitory factor cachexia mouse model Knowing that rLIF is able to alter serum leptin levels, we hypothesized that it also increases IL-6 serum levels matching changes observed in in vivo cancer cachexia models. To test this hypothesis, we evaluated serum levels and adipose mRNA expression levels of LIF, IL-6, and leptin in our rLIF-driven cachexia model ( Figure 2). As rLIF-injected mice lose fat mass (Figure 2A), there was a parallel increase in serum LIF ( Figure 2B) and IL-6 ( Figure 2C) with a corresponding decrease in serum leptin ( Figure 2D).
To determine whether adipose mRNA expression correlated with rLIF-induced changes in serum LIF, IL-6, and leptin, we collected RNA from WAT of rLIF-treated and vehicle-treated mice for quantitative reverse transcription polymerase chain reaction analysis. Adipose tissue from mice treated with rLIF had a >50-fold increase in IL-6 mRNA expression ( Figure 2F) and an approximately five-fold decrease in leptin mRNA expression ( Figure 2G) with no significant change in LIF expression ( Figure 2E). The changes of IL-6 and leptin adipose mRNA expression correlated with the changes observed in their respective serum levels (compare Figure 2C and 2F with Figure 2D and 2G). Overall, the simple model of rLIF-induced cachexia had a similar serum and Figure 1 Serum and adipose mRNA levels of cytokines/adipokines in a colon cancer cachexia mouse model. Chow-fed Balb/c mice (10-week-old male mice) were injected s.c. in the right flank with PBS in the absence or presence of C26c20 cells. (A and B) Serum cytokine/adipokine levels. Blood was collected by tail vein bleed on Day 0 and by cardiac puncture after each mouse lost 30-50% of its adipose mass as measured by ECHO magnetic resonance imaging. Serum was isolated, and the cytokine and adipokine levels were measured as described in the Materials and Methods. (C-E) Adipose leptin, LIF, and IL-6 mRNA levels. Epididymal white adipose tissue was harvested at sacrifice for measurement of the indicated mRNA by quantitative reverse transcription polymerase chain reaction. For each gene, the amount of mRNA from PBS-treated mice is set to 1, and mRNA amounts from C26c20-bearing mouse adipose tissue are expressed relative to this reference value. The average C t values for β-actin (invariant control) for PBS-administered and C26c20-administered mice were 18.0 and 18.7, respectively. The average PBS-administered mice C t values for leptin, LIF, and IL-6 were 22.1, 28.2, and 33.1, respectively. Data are shown as dot plots with bars representing mean ± SEM (A-E) of five (A) or eight (B-E) mice. * P < 0.05 and *** P < 0.001 based on use of a ROUT method (Q = 0.001) to remove outliers followed by an analysis of variance and Dunnett's multiple comparison post-test comparing the relative change in the indicated serum cytokine relative to VEGF (A) or to PBS control (B-E). These results were confirmed in at least two independent experiments. IL, interleukin; LIF, leukaemia inhibitory factor; mRNA, messenger RNA; PBS, phosphate-buffered saline; TNF-α, tumour necrosis factor alpha; VEGF, vascular endothelial growth factor. Figure 2 Serum and adipose mRNA levels of LIF, IL-6, and leptin in the rLIF cachexia mouse model. Chow-fed C57BL/6 J mice (10-week-old male mice) were injected i.p. with PBS in the absence or presence of recombinant LIF at 80 μg/kg of body weight twice daily, and fat mass (A) was measured by ECHO magnetic resonance imaging. Fat mass is shown relative to the average Day 0 reference value for each respective cohort, which was 3.07 and 3.16 g for the PBS-treated and rLIF-treated mice, respectively. (B-G) Serum levels and adipose tissue mRNA expression of LIF, IL-6, and leptin. Every 3 days, three mice from each cohort were sacrificed for harvesting of serum for the respective cytokine/adipokine ELISA (B-D) and epididymal white adipose tissue for measurement of the indicated mRNA by quantitative reverse transcription polymerase chain reaction (E-G) as described in the Materials and Methods. For each gene, the amount of mRNA from Day 0 is set to 1, and mRNA amounts from adipose tissue from the indicated days are expressed relative to this reference value (E-G). The average C t values for β-actin (invariant control) for Days 0, 3, 6, 9, and 12 for PBS-treated mice were 18.8, 18.8, 18.9, 18.5, and 18.7, respectively, and for rLIF-treated mice were 18. adipose signature to that of the complex C26c20 colon cancer cachexia model ( Figure 1) with a net result of increased serum LIF and IL-6, with a corresponding decrease in leptin, making it a powerful system to understand the biology of cachexia.
Evaluation of recombinant leukaemia inhibitory factor-induced cachexia in an IL-6 À/À mouse model As described previously, rLIF injected into C57BL/6J mice causes an~5-10% muscle loss and an~30-40% adipose loss resulting in an~10-15% body weight loss mimicking a cachexia phenotype. 6 IL-6 has also been reported to induce cachexia in vivo. [18][19][20] We have shown that LIF treatment leads to an increase in serum IL-6 ( Figure 2C) and adipose IL-6 expression ( Figure 2F) in vivo. Therefore, we next determined whether rLIF-associated cachexia is dependent on its induction of IL-6. PBS in the absence or presence of rLIF was injected into IL-6 +/+ or IL-6 À/À mice, and changes in food intake, body weight, and fat and lean mass by ECHO MRI were measured over time ( Figure 3). Figure 3A shows that circulating concentrations of LIF were elevated in both IL-6 +/+ and IL-6 À/À mice administered rLIF at Day 21 compared with Day 0. As expected, circulating levels of IL-6 were also increased with rLIF administration in IL-6 +/+ mice, but not in IL-6 À/À or PBS-treated IL-6 +/+ mice at day 21 ( Figure 3B). In mice receiving rLIF, there is a reduced food intake in both IL-6 +/+ and IL-6 À/À mice as compared with mice receiving PBS during the first 9 days of the experiment ( Figure 3C). Both IL-6 +/+ and IL-6 À/À mice also demonstrated decreased fat mass ( Figure  3D) and body weight ( Figure 3E) when treated with recombinant rLIF compared with control conditions treated with PBS.
Although an increase in IL-6 can promote a cachexia phenotype, LIF's induction of cachexia was not dependent on its ability to increase serum IL-6. Therefore, both molecules are likely driving the cachexia phenotype in cancer models.

Interleukin-6 messenger RNA expression in interleukin-6 family-stimulated differentiated adipocytes
To better understand how LIF and the subsequent increase in IL-6 can stimulate adipose tissue to promote cachexia-associated wasting, we studied the effect of these  cytokines on in vitro differentiated adipocytes in relation to lipolysis and changes in mRNA expression of IL-6, LIF, and TNF-α. As shown in Figure 4A, both wild-type rLIF and recombinant IL-6 (rIL-6) increased adipocyte lipolysis, whereas the mutant rLIF K159A had no effect. The point mutation (K159A) in LIF disrupts its interaction with it receptor, LIF receptor-α. 22 The mutant rLIF K159A is unable to stimulate lipolysis of differentiated adipocytes in vitro or promote cachexia-associated adipose wasting when administered in mice. 6 Compared with rIL-6, rLIF stimulated lipolysis at much lower concentrations. However, rIL-6 caused approximately five-fold higher level of maximum lipolysis than rLIF. Wild-type rLIF and rIL-6, but not rLIF K159A, stimulated the phosphorylation of STAT3 in adipocytes ( Figure 4B, top immunoblot) at similar concentrations to those necessary to stimulate lipolysis ( Figure 4A).
In adipocytes, triglycerides are sequentially hydrolysed by adipose triglyceride lipase, hormone sensitive lipase, and monoacylglycerol lipase, each of which sequentially removes one fatty acid molecule to produce glycerol and non-esterified fatty acids. CAY10499 is a commercially available pan inhibitor of these lipases. 23,24 To evaluate if IL-6-stimulated or LIF-stimulated adipocyte lipolysis is required for the induction of IL-6 mRNA expression, we performed quantitative reverse transcription polymerase chain reaction to assess the fold change of mRNA expression in differentiated adipocytes that were treated with either vehicle, isoproterenol, wild-type rLIF, mutant rLIF K159A, or rIL-6 in the absence or presence of lipase inhibitor CAY10499. Isoproterenol is a β-adrenergic agonist that enhances lipolysis by increasing cyclic adenosine monophosphate, stimulating the phosphorylation and activation of the lipase hormone sensitive lipase. 25 CAY10499 was able to suppress adipocyte lipolysis induced by isoproterenol, rLIF, and rIL-6 ( Figure 4C). However, these cytokines were still able to activate adipocytes in the presence of CAY10499 as demonstrated by cytokine-induced phosphorylation of STAT3 ( Figure 4E). Although isoproterenol-treated adipocytes had a greater than two-fold increase in lipolysis relative to rIL-6-treated or rLIF-treated adipocytes ( Figure 4C), there was no significant increase in IL-6 mRNA expression ( Figure 4D, upper  panel). Contrarily, adipocytes treated with wild-type rLIF or rIL-6 had a >10-fold increase in relative IL-6 mRNA expression, which remained elevated even in the presence of lipase inhibitor CAY10499. The mRNA expression of LIF (middle panel) and TNF-α (lower panel) was unchanged in β-adrenergic-induced or IL-6 family-induced adipocytes in the absence or presence of CAY10499. IL-6 family of cytokines did not affect the mRNA expression of leptin in the well-established in vitro differentiated adipocyte model we utilized because of its inherent low baseline leptin mRNA expression (data not shown). Combined, the results of these experiments demonstrated that IL-6 family of cytokines are able to stimulate IL-6 expression in adipocytes independent of their induction of lipolysis.

Janus kinase inhibition blocks lipolysis, JAK/STAT activation, and interleukin-6 messenger RNA expression in interleukin-6 family-stimulated differentiated adipocytes
Having demonstrated that cachexia factors LIF and IL-6 are both increased in cancer cachexia, an ideal inhibitor would block a common pathway these cytokines activate to induce adipose loss, anorexia, and muscle wasting. Because LIF and IL-6 both induce JAK/STAT activation in adipocytes ( Figure 4B) and in the hypothalamus, 6,26 we next screened JAK inhibitors for their ability to block in vitro IL-6 family cytokine-stimulated adipocyte lipolysis, JAK/STAT activation, and IL-6 mRNA expression. We incubated adipocytes with isoproterenol or rIL-6 in the presence of increasing concentrations of multiple JAK inhibitors including tofacitinib, ruxolitinib, decernotinib, and filgotinib ( Figure 5A). Beta-adrenergic agonist isoproterenol-induced adipocyte lipolysis was not affected by any JAK inhibitor. Adipocyte lipolysis induced by rIL-6 was significantly inhibited when adipocytes were incubated with tofacitinib or ruxolitinib.
To evaluate whether cytokine-induced adipocyte IL-6 expression was dependent on JAK-STAT activation, we incubated adipocytes with PBS, isoproterenol, rLIF, mutant rLIF K159A, or rIL-6 in the absence or presence of tofacitinib or ruxolitinib. As shown in Figure 5B, tofacitinib significantly suppressed rIL-6 and rLIF-mediated, but not isoproterenol-mediated adipocyte lipolysis. To verify that tofacitinib inhibited JAK-mediated STAT activation, immunoblot analysis of adipocyte lysates demonstrated reduced IL-6 family cytokine-mediated STAT3 phosphorylation in the presence of tofacitinib ( Figure 5D, top immunoblot). Tofacitinib also completely suppressed cytokine-mediated induction of adipocyte IL-6 expression ( Figure 5C, top panel). Adipocyte mRNA expression of other cytokines, LIF (middle panel) and TNF-α (bottom panel), was unchanged in adipocytes treated with PBS, isoproterenol, or cytokines in the absence or presence of tofacitinib. Another JAK inhibitor, ruxolitinib, also suppressed IL-6 family cytokine-mediated lipolysis ( Figure 5E), phosphorylation of STAT3 ( Figure 5G), and IL-6 mRNA expression ( Figure 5F, top panel).

Janus kinase inhibition suppresses anorexia and adipose loss in the recombinant leukaemia inhibitory factor cachexia mouse model
To demonstrate that JAK inhibition blocks anorexia and adipose wasting in the rLIF-induced cachexia model, we injected mice with PBS or rLIF in the absence or presence of tofacitinib ( Figure 6A-D) or ruxolitinib ( Figure 6E-J). The addition of tofacitinib or ruxolitinib to rLIF-treated mice restored food intake ( Figure 6A and 6E), body weight ( Figure 6B and 6F), and fat mass ( Figure 6C and 6G) towards levels observed in vehicle-treated mice. There was no change in lean mass among all cohorts ( Figure 6D and 6H) within the limited time frame of the experiment. To confirm that JAK inhibition suppressed STAT activation in target tissues relevant to appetite and wasting, STAT3 phosphorylation was measured in hypothalamic and adipose tissues, respectively, by immunoblot analysis. Mice treated with rLIF and ruxolitinib (lanes 10-12) had decreased STAT3 phosphorylation compared to rLIF-treated mice in the absence of ruxolitinib (lanes 7-9) in hypothalamic ( Figure 6I) and adipose ( Figure 6J) tissues.

Janus kinase inhibition suppresses anorexia and adipose loss improving survival in the C26c20 cachexia mouse model
To test if JAK/STAT signalling pathways are an integral component of cachexia, we tested JAK inhibitors in the in vivo Figure 6 JAK inhibition suppresses rLIF-mediated cachexia. Chow-fed C57BL/6 J mice (10-week-old male mice) were injected i.p. with PBS in the absence or presence of rLIF at 80 μg/kg of body weight twice daily. Ninety minutes after each rLIF injection, mice also received either oral gavage of 200 μL of PBS containing 0.5% (v/v) methylcellulose and 0.1% (v/v) Tween 20 in the absence or presence of 25 mg/kg tofacitinib (A-D) or i.p. of 150 μL of PBS containing 2% DMSO (v/v) and 30% PEG300 (v/v) in the absence or presence of 25 mg/kg of ruxolitinib (E-J) twice daily. Food intake (A and E), body weight (B and F), ECHO magnetic resonance imaging measurement of fat mass (C and G) and lean mass (D and H), or IB analysis of harvested hypothalamic (I) or adipose (K) tissue were measured at the indicated time points. Body weight, fat mass, and lean mass are shown relative to the average Day 0 reference value for each respective cohort. The average values for Day 0 for the PBS with vehicle, PBS with tofacitinib, rLIF with vehicle, and rLIF with tofacitinib were as follows: body weight (27, 25.4, 26.7, and 26.8 g), fat mass (3.6, 3.4, 3.7, and 3.6 g), and lean mass (19, 18, 18.7, and 19 g), respectively. The average values for Day 0 for the PBS with vehicle, PBS with ruxolitinib, rLIF with vehicle, and rLIF with ruxolitinib were as follows: body weight (26.5, 28, 26.4, and 27 g), fat mass (2.7, 3.0, 3.0, and 2.8 g), and lean mass (20,21,20, and 21 g), respectively. Data are shown as dot plots with mean ± SEM (A and E) or each value represents the mean ± SEM (B-D, F-H) of four (A-D) or three (E-H) mice. These results were confirmed in at least three independent experiments. IL, interleukin; JAK, Janus kinase; LIF, leukaemia inhibitory factor; PBS, phosphate-buffered saline; rLIF, recombinant LIF; STAT, signal transducer and activator of transcription. C26c20 cancer cachexia model. Mice were injected with PBS or C26c20 cells on Day 0 in the absence or presence of ruxolitinib (Figure 7). Mice implanted with colon adenocarcinoma C26c20 cells and treated with ruxolitinib had an~20-30% increase in median survival compared to mice receiving C26c20 in the absence of ruxolitinib ( Figure 7A), despite the absence of any significant ruxolitinib effect on tumour growth ( Figure 7B). Mice receiving both C26c20 cells and ruxolitinib had a reduction in the cachexia-associated-adipose loss at intermediate time points ( Figure 7D, right panel), coinciding with decreased adipose tissue STAT3 phosphorylation ( Figure  7F, top blot, compare lanes 4 and 5). At the time that animals were sacrificed due to cachexia morbidity, there was no longer a significant difference in adipose mass either in the absence or presence of ruxolitinib ( Figure 7D, right panel). This endpoint coincided with increasing adipose tissue STAT3 phosphorylation even in the ruxolitinib-treated cancerbearing mice ( Figure 7F, top blot, compare lanes 9 and 10). Mice receiving both C26c20 cells and ruxolitinib also had a reduction in the cachexia-associated anorexia between Days 5 and 8 ( Figure 7C, middle panel), coinciding with decreased hypothalamic tissue STAT3 phosphorylation ( Figure 7F, fourth  blot, compare lanes 4 and 5). However, these mice reached the same levels of anorexia before sacrifice (Days 9-12) as the cancer-bearing mouse cohort receiving vehicle ( Figure 7C, right panel), coinciding with increasing adipose tissue STAT3 phosphorylation even in the ruxolitinib-treated cancerbearing mice ( Figure 7F, fourth blot, compare lanes 9 and 10). During intermediate time points when JAK inhibition was effective at blocking STAT3 phosphorylation in adipose tissue, there was also a significant suppression of cachexia-associated changes in circulating leptin ( Figure 7G) and IL-6 ( Figure 7H). As STAT3 phosphorylation returned in adipose tissue even in the presence of JAK inhibition, the serum levels of IL-6 and leptin approached those observed in a cachexia mouse without JAK inhibition.

Discussion
LIF is a tumour-secreted factor that induces cachexia-associated anorexia and adipose loss through actions on the hypothalamus and increased lipolysis in adipose tissue. 6 In this study, we demonstrated that the murine colon adenocarcinoma C26c20 model has a serum signature of increased LIF and IL-6 and decreased leptin, generalizable across multiple in vivo cachexia tumour models. Consistent with the serum findings, there was also an increase in IL-6 and a decrease in leptin mRNA expression in adipose tissue of cancer bearing mice. These IL-6 and leptin changes in serum and adipose mRNA expression levels were consistently altered in both the C26c20 cancer cachexia model and the rLIF-administered cachexia model. We next considered whether LIF's induction of cachexia was a consequence of its direct effect on target tissues or due to the up-regulation of the other cachexia factor, IL-6. LIF's ability to induce cachexia in vivo was independent of IL-6 because it still promoted cachexia-associated anorexia and wasting in IL-6 À/À mice. However, this LIF-induced systemic increase of IL-6 likely enhanced LIF's overall contribution to cachexia development, with IL-6 able to also stimulate adipocyte lipolysis ( Figure 4A), alter appetite through hypothalamic signalling, 26 and promote the further amplification of systemic IL-6 expression ( Figure 4C). The cytokine-mediated induction of IL-6 mRNA expression was also observed in vitro when differentiated adipocytes were stimulated by LIF or IL-6. We used this model to show that cytokine-mediated IL-6 mRNA induction was not dependent on lipolysis, but rather on JAK-STAT pathway activation. Use of inhibitors of JAK, a common signalling pathway of LIF and IL-6, suppressed cachexia development in rLIF-treated mice and C26c20 tumour-bearing mice, resulting in decreased cachexia-associated anorexia ( Figures 6A, 6E, and 7C) and adipose loss (Figures 6C, 6G, and 7D) that correlated with a parallel decrease in STAT3 phosphorylation in the hypothalamus (Figures 6I and 7F) and adipose tissue (Figures 6J and 7F). Furthermore, JAK inhibition in these cachexia models normalized cytokine-driven alterations in IL-6 and LIF serum levels ( Figure  7G and 7H) during time points of effective inhibition of STAT3 phosphorylation of the adipose tissue ( Figure 7F). Inhibiting the JAK-STAT pathway in target tissues resulted in an~20-30% increase in median survival ( Figure 7A), importantly, without significant change in primary tumour size ( Figure 7B).
Targeting single molecules such as TNF-α, IL-6, or ghrelin has not met the threshold for becoming standard of care treatments for cachexia. [27][28][29][30] The lack of a durable therapy could be due to the following: (i) the previously targeted molecules are not relevant to all types of cachexia; (ii) each patient's cachexia may be driven by a unique set of factors; and/or (iii) there are multiple factors up-regulated in cachexia that contribute to the cachexia phenotype. Our results favour the third hypothesis. They suggest how a single cytokine, such as LIF, amplifies its own signal through targeting of multiple tissues including the hypothalamus and adipose, changing circulating levels of multiple cytokines including IL-6 and the adipokine leptin that work synergistically to promote the cachexia-associated appetite and body composition changes. We hypothesize that cancer cachexia-associated tumours and/or chronic inflammation exploit an intrinsic signalling axis between the immune system, adipose, and the brain to enhance anorexia and wasting, reducing survival in cancer models. Tumour-secreted or immune-secreted cytokines, such as LIF, act on the adipose tissue to promote lipolysis and alter the release of appetite regulating molecules IL-6 and leptin. LIF can also induce expression of IL-6 in other cell types, such as fibroblasts 31 and myotubules. 32 These altered levels of cytokines/adipokines can also act directly on the hypothalamus and other parts of the brain to regulate appetite. 26,33 Additionally, others have shown that IL-6 can also cause cachexia-associated muscle atrophy in a STAT3 dependent manner. 34 We are encouraged that a simplified rLIF-injected mouse model is appropriate to study cachexia since a recent publication showed that genetic silencing of LIF from the C26 parental tumour line led to an anticipated decrease in systemic levels of IL-6 with suppression of the cachexia phenotype. 35 Therefore, our studies support the premise that cachexia patients have multiple circulating cachexia-inducing factors at any given time explaining why therapeutic interventions targeting a single molecule have been ineffective.
Because of the heterogeneity in potential factors driving the cachexia phenotype, there is a need to identify the common downstream signalling pathways to elucidate targets to permit sustained responses. To block LIF and its activation of target tissues, an inhibitor would have to block not only LIF's direct effects centrally on anorexia and peripherally on adipose tissue but also its indirect effects on these tissues from changing other cytokine/adipokine serum levels. Therefore, we inhibited JAK since both LIF and IL-6 use this pathway when signalling target tissues. With evidence that both tofacitinib and ruxolitinib could block lipolysis and induction of IL-6 in our in vitro cachexia adipocyte assay, we tested these compounds in our in vivo cachexia models. Both JAK inhibitors independently blocked rLIF-induced and cancer-induced cachexia, with suppression of anorexia and adipose/body weight loss. In the C26c20 cancer cachexia model, the JAK inhibitors improved median overall survival. The compounds were most effective in blocking cachexia-induced anorexia and adipose loss when they were able to suppress adipose and hypothalamic STAT3 phosphorylation. These data support the importance of both anorexia and adipose loss to the cachexia phenotype. As shown in the aggressive C26c20 cancer cachexia model, these JAK inhibitors were not as effective in sustaining the cachexia suppression durably. There are several reasons why these JAK inhibitors are less effective long term: (i) the pharmacokinetics of these JAK inhibitors are not optimized, (ii) these JAK inhibitors are not optimally blocking the activated JAK subclass upregulated by the cachexia factors, (iii) these JAK inhibitors could be adversely affecting the negative regulators of JAK-STAT signalling, and (iv) JAK independent pathways are being upregulated in the cancer cachexia model. Further studies will need to be done to optimize the pharmacokinetics of the JAK inhibitors used in these studies and to identify other selective inhibitors that block the subclass of JAK molecules regulating the cytokine-mediated signalling. It will also be important to evaluate the effect of these JAK inhibitors on the STAT3 negative feedback regulators including protein tyrosine phosphatases, suppressors of cytokine signalling, and protein inhibitor of activated STAT. The effect of each JAK inhibitor on tumour growth kinetics and immune modulation will also need to be evaluated.
Despite the completion of hundreds of prospective clinical trials, there are no universally accepted therapies for cancer cachexia. One hypothesis for this absence of clinical options is that patients accrued on these studies had cachexia that was end-stage, that is, treatment refractory. These failed outcomes were part of our rationale in treating our cancer cachexia models with JAK inhibition at tumour induction and also why it could be a consideration when managing patients with cancers highly associated with cachexia. Obviously, being able to use biomarkers, such as LIF, IL-6, and leptin, to predict cancer patients at very high risk for subsequent cachexia development could potentially allow us to optimize JAK inhibition use clinically. With the work from this paper and further studies, we anticipate identifying such a cachexia signature that could be validated in future clinical trials to risk stratify therapeutic use.
In patients with myelofibrosis, ruxolitinib improved clinical symptoms and survival. 14 Those who received ruxolitinib also had an~3% weight gain than those receiving placebo who had~2% weight loss. The patients receiving ruxolitinib who gained weight also had a decrease in serum IL-6 and an increase in serum leptin. The changes in serum cytokine/adipokine levels of ruxolitinib-treated patients are consistent with a reversed cachexia cytokine/adipokine serum signature. These clinical findings support the importance of cytokine signalling to human physiology and pathology.
In summary, our studies demonstrate the inherent complexity of treating patients with cancer cachexia, because the presence of even one cachexia factor can amplify its signal by altering the levels of other independently acting molecules that induce cachexia/anorexia. Owing to these changes of serum cytokines/adipokines in cancer cachexia, our data offer an explanation for why therapies targeting single molecules have been ineffective in curtailing cachexia progression. Our current findings indicate that targeting common pathways, such as JAK, of cytokine-mediated signalling in the adipose and centrally will suppress the cancer cachexia phenotype of hypophagia, muscle atrophy, adipose loss, and body weight loss, resulting in improved quality of life and survival.

Online supplementary material
Additional supporting information may be found online in the Supporting Information section at the end of the article. Data S1. Supporting Information