1. IntroductionHyperthermia (i.e., the exposure of tumor tissue to elevated temperatures of 39–44 °C) is gaining interest as an adjunct to radio- and chemotherapy in a wide spectrum of cancers, including cervix, colon, skin and bladder [1], with efficacy supported by an increasing body of clinical trials [2,3,4,5,6]. The therapy-promoting effects of hyperthermia are attributed to various mechanisms, both at the macroscopic and cellular level [7,8,9]. For instance, the elevation of tissue perfusion and oxygenation, in both well- and poorly vascularized regions, are thought to be the main drivers of radio- and chemosensitization by hyperthermia [10,11,12,13,14,15,16,17], as they affect some aspects of the tumor microenvironment that are relevant in disease progression and metastasis, such as hypoxia, acidity and nutrient shortage [18,19]. Moreover, the perfusion-driven increase in oxygen maximizes the cytotoxicity of radiotherapy and different types of chemotherapeutics that show a dependence on oxygen for their efficacy [20]. Accumulating preclinical evidence also suggests that hyperthermia has immunostimulatory effects [21], which might be exploited in novel treatment strategies that include, among others, immune checkpoint inhibition [22]. At the cellular level, heat alters membrane characteristics and induces proteotoxic stress, negatively affecting most compartments and multiple pathways [8].In response to hyperthermia, as well as to many other forms of stress, cells activate a protective mechanism known as the heat stress response. This process was initially characterized as an activation of a subfamily of chaperone proteins, nowadays known as heat shock proteins [23], and many subsequent studies have confirmed and further explored their role in stress tolerance [24,25,26]. In recent decades, however, Omics-based applications, such as microarrays and proteomics, have been applied to systematically uncover heat stress-induced changes in gene expression in different organisms [27,28,29,30], which led to the consensus that heat stress activates functionally different classes of genes that are involved in proteostasis, metabolism, DNA-repair and detoxification [8]. More recent techniques, such as Precision Run-On sequencing (PRO-seq), which identifies the exact positions of transcribing polymerase II complexes across the genome, have provided detailed additional insights into the transcriptional alterations after heat stress [31,32,33,34]. Furthermore, the role of histone modifications [32,35,36] and key transcriptional regulators in the heat stress response, such as HSF1 [30,33,37], have been assessed by chromatin immunoprecipitation sequencing (ChiP-seq) in different organisms.Among these various aspects of the heat stress response, gene expression is among the best studied, yielding a considerable body of data that has been mostly generated using different transcriptomics techniques. Some early contributions reported on gene expression changes in organisms such as A. fulgidus, E. coli, S. cerevisiae, C. elegans and A. thaliana [38,39,40,41,42], as well as in a human (lymphoma) cancer cell line [43]. Subsequent studies explored the transcriptome alterations in different cancer cell lines [43,44,45,46,47,48], and found multiple effects in processes related to protein folding, cell cycle, mitosis, and cell death (Table 1). Nevertheless, as the different experiments and studies focus on, and highlight, different processes, pathways or genes in relative isolation, systemic interpretation and systematic understanding of the emerging gene regulation landscape remain challenging. Meta-analysis has, historically, been among highly useful tools in such cases and, indeed, it has been applied over a decade ago by Richter and colleagues to review and compare gene regulation in response to heat stress [8]. The authors of this study relied on a few then-available datasets to reveal the diverse and largely non-overlapping patterns of gene expression after heat stress in different organisms.

Here, we apply a comparable meta-analytic approach, focusing on 16 public and two unpublished datasets that quantified hyperthermia-induced transcriptome alterations in various human cancer cell lines. Our results demonstrate a high degree of inter-study variability in the transcriptome landscape, and an apparent absence of a universal heat stress response signature. This is likely caused, at least in part, by the different experimental conditions adopted (including cell line, heating technique, thermal dose, time after heat stress, experimental and data analysis pipeline), but due to the limited number of datasets, it is not feasible to confirm which parameters are major drivers of variability. Our analysis highlights, therefore, that the results of individual gene expression studies should generally be interpreted in the context of their particular experimental setup, and that extrapolation of these results to other conditions should be exercised with caution. It also calls for a more extensive, controlled study that would examine the effects of some key parameters, notably the cell line, the thermal dose, the heating technique, and the time after heat stress, in direct comparison.

4. DiscussionsStudies that evaluated global gene expression alterations, or specifically genes regulated by certain transcription factors, have shown that cells, in response to hyperthermia, tune a broad range of pathways that involve thousands of genes [8,27,28,30,31,32,33,34,36,37,61]. Since the publication of a meta-analysis by Richter and colleagues [8], which focused on comparing expression alterations across organisms, multiple studies have reported transcriptomics data of heat-stressed cancer cells [43,44,45,47,48,49]. Here, we leveraged this data, as well as our unpublished results, in an attempt to uncover universal patterns in the heat stress response. Our systematic literature search yielded datasets with diverse characteristics of the thermal dose, the used cell line, the heating technique, and the timing of the transcriptomic analysis (Table 2). An initial global analysis of the data (Figure 1) revealed a high degree of variability between the transcriptome alterations reported by the studies, with several thousands of genes being altered in a subset of datasets, and almost no alterations in others. While it is known that the status of the transcriptome can change rapidly after heat stress [31,33], there was no clear relation between the expression profiles and any of the key experimental parameters. It should be noted, however, that, due to the limited number of studies, such analysis was considerably underpowered, and a correlation cannot be excluded. Thus, the high degree of variation and the inability to pinpoint the source of this variability limits the strength of the conclusions that can be drawn from the analysis. Through pathway analysis, we could confirm the alteration in expression of genes involved in protein folding, cell cycle, mitosis and cell death, in line with the conclusions drawn by the authors of the studies included in the meta-analysis (Figure 2). For instance, Amaya and colleagues [45] pointed out the changes in cell cycle-related genes, in particular those involved in G2/M transition and mitosis. In addition, perturbations of proteostasis by hyperthermia have been observed in microarray data from the studies of Tabuchi and Furusawa [43,48]. With the exception of global cellular stress, we did not observe, however, transcriptome perturbations that were consistent across all datasets. While GSEA initially revealed shared alterations in pathways that are understudied in the context of hyperthermia, such as various aspects of cellular metabolism (e.g., oxidative phosphorylation and glycolysis), as well as mTOR and MYC signaling [62,63,64,65], an overrepresentation analysis showed less commonalities. Interestingly, despite this high variation, we consistently found an alteration of genes related to KRAS signaling, TNF-a signaling and EMT, which connects with recent preclinical and clinical studies that assessed the role of KRAS signaling in cellular responses to hyperthermia [66,67], and in vitro data that shows attenuation of TNF-a signaling [68,69]. It is noteworthy, however, that even the expression of genes that are involved in these pathways varied greatly across the datasets (Figure 3). These inconsistencies also have a negative impact on attempts to pinpoint genes that play a key role in these pathway-level effects by leading edge analysis (Figure 4). In short, although the various types of analyses applied here revealed interesting patterns, the patterns are not universally shared by the analyses, studies or datasets. These results call for a more controlled and comprehensive study that could evaluate the effects of various parameters, most notably the thermal dose, the cell line, and the time after heat exposure, through Omics-based approaches or global genome intervention techniques (i.e., CRISPR KO screening). Such study could not only reveal novel, universal mechanisms driving cellular responses to heat stress, but also potential druggable targets to improve the existing, or develop novel therapies based on hyperthermia.