Lung cancer is the most prevalent cancer type worldwide and responsible for an estimated 1.8 million deaths, each year [1
]. Most patients (~ 85%) develop non-small cell lung cancer (NSCLC) [2
], which is frequently diagnosed in an advanced stage, and consequently has an unfavorable prognosis [3
]. Cancer cachexia is a syndrome that affects a considerable proportion of NSCLC patients [4
]. It is characterized by an ongoing loss of skeletal muscle mass (with or without loss of fat mass) that cannot be fully reversed by conventional nutritional support and is associated with significant functional impairments [5
The loss of skeletal muscle mass in cancer cachexia may lead to substantial weight loss and decreased body mass index (BMI), which are associated with worse outcome in NSCLC patients [6
]. Studies using computed tomography (CT) images have revealed occult muscle depletion in NSCLC patients, regardless of overall body weight [4
]. Also, both the detection of muscle depletion or low muscle mass by CT images have been associated with shorter time to tumor progression, increased risk of chemotherapy toxicity, and shorter survival in NSCLC patients [4
]. Skeletal muscle depletion detected by CT images in these patients also negatively affects their functional status and quality of life [15
]. Indeed, CT-derived pectoralis muscle area (PMA) analysis has been already used to evaluate sarcopenia and to correlate low PMA with shorter survival and inflammation in NSCLC patients [14
]. To our knowledge, tumor-secreted factors with the prognostic value associated with low PMA as detected by CT in NSCLC are unknown.
Several studies have highlighted that macromolecules secreted from cancer cells and cells within the tumor microenvironment (secretome), including many pro-inflammatory cytokines, act systemically leading to muscle wasting in cancer cachexia [18
]. However, the secretome complexity and differences found in distinct lung cancer and cells lines [21
] illustrate the need to apply global approaches, to identify tumor-specific secreted molecules associated with skeletal muscle depletion. Moreover, previous “omics” studies of cancer secretome in cachexia have focused on the analysis of cachectic conditioned media of single cancer cells lines to identify mediators of the syndrome [24
]. However, in vitro systems ignore the contributions of the host–tumor microenvironment and the tumor heterogeneity as well as provide no insight into the disease progression [23
]. These findings emphasize the importance of cancer cachexia studies in exploring the tumor secretome. Thus, we hypothesized that a tumor transcriptome-based secretome analysis in NSCLC patients with low-muscularity is a strategy capable of identifying prognostic biomarkers and mediators of cancer-associated muscle loss.
Herein, we analyzed a cohort of NSCLC patients with CT images, clinical findings, and tumor expression microarrays data from a previous study that decoded tumor radiomics features associated with gene expression levels [27
]. For these patients, we compared the pectoralis muscle area with muscle normalizations based on different radiomics features to select an approach for screening muscularity. Next, we identified genes predicted to be secreted in patients with low-muscularity and assessed their prognostic value as tumor markers of recurrence-free survival and overall survival. Finally, we demonstrated the potential of interleukin (IL)-8 as a putative secreted marker capable of inducing atrophy in C2C12 myotubes.
Using a tumor transcriptome-based secretome analysis in NSCLC patients with low-muscularity, we aimed to identify potential cancer biomarkers of prognostic value and mediators of cancer-associated muscle loss. This strategy revealed increased expression levels of cachexia-related genes predicted to be secreted in NSCLC from patients with lower PMA. These genes were further associated with shorter recurrence-free survival and decreased overall survival in different validation sets of patients with NSCLC. Importantly, increased expression levels of IL-8 were detected in the high-risk group in all NSCLC validation sets, and IL-8 was sufficient to trigger atrophy in C2C12 myotubes.
Muscle depletion or low muscle mass in NSCLC patients identified by CT images has been extensively associated with poor outcome [4
]. Previous studies using the same methodology to ours—the objective assessment of the PMA on CT scans—reported lower PMA associated with worse overall survival in NSCLC patients or cases with chronic obstructive pulmonary disease, despite normalization for BMI and performance status [14
]. Teigen et al. reported that the PMA divided by height (used to standardize for body size) is a powerful predictor of outcome after left ventricular assist device implantation [38
]. Unfortunately, the height in our cohort of PMA CT-based analysis was not available. However, the high quality of these CT images previously allowed the identification of new tumor radiomics features with prognostic value in NSCLC patients [27
]. Thus, we hypothesized that the comparison of the PMA with muscle normalizations based on different radiomics features aiming the standardization for body size could reveal new approaches for screening muscularity in NSCLC patients. Interestingly, PMA distinguished NSCLC patients with low- and high-muscularity in all muscle normalizations tested. Considering that CTs images of lung cancer patients are preferentially performed in the thoracic region, our data additionally confirm that PMA is a feasible measurement easily applied to the clinical practice to distinguish NSCLC patients with different muscularity.
Although a large range of changes in body composition has been associated with tumor-derived factors, including many pro-inflammatory cytokines [18
], only few NSCLC studies associated CT-derived body composition with systemic inflammatory response [40
]. These studies showed that lower muscularity was associated with systemic inflammatory response (IL-6, C-reactive protein, and albumin blood levels, and neutrophil-to-lymphocyte ratio). However, the specific tumor-derived factors that induce muscle loss in NSCLC patients are still unknown. Using the tumor transcriptome analysis of NSCLC patients with low-muscularity, we found 105 deregulated genes, of which 75 were upregulated and 30 downregulated. The functional enrichment analysis revealed upregulated genes related to cytokine activity (CSF3, IL-8, IL-6, BMP6, SCG2, CCL8, BMP2
) and extracellular space (CSF3, FLRT2, IL-8, PLA2G3, IL-6, ATP1B1, COL14A1, LPL, HBB, ADAMTS4
). These results suggest that tumor of patients with low-muscularity possibly secrete cachexia-associated factors.
The in silico analysis confirmed that a set of over-expressed genes were translated into proteins presented in plasma or secretome of NSCLC patients. Seven of these predicted proteins (NCAM1, CNTN1, SCG2, CADM1, IL-8, NPTX1, and APOD) were identified in five databases (SignalP 4.1, SecretomeP 2.0, Vesiclepedia, Human Cancer Secretome, and Plasma Proteome), giving support to their relevance in NSCLC. Although not all NSCLC patients with low-muscularity were cachectic, the tumor gene expression profile identified molecules, such as IL-6
, consistently linked to inflammation and cancer cachexia pathogenesis [29
]. The low muscle mass detected by CT images can occur in the absence of systemic inflammation in other malignancies, such as colorectal cancer, but the proportion of patients with low-muscularity is substantially greater in the presence of systemic inflammation [49
]. In cases where the inflammation coexists with low muscle mass, the prognosis is especially poor [50
]. Taken together, we identified a specific set of upregulated genes coding for secreted proteins that may constitute potential mediators of muscle loss in NSCLC.
Based on the fact that circulating levels of tumor-derived factors were correlated with cachexia development and predicted outcome in cancer [29
], we also investigated the predictive potential of seven transcripts (NCAM1
, and APOD
). All of them were associated with shorter overall survival and recurrence-free survival for the predicted high-risk groups in the NSCLC validation set. However, only IL-8
was over-expressed in the high-risk group in all cohorts of our NSCLC validation set. We further confirmed that high IL-8
expression level in tumor tissue is a strong predictive biomarker significantly associated with worse survival (validation cohort of 1053 NSCLC patients). In agreement with our results, IL-8 expression in tumor tissues was recently associated with cachectic status and outcome in pancreatic cancer; cachectic patients with high IL-8 expression in tumor tissues had shorter overall survival or disease-free survival [31
]. Importantly, these authors also showed that IL-8 expression level in tumor specimen paired with a serum sample from the same patients was associated with tumor size.
We demonstrated that IL-8 directly induced myotube atrophy, reinforcing its potential as a new mediator of cancer cachexia. Muscle wasting in cancer cachexia has been attributed to the combinatorial action of mediators from host and tumor microenvironment [18
]. Also, tumor expression and serum levels of IL-8 have been associated with muscle wasting in patients with different tumor types [29
]. The potential direct effect of IL-8 in inducing muscle cell atrophy is still unknown. In this study, we provide evidence that IL-8 is a biomarker of worse prognosis that has the potential to define the cachectic state in NSCLC patients
The main strength of the present investigation is the identification of potential tumor-derived mediators of muscle wasting in patients with low-muscularity, which have prognostic value in NSCLC. However, our study is based on the reuse of transcriptomic and clinical data, which results in limitations that can be pointed out. Firstly, the validation of the findings at protein levels in NSCLC patients with low-muscularity would be a strategy to define the cachexia blood biomarkers useful for clinical routine. Secondly, our survival analyses were restricted to the validation set; the survival information was not available in our discovery dataset. Finally, since the IL-8
gene is not present in the rodent genome, the atrophy phenotype observed in mice myotubes was likely induced by orthologue receptors to the human IL-8 [51
]. In agreement with our study, Gerber et al. reported that IL-8 protein expression was significantly associated with tumor-free body weight and skeletal muscle weight in a human pancreatic cancer xenograft mouse model [52
]. Further studies are needed to elucidate the mechanisms of action of IL-8 in human muscle cells.