Impact of extracellular matrix stiffness on genomic heterogeneity in MYCN-amplified neuroblastoma cell line

Tumors are not homogeneous structures, but rather highly complex tissues involving many cell types, such as tumor, stromal and inflammatory cells [1‐3]. Cells are primarily supported in solid tumors by the extracellular matrix (ECM), a three-dimensional (3D) dynamic network composed mainly of fibrous proteins, glycoproteins and proteoglycans [1, 4]. In tumors, cells respond to the biochemical signals of their ECM, but also to physical forces such as tension (traction and compression forces that maintain their stability), known as biotensegrity [2]. In fact, increased stromal stiffness is a classic hallmark of cancer [1, 5], as is transformation of this stiffness into chemical signals through mechanotransduction for cell advantage [5‐7]. Data suggest that an aberrant ECM may promote genetic instability and can even compromise DNA repair pathways necessary to prevent malignant transformation [6, 8‐10]. Among the cell population trying to respond to ECM stiffness, only the most genotypically and phenotypically adapted ones will succeed in invading new metastatic niches [11, 12]. The most advantageous cells will also segregate anchor proteins and transform the ECM into a more biochemically and tensegrally favorable microenvironment, which will also promote growth and migration [6, 13]. A continuous interplay therefore exists between tumor ECM and cell genomics and behavior [13].

ECM stiffness and composition are currently of such importance that, together with mechanotransduction, are considered promising therapeutic targets in many cancer types [14‐16]. This is also the case of neuroblastoma (NB), a peripheral sympathetic tumor that causes 15% of childhood deaths from cancer [17]. We previously defined an aggressive pattern of rigid ECMs in high-risk NB (HR-NB)[18]. NB stiff ECM is rich in cross-linked collagen III fibers, poor in glycosaminoglycans, supporting sinusoidal vascular structures (blood and lymphatic) and with a high quantity of territorial vitronectin (VN, located in the cytoplasmic compartment and in a thin layer around the tumor cells) [19‐22]. NB stiff ECM is also rich in collagen I fibers and fibronectin but in lesser amounts and with a more physiological-like topology than collagen III and VN [18, 23]. Moreover, VN is a glycoprotein containing multiple cell receptor binding sites including integrins, uPAR and PAI-1 [24], and which can therefore form transitory ECM element-cell junctions. It seems to anchor to ECM fibers and proteoglycans, and also disrupt cell adhesion, promoting spread and metastasis in NB [22, 25]. Some studies have shown an important role for VN in initiation and progression of hepatic [26], breast [27] and lung [28] cancers, and its particular involvement in migration in ovarian cancer [29, 30]. In a previous report we used inoculation of the NB cell lines studied here into the adrenal gland in VN knockout mice (VN-KO) to study the systemic role of VN in tumor growth. Interestingly, we found VN-dense cytoplasmic expression in tumor cells from both VN-KO and wild type (VN-WT) mice, with no differences in VN staining pattern or in tumor growth rate in early passages [22]. However, in this study we sought to resolve the particular question of whether host lack of VN has any impact on tumor genomics.

Besides MYCN amplification (MNA), segmental chromosomal aberrations (SCAs), ALK mutation and amplification, TERT rearrangement and ATRX mutation and deletion are also genetic characteristics of HR-NB [31, 32]. In HR-NB patients these genomic aberrations usually show high intratumor heterogeneity, resulting in resistance to treatment and metastasis [33‐35]. We have reported a link between a SCA (1p chromosome arm deletion, 1p-) and reduction of glycosaminoglycans in the ECM of HR-NB [36]. Nevertheless, the association between other ECM features and HR-NB genomic aberrations, and clonal selection, is a neglected area of study.

3D cultures are emerging as good models for ECM-related biological studies, also reducing use of animal models. These scaffolds can reproduce complex ECM features absent from traditional, 2D or monolayer cultures, and allow us to add degrees of complexity to delve deeper into the physiopathological tumor microenvironment [37, 38]. We recently published a study using 3D bioprinted hydrogels composed of methacrylated gelatin (GelMA) and increasing concentrations of methacrylated alginate (AlgMA) to recreate different ECM stiffness. We described a pattern of aggressive behavior in NB cells cultivated for long times in rigid hydrogels [39]; however, that study did not analyze how scaffolding stiffness and growing time affected genomic heterogeneity.

In this work the described models (xenograft VN-KO mice and 3D bioprinted hydrogels) are used to discern how tumor genomics patterns and clonal heterogeneity in two NB cell lines are affected by the VN host status and tumor microenvironment stiffness, and also to extrapolate the results to primary human HR-NB. This study shows that tumor genomics and clonal genetic heterogeneity are directly related to tumor surroundings.

HR-NB is a disease with risky spatial-temporal genetic heterogeneity and high chromosome instability related to poor prognosis and therapy resistance [49, 50]. It is not yet clear whether this genomic variability could be the cause or consequence of selective evolutionary pressure in the tumor microenvironment [11]. We assumed that clonal genetic selection and/or evolution of aggressive NB cells could be promoted in part by the stiffness and composition of their surrounding matrix. To get closer to confirming this hypothesis, we selected two representative NB cell lines characterized by genomic alterations typically observed in HR-NB (MNA in SK-N-BE(2) [42] and ALK mutation in SH-SY5Y) [47]. Both cell lines were grown separately in hydrogels with ECM rigidity similar to the ECM of human HR-NB and xenograft tumors [39]. To make the 3D models even more biomimetic, malignant neuroblasts were cocultured with 10% Schwann cells in some cases, since even in poor stroma NB a low proportion can be found accompanying the undifferentiated or poorly differentiated neuroblasts [51]. To analyze the impact of ECM properties on genomic heterogeneity, we selected the HD-SNPa approach, complemented by the NGS technique, as one of the most commonly used procedures for detecting both MNA and SCAs in routine genetic diagnosis of NB [52]. Unlike this study, most research related to ECM tumor mechanotransduction has focused on uncovering changes in gene expression, which are difficult to translate to the clinic [53‐55]. We looked for typical and/or atypical HR-NB SCA changes that could be associated with NB physics, and rapidly used for clinical translation to diagnosis and therapy. We also explored whether MNA and/or ALK-mutated cells respond similarly to ECM properties.

In previous studies, we demonstrated that elevated VN secretion by tumor cells was found in rigid ECMs, and was related to poor patient outcomes and to growth of orthotopically inoculated NB cell lines [22]. Moreover, we observed that a high amount of VN is secreted forming tracks and we postulated that VN could participate in stiffness, mediating biotensegrity and promoting tumor migration in HR-NB [22, 25]. We also reported that SK-N-BE(2) cells grown in stiff 3D hydrogels show more efficient adaptation, increased cell proliferation, high expression of the Bcl2 antiapoptosis marker, and high mRNA processing rate [39]. In the present study, genomic analysis of SK-N-BE(2) cell line revealed a number of equivalent aberrations in VN-KO tumors and in cells grown in stiffest and/or longer cultured times hydrogels. In these samples, we observed a positive selection of cells that in addition to MNA, contained: i) stable SCAs (1p-, 3p- and 17pq-); ii) SRO generated from previous SCAs (+2p and chromotripsis-like of chromosome 21) and iii) new SCAs and FSCAs (1p-, +1q, 9p-, +9p, 9q-, +9q). We also observed negative pressure which triggered progressive loss of the remaining SCAs (+7q, +11q, 13q-, 19q-, 20p-). Schwann cells seem to enhance clonal selection in NB in a more biomimetic context; however, the genomic differences found between hydrogels with vs without Schwann cells are minor and affect only the percentage of cells with some SCAs.

The evolutionary adaptation we have detected may be due to stochastic processes (generation of new aberrations and genetic drift) or also to deterministic events (clonal selection). However, the fact that cells genomes in two independent models, both with tumor spatial structures related to cellular aggressiveness, show similar genomic aberrations suggests a predominance of the deterministic evolutionary force. The increase in selection pressure, driven by adaptation to niches deliberately altered (by their lack of systemic VN, and by high stiffness and/or long 3D growth time) in both models, could prompt this similar clonal selection towards genetic aberrations detected by probably advantageous cellular phenotypes [56]. Both niches, which initially lacked VN, showed a large amount of VN secreted by neuroblasts [22]. Our results show new evidence for a probable role of VN in biotensegrity mechanotransduction by mediating the cellular response to ECM stiffness [22, 25]. Clonal evolution was faster in VN-KO tumor cells but less manageable than in hydrogels. Furthermore, cells grown in rigid 3D hydrogels showed faster selection of some SCAs and less intratumoral heterogeneity than cells with long and soft 3D growth, which could reflect a more efficient adaptation to niche [39], as occurs in aggressive patient tumors with rigid ECMs [57]. Conversely, in 3D hydrogels with short culture times, cells may not yet have been subjected to the selective process, due to their incipient adaptation to ECM. As in the early stages of patient tumors, cells in 3D bioprinted hydrogels need to grow and survive, and then adjust genetically and epigenetically to acquire the features they need to take on new roles such as migration. For this among other characteristics, 3D models are increasingly important, reducing the use of animal models and controlling the parameters focused on the study, adding degrees of complexity to gain deeper insight into the tumor microenvironment.

Interestingly, some genes of the new SCAs undergoing positive selection in VN-KO tumors and stiff and/or long-cultured hydrogels are involved in ECM composition and architecture, mechanotransduction and cell migration. In chromosome 1 there is special focus on the collagen gene COL11A1 (in the nearer centrometic fragment of 1p-), the genes related to actin ARPC5 and ACTN2, some laminins and the integrin α10 (in +1q), while the deleted FSCA of chromosome 9p contains the genes KANK1 and DOCK8, both involved in actin polymerization [58‐60]. In our customized NB-mechanopanel, COL11A1 and DOCK8 genes specifically showed certain gene variants that had disappeared in the tumors from VN-KO mice with the mentioned heterozygous deletions. The extent to which this occurred, even when the cell population with these mutations was not inconsiderable in the other growing conditions (allelic frequency: 0.4), raises the question of whether these variants might play a role in adaptation. VN of ECM could have a swift impact on the mutational profile of these migration-related genes.

Chromosome 9 aberrations have not been a focus of interest so far in HR-NB, unlike those of chromosome 1 (1p- and + 1q) [32, 61, 62]. Therefore, we decided to delve deeper into 9 chromosome abnormalities in a cohort of primary human HR-NB. Chromosome 9 aberrations involving KANK1 and/or DOCK8 genes were detected in 39.53% of the MNA primary HR-NB analyzed. Consequently, we recommend a spotlight on chromosome 9 aberrations [63, 64] affecting DOCK8 and KANK1 in NB, and especially on the sometimes denigrated FSCAs, as they could involve invasion in MNA HR-NB, as occurs in other malignant tumors [65]. The significance of these new genetic alterations found in human HR-NB, especially the frequency and prognostic impact of DOCK8 and KANK1, should be determined in collaborative studies.

The SK-N-BE(2) cell line has been characterized as particularly heterogeneous [42]. Likewise, high genetic instability of xenograft mice, showing gains and losses of SCAs with passages, has already been described [56]. In all our in vitro and in vivo samples only two SCAs remained stable (with the same length and break points) in most cells: partial deletion of 3p and 17pq, adding MNA. 3p loss is a common event in HR-NB [66]; the region contains certain important suppressor genes, including RHOA. RhoA protein plays an significant role in mechanotransduction, as it mediates focal adhesions and stress fiber formation [67]. Its reduced signal (associated with 3p) has been linked to growth and metastasis in colorectal tumors [68, 69]. Moreover, in these tumors RHOA downregulation is used to match the activation of Wnt signal pathway and inactivation of TGFβ. All together they participate in initiation and progression of tumors [68]. According to Gene Onthology, the Wnt pathway is one of the most affected by genomic aberrations of SK-N-BE(2) cells, and most of the altered genes of these pathways are in SCAs that changed (were positive or negative selected) in VN-KO tumors and 3D bioprinted models (+7q, +9q, +11q, 13q-). TGFβ is located in the 19q deleted region which also had reduced presence in the samples mentioned. This gene has dual action: progression of tumors when underexpressed, and degradation of the ECM when its levels increase [70]. Meanwhile, 17pq- contains the tumor suppressor gene TP53, which plays an important role in tumorogenesis and aggressiveness [71]. This gene was also mutated in all samples from SK-N-BE(2) cell line with an alleleic frequency close to 1. The double mechanism involving a mutation in one of the alleles and a deletion or CNLOH of the other one has been highlighted for its combined negative impact [71]. In addition, the reported imbalance of chromosome 17, in which a loss of 17pq has been detected, is usually found with the gain of the remaining fragment of the chromosome, that is, a typical +17q [72]. 17q gain is precisely the most common genomic aberration of NB [31]. In fact, the hidden +17q became evident in some of our hydrogels and in VN-WT tumors. A probable explanation for this is that in addition to overexpression of tumor progression–promoting genes within the 17q fragments, the loss of function of genes localized on 17p- or CNLOH of 17p (such as TP53) may play a significant role in NB development [72].

Previously published findings could shed some light on the different impact of ECM properties on SH-SY5Y and SK-N-BE(2) cell lines when cultured in the stiffness models, much less marked on the former than the latter. PI3K/AKT is one of the most important pathways in mechanotransduction of mechanical forces, leading to gene expression and protein synthesis [73], and increased matrix stiffness has been reported to activate the PI3K/AKT signal pathway [74]. Various signaling pathways including PTEN/PI3K/AKT and RAF/MEK/ERK have been identified which control MYCN stabilization and act as major mediators of uncontrolled tumor growth, angiogenesis, invasion, apoptosis and cellular metabolism in NB [75‐77]. An over-activated PI3K/AKT pathway could underlie the impact of ECM stiffness on SK-N-BE(2) cell line and other MNA tumor cells. Although ALK mutations signal via numerous downstream pathways, their relationship with ECM stiffness is understudied [78]. Two research groups have reported that morphological and gene expression changes in cytoskeleton and migration-related genes of SH-SY5Y cell line in stiff environments varied with the type of material [53, 54]. Moreover, Rac GTPase has been shown to regulate 3D invasion in NB lacking MYCN amplification [79]. It is plausible that ALK-mutated cells respond to ECM changes through epigenetic alterations, genetic mutations or post-translational modifications undetected by the techniques we employed [53, 80‐82]. Various mechanisms of ECM’s influence on genomic heterogeneity have also been outlined. Some authors have suggested that in stiff solid tissues, tumor cells are squeezed (especially those that are migrating) and can suffer DNA damage, leading to new aberrations and heterogeneity [83, 84]. In this context, it been put forward that SCAs, more than mutations, could be a signature of tissue stiffness [83]. In NB, MNA and chromothripsis-like phenomenon have been described as consequences of genomic instability possibly associated with a deficient DNA repair mechanism, that could have promoted the high genetic heterogeneity observed in SK-N-BE(2) cell line in this and previous studies [42, 85, 86]. Moreover, the stiff microenvironment of the models may have prompted new chromosome breaks in SK-N-BE(2) cells with an already unstable genome. This could explain the high genomic heterogeneity found and the subsequent clonal selection of new SCAs and others already present in a minority in the cell, by leading advantageous phenotypes in the analyzed ECM stiff conditions in MNA NB cell line.