**The DYRK Family of Kinases in Cancer: Molecular Functions and Therapeutic Opportunities**

**Jacopo Boni 1,2, Carlota Rubio-Perez <sup>3</sup> , Nuria López-Bigas 3,4, Cristina Fillat 2,5 and Susana de la Luna 1,2,4,6,\***


Received: 6 July 2020; Accepted: 27 July 2020; Published: 29 July 2020

**Abstract:** DYRK (dual-specificity tyrosine-regulated kinases) are an evolutionary conserved family of protein kinases with members from yeast to humans. In humans, DYRKs are pleiotropic factors that phosphorylate a broad set of proteins involved in many different cellular processes. These include factors that have been associated with all the hallmarks of cancer, from genomic instability to increased proliferation and resistance, programmed cell death, or signaling pathways whose dysfunction is relevant to tumor onset and progression. In accordance with an involvement of DYRK kinases in the regulation of tumorigenic processes, an increasing number of research studies have been published in recent years showing either alterations of DYRK gene expression in tumor samples and/or providing evidence of DYRK-dependent mechanisms that contribute to tumor initiation and/or progression. In the present article, we will review the current understanding of the role of DYRK family members in cancer initiation and progression, providing an overview of the small molecules that act as DYRK inhibitors and discussing the clinical implications and therapeutic opportunities currently available.

**Keywords:** DYRK kinases; cellular signaling; expression dysregulation; cell cycle; cell survival; tumor progression; kinase inhibitors

#### **1. Background**

The first cancer gene identified, the proto-oncogene *c-Src*, was found to encode a protein kinase [1]. Yet, since then, almost a hundred kinase genes have been attributed a tumor suppressor or oncogenic role, and they represent the most abundant class of cancer driver genes known to date [2]. Dual-specificity tyrosine-regulated kinases (DYRKs) belong to the CMGC group of kinases, which includes cyclin-dependent kinases (CDKs), mitogen-activated protein kinases (MAPKs), CDK-like kinases, the serine-arginine-rich protein kinase, Cdc2-like kinases (CLKs) and members of the RCK family [3]. The DYRK family is formed by three subfamilies: the DYRK subfamily, the homeodomain-interacting kinases (HIPKs), and the pre-messenger RNA-processing protein 4 kinases (PRP4Ks) [3]. Here, we will use "DYRK" to refer specifically to the DYRK subfamily, which contains five members in humans that are clustered into two classes based on their phylogenetic relationships [4]: class I DYRKs, DYRK1A and DYRK1B (also known as Mirk from minibrain-related kinase) and class II DYRKs, DYRK2, DYRK3 (also known as REDK from regulatory erythroid kinase) and DYRK4 (Figure 1A).

**Figure 1.** Dual-specificity tyrosine-regulated kinase (DYRK) protein kinases: primary structure and expression. (**A**) Scheme of the mammalian family of DYRKs, indicating their phylogenic relationships, degree of homology and protein domains. The catalytic domain (KINASE) and the DYRK homology box (DH) are common to all members of the family. Class I DYRKs have two nuclear localization signals (NLSs) (NLS1 and NLS2) and a proline-, glutamic acid-, serine- and threonine-rich (PEST) motif. DYRK1A also includes a tract of 13 consecutive histidine residues (His) and a region enriched in serine/threonine residues (S/T) at the C-terminus. Class II DYRKs have a common structure, with the characteristic N-terminal autophosphorylation accessory (NAPA) domain at the N-terminus. In the case of DYRK2 and DYRK4, functional NLSs have been described within the noncatalytic N-terminus. (**B**) The expression of human DYRKs based on the Genotype-Tissue Expression (GTEx) data represented as the median TPMs (transcripts per million: GTEx Analysis Release V8, www.gtexportal.org/home, dbGaP Accession phs000424.v8.p2). Tissues represented in the tumor data in Table S1 were chosen (brain: cortex; cervix: ectocervix; colon: sigmoid colon; esophagus; mucosa, kidney: cortex; skin: suprapubic—not sun-exposed).

DYRK kinases phosphorylate a broad set of substrates that are involved in a wide range of cellular processes, and they are thought to fulfill essential biological functions both during development and in maintaining homeostasis during the adult life. Consequently, the aberrant regulation or expression of DYRK kinases has been associated with several human pathologies, including cancer. In the present article, we will review our understanding of the role of DYRK family members in cancer initiation and progression, providing an overview of the small molecules that act as DYRK inhibitors and discussing the clinical implications and therapeutic opportunities currently available.

#### **2. The DYRK Family of Kinases**

The members of the DYRK family all share a highly conserved catalytic domain with special features within the CMGC group [5] and the so-called DYRK homology (DH) box motif located upstream of it (Figure 1A). In addition, DYRK kinases present class-specific domains: DYRK1A and DYRK1B harbor a proline-, glutamic acid-, serine- and threonine-rich (PEST) motif in the noncatalytic C-terminal region and equally positioned nuclear localization signals (NLS) (Figure 1A). On the other hand, class II DYRKs present a N-terminal autophosphorylation accessory region (NAPA) domain, essential for catalytic activation [6] (Figure 1A). All human DYRKs accumulate in the cytosol of cells, and DYRK1A, DYRK2 and DYRK4 can be imported into the nucleus by means of dedicated NLSs [7–9]. DYRK1A translocation to the nucleus acquires special biological significance, since it has been described as a chromatin-associated kinase capable of regulating the gene expression [10,11], and it is functionally linked to the DNA damage response (DDR) [12–14]. Chromatin association in the DDR context has also been recently described for DYRK1B [15]. Moreover, a DYRK1A-specific run of histidine residues targets this family member to the subnuclear splicing compartment [7], and the noncatalytic N-terminal domain of DYRK3 serves to localize it to stress granules [16]. Both the histidine run in DYRK1A and the N-terminus of DYRK3 participate in the generation of phase-separated subcellular compartments [17,18]. Changes in the subcellular localization of DYRK proteins have been observed in response to different signals, such as that of DYRK2 in response to DNA damage or proinflammatory signals [19,20] or DYRK1A in response to Wnt signaling [21]. However, how the subcellular localization of DYRKs is regulated or how it contributes to their activity is still not well-understood.

A high-throughput transcript data analysis indicates that DYRK1A and DYRK1B are expressed ubiquitously in human tissues, whereas class II DYRKs are generally expressed more weakly and in a more tissue-restricted pattern (Figure 1B). The expression of DYRKs is regulated through alternative promoters that generate transcripts with distinct 5′ -untranslated regions and/or encoding different N-terminal regions [4]. In addition, alternative splicing generates multiple protein isoforms of unclear functional significance [4,8,22–24]. DYRKs are also subject to other post-transcriptional events, such as microRNAs (miR)-mediated gene silencing [25–27] or local translation [28].

DYRK kinases are "dual specificity" kinases, as they can phosphorylate both tyrosine (Y) and serine/threonine (S/T) residues, although Y-phosphorylation is limited to their autophosphorylation activity [29]. These kinases are activated by the phosphorylation of residues within the activation loop, which drives a conformational switch from the inactive to active state [30,31]. Unlike other kinase families, this key event in DYRKs is an autocatalytic reaction that occurs during protein synthesis and that generates a constitutively active kinase [32]. As DYRK activation does not depend on upstream kinases, other regulatory mechanisms are thought to operate. These include: the dephosphorylation of residues in the activation loop, although no phosphatase has been attributed this role to date, allosteric phosphorylation performed by other kinases [9,33–36], interactions with scaffolding proteins [37–39] or accessibility to substrates due to changes in the subcellular localization. In this regard, and given the constitutive nature of DYRK kinase activity, the regulation of their intracellular levels becomes crucial to modulate their functions, and thus, altering the DYRK expression acquires additional importance in terms of their impact on normal cell fitness.

#### **3. The Role of DYRKS in Cancer**

DYRKs phosphorylate a wide range of substrates, including factors associated with one or several of the hallmarks of cancer [40] (Figure 2). Of all the DYRKs, only DYRK1A has been identified in high-throughput cancer studies, initially as a potential tumor suppressor using Tumor Suppressor and Oncogene Explorer (TUSON), a method developed to predict the potential of a given gene to act as a tumor suppressor, or oncogene, by computing somatic mutation profiles and copy number alterations (CNAs) [41]. Subsequently, it was proposed as a driver in liver cancer through a study that identified such drivers according to mutations in unusual nucleotide contexts [42]. Although these results would suggest that DYRKs are not major drivers of cancer, further evidence that they play a role in oncogenic processes has emerged over the past two decades. In the following sections, we will discuss the evidence indicating that each member of the DYRK family is involved in cancer by considering two main aspects: (i) alterations to the DYRK expression in tumor tissues, either based on published reports or on our own analysis of The Cancer Genome Atlas (TCGA: see Table S1; only cancer type cohorts with at least 10 paired samples, matched tumor-healthy tissue, were considered

in the analysis), and (ii) the impact of DYRK-dependent phosphorylation on substrates involved in cancer-related events.

**Figure 2.** DYRKs are involved in cancer-associated processes. DYRK kinases participate in the regulation of crucial cell events, the perturbation of which is responsible for producing important features in cancer cells or the hallmarks of cancer.

#### **4. DYRK1A**

The *DYRK1A* gene maps to chromosome 21, and it is the most extensively studied member of the family, mainly due to its key role in neurogenesis and in the etiology of some of the pathological traits associated to Down syndrome (DS: recently reviewed in [43]). In fact, *DYRK1A* is a dosage-sensitive gene, since small variations in the amount of its protein produce clinical phenotypes. On the one hand, DYRK1A is overexpressed 1.5-fold in DS individuals [22], and indeed, some of the morphological and cognitive defects of DS are reproduced when it is overexpressed in mouse models [43]. On the other hand, DYRK1A haploinsufficiency caused by de novo truncation or by missense-inactivating mutations was recently seen to underlie a rare, severe disorder, the DYRK1A haploinsufficiency syndrome (also known as MRD7 or Mental Retardation, Autosomal Dominant 7: OMIM#614104 and ORPHA:464311 and 268261; [44,45] and references therein).

DYRK1A is a pleiotropic factor that phosphorylates a broad set of proteins involved in many different cellular processes. These include factors that have been associated with all the hallmarks of cancer, from genomic instability to increased proliferation and resistance to programmed cell death or signaling pathways whose dysfunction is relevant to tumor onset and progression (e.g., Wnt, Notch and Hedgehog (Hh); Figure 3 and Table 1). Notably, the role of DYRK1A in specific cell responses has contrasting outputs, suggesting that it can act as a bimodal signaling regulator. For instance, DYRK1A stimulates the transcriptional activity of the Hh-signaling effector GLI1 through direct phosphorylation (Figure 3), although it also represses the Hh pathway through an indirect mechanism involving regulators of the actin cytoskeleton [46,47]. Likewise, DYRK1A negatively regulates the nuclear factor of activated T-cell (NFAT) transcription factors by inducing their phosphorylation-dependent nuclear export [48], yet it serves as a positive modulator of NFAT signaling in primary endothelial cells stimulated by vascular endothelial growth factor (VEGF) [49] (Figure 3). DYRK1A bimodal activity has also been reported in Wnt signaling, where DYRK1A acts as a positive regulator of the activated pathway, but it represses basal Wnt-signaling activity [21]. Finally, DYRK1A may induce cells to

either enter or exit the cell cycle by controlling the Cyclin D1-to-p21 ratio [50]. All these observations may reflect the different experimental systems used by different groups, and specifically, the ectopic expression might produce confounding effects, since dramatic changes in the DYRK1A protein might be transiently induced in these cells over and above the endogenous levels. Alternatively, these findings might actually support the bimodal activity of DYRK1A in vivo, with the different outcomes depending on specific conditions such as cell identity, subcellular localization or the levels of kinase expression. Along similar lines, several studies have ascribed opposite functions to DYRK1A in cancer, reflecting a very complex scenario. Therefore, as will become evident below, it remains unclear as to whether DYRK1A acts as a tumor suppressor or a tumor promoter or, more probably, as either, depending on the tumor context.

**Figure 3.** DYRK1A modulates the cellular factors involved in oncogenic processes. An overview of the DYRK1A interactions with the cellular factors involved neoplastic transformation and cancer-related pathways. CycD1: cyclin D1; DVL1: dishevelled 1; DREAM: dimerization partner (DP), RB-like, E2F and multi-vulval class B (MuvB); EGFR: epidermal growth factor receptor; HPV: human papilloma virus; ID2: inhibitor of DNA binding 2; NFAT: nuclear factor of activated T-cells; REST: RE1 silencing transcription factor; VEGFR2: vascular endothelial growth factor receptor 2.


#### **Table 1.**Signaling molecules targeted by DYRK1A and other dual-specificity tyrosine-regulated kinases (DYRKs).

\* Not a direct phosphorylation target. ABLIM1: actin binding LIM protein 1; CBP: CREB binding protein; DKK3: dickkopf WNT signaling pathway inhibitor 3; DVL1: dishevelled 1; EGFR: epidermal growth factor receptor; EGLN2/PHD1: egl-9 family hypoxia inducible factor <sup>2</sup>/prolyl hydroxylase 1; HIF2α: hypoxia-inducible factor 2-alpha; ID2: inhibitor of DNA binding 2; NFAT: nuclear factor of activated T-cells; RTKs: receptor tyrosine kinases and VEGFR2: vascular endothelial growth factor receptor 2. n.d.: not determined.

#### *4.1. DYRK1A and Cell Cycle Regulation*

The first indications of a role for DYRK1A in cell immortalization were obtained in studies on oncogenic viruses, indicating that DYRK1A potentially affects cell transformation in oncovirus-associated cancer models. Both DYRK1A and DYRK1B interact with the adenovirus oncoprotein E1A, a feature conserved in the *Saccharomyces cerevisiae* DYRK Yak1p [71,72] (Figure 3). Mutations in E1A that interfere with DYRK1A binding produce hyper-transformation in conjunction with G12V HRAS proto-oncogene [72]. Moreover, the interaction between DYRK1A and E1A is dependent on the DCAF7 scaffold protein, which favors E1A phosphorylation at S89 [39] and contributes to the ability of the adenovirus to regulate the interferon response [73]. DYRK1A also interacts functionally with human papilloma virus (HPV), and *Dyrk1a* mRNA levels increase when primary mouse keratinocytes are immortalized by HPV infection (HPV high risk strain 16) [74]. Indeed, there is more DYRK1A protein in cervical lesions from HPV-derived patient samples than in the respective normal tissues. Alterations to the DYRK1A expression might involve the miR-1246 known to target DYRK1A [26], which is significantly downregulated in lesions from cervical cancer patients in a manner associated with HPV infection [75]. DYRK1A interacts and phosphorylates HPV16 E7, stabilizing E7 and thereby potentially promoting E7-dependent cell proliferation [76] (Figure 3). Moreover, DYRK1A interacts with beta-HPV E6 proteins (Figure 3), and this DYRK1A interaction is defective in HPV E6 variants found in invasive cervical carcinoma [77].

The link between DYRK1A and cell proliferation is based on its ability to phosphorylate crucial cell cycle regulators, like Cyclin D proteins or p27 (Figure 3 and Table 1), modulating their stability and, hence, their cellular levels [50,53,58,78]. It should be noted that these regulatory mechanisms have mainly been observed in nontransformed cells, and as mentioned above, the effect of DYRK1A on the cell cycle is not straightforward, as it depends on the Cyclin D-induced stabilization of the CDK inhibitor p21, at least for Cyclin D1 [50]. In addition, DYRK1A is a kinase in the DREAM complex (dimerization partner (DP), RB-like, E2F and multi-vulval class B (MuvB)). DYRK1A promotes the assembly of this complex by phosphorylating the DREAM component Lin52 on S28, thereby triggering cell cycle exit [54] (Figure 3). Notably, DYRK1A-mediated DREAM complex formation was proposed to be responsible for ovarian cancer cell dormancy [79] and for the quiescence of gastrointestinal stromal tumor (GIST) cells induced by treatment with imatinib [80]. Another DYRK1A cell cycle-related target is the p53 tumor suppressor (Figure 3). DYRK1A positively regulates p53 transcriptional activity by the direct phosphorylation of S15 [60], but it also negatively regulates this factor by enhancing sirtuin (Sirt)1-dependent deacetylation [81]. The functional interaction of DYRK1A with p53 promotes cell cycle arrest in embryonic neuronal cells [60], as well as the survival of osteosarcoma and colorectal cancer (CRC) cell lines in response to genotoxic stress [81]. The cross-talk between DYRK1A and p53 also involves a negative feedback loop that engages two distinct regulatory mechanisms: (i) the p53-dependent induction of miR-1246, which suppresses DYRK1A expression [26], and (ii) the degradation of the DYRK1A protein mediated by the E3 ubiquitin ligase mouse double-minute 2 homolog (MDM2) [82].

#### *4.2. DYRK1A and Receptor Tyrosine Kinase (RTK)-Dependent Signaling*

An important aspect of the participation of DYRK1A in oncogenic processes is related to the regulation of RTK-dependent signaling. This class of protein kinases is frequently altered in tumors, with almost half of them included in the list of driver kinases assembled in 2016 [2]. We found a conserved regulatory pattern that involves the positive effects of DYRK1A on the stability of several RTKs (Figure 3 and Table 1), yet it is unclear whether these effects are mediated by a shared DYRK1A target or one specific to each RTK. Thus, DYRK1A prevents epidermal growth factor receptor (EGFR) endocytosis-mediated degradation in neural stem cells [65] and indeed, DYRK1A-dependent EGFR stabilization has been described in glioblastoma (GBM) and non–small cell lung cancer (NSCLC) cell lines [66,67]. Indeed, DYRK1A and EGFR protein levels correlate in tissues from glioma patients [66]. In pancreatic ductal adenocarcinoma (PDAC) tumor tissue, a similar relationship was found between

the expression of DYRK1A and c-MET, the hepatocyte growth factor receptor [68]. DYRK1A exerts a positive role on the c-MET protein levels in cell models of PDAC and NSCLC, which might contribute to the protumorigenic role of DYRK1A in these types of tumors [67,68]. Given that RTKs are common targets in cancer therapy [83], the inhibition of DYRK1A (and its paralog DYRK1B) could be considered an element in combinatorial therapies to simultaneously target several deregulated RTKs. Finally, DYRK1A depletion reduces the levels of membrane-bound VEGF receptor 2 (VEGFR2), and it causes defects in VEGFR2-dependent signaling and the downstream NFAT-dependent transcriptional response in endothelial cells [49]. These results are correlated with the defects in developmental angiogenesis in a mouse model in which the *Dyrk1a* dosage is reduced [49], although whether DYRK1A has a proangiogenic role in the tumor microenvironment needs to be further explored.

DYRK1A regulates other cell factors known to participate in malignant transformations, including the stemness-related RE1 silencing transcription factor REST [84] or key effectors of cancer-promoting signaling pathways, like the Hh, Wnt and Notch pathways (Table 1). However, whether DYRK1A is connected to alterations in these pathways during tumor initiation/progression has not yet been established.

#### *4.3. DYRK1A in Cancer*

Changes in the *DYRK1A* expression have been analyzed in tumor samples, and as such, *DYRK1A* was seen to be downregulated in breast cancer [27] and in acute myeloid leukemia (AML) tissue [55], and it is upregulated in GBM [66], lung cancer [67] and head and neck squamous cell carcinoma (HNSCC) [85], as well as in PDAC [68]. Indeed, a weaker DYRK1A expression was correlated with a worse overall survival in breast cancer patients [27] and a poorer prognosis in CRC and GBM patients [62,86], whereas more DYRK1A was associated with a reduced survival time in patients with lung cancer [67]. Our analysis of the TCGA RNA-Seq data revealed a clear trend towards *DYRK1A* downregulation in tumor tissues, with a significant downregulation of *DYRK1A* in 11 out of the 15 tumor types considered (Table S1): colon (COADREAD), esophagus (ESCA), HNSCC, kidney (KIRP and KIRC), liver (LIHC), lung (LUSC and LUAD), stomach (STAD), thyroid (THCA) and uterus (UCEC). No significant CNAs associated with changes in the gene expression were observed (Table S1), suggesting that the changes in RNA levels could be due to epigenetic, transcriptional or post-transcriptional alterations. The general trend towards a reduced *DYRK1A* expression in tumor samples would be in agreement with a more prominent tumor-suppressor role, even though the correlation between *DYRK1A* mRNA and the protein levels has not been properly evaluated in any cancer study.

A direct role for DYRK1A in tumor progression has been proposed in several studies. In cell models, DYRK1A knockdown or enzymatic inhibition reduced the proliferation of HNSCC cell lines [85], luminal/HER2 breast cancer [87] or PDAC [68], as well as impaired the self-renewal capacity of GBM cells [66] and compromised ovarian cancer spheroid cell viability [79]. The pro-oncogenic role suggested by these findings is in accordance with results obtained from xenografts in mouse models [66,68,85]. However, a tumor suppressor role was also proposed on the basis of DYRK1A overexpression experiments in AML cells [55]. The antitumor role of DYRK1A was suggested to be related to the lower incidence of cancer in DS individuals [88], deviating from that observed in the normal population. Indeed, epidemiological studies have demonstrated that individuals with DS have a markedly lower incidence of most solid tumors [89] and reduced cancer-associated mortality [90] relative to the age-adjusted non-DS population. However, childhood leukemia represents a strong exception to this trend, as DS children have a 10 to 50-fold increased risk of developing AML, as well as a 500-fold increased incidence of developing acute megakaryoblastic leukemia (AMKL) [91]. In this regard, DYRK1A was proposed to be a potent, megakaryoblastic oncogene, suggesting that NFAT-negative regulation through an imbalance in DYRK1A might perturb myeloid differentiation and promote AMKL in DS individuals [92].

In summary, the literature reflects a complex picture in which DYRK1A may fulfill opposite roles in different tumor contexts. Thus, more research is clearly required to fully understand how DYRK1A contributes to tumor initiation or progression.

#### **5. DYRK1B**

DYRK1B is the closest paralog to DYRK1A, sharing 85% homology that extends beyond the kinase domain (Figure 1A). Although both kinases share substrates (Table 1), the distinct clinical outcome of inactivating mutations indicates they are not functionally redundant, i.e., a disorder within the autism spectrum for DYRK1A and a metabolic syndrome for DYRK1B (abdominal obesity metabolic syndrome-3, OMIM#615812; [93]). A recent review of DYRK1B has offered extensive information on this kinase [94], and thus, here, we will focus on those aspects of the kinase that are related to its role in cancer, which, unlike DYRK1A, point mostly to a prosurvival and protumorigenic role for DYRK1B (Figure 4).

**Figure 4.** DYRK1B promotes survival and chemoresistance in cancer cells. Environmental stress conditions induce DYRK1B expression or activity in tumor cells, which, in turn, promotes cell cycle exit, quiescence (entry in G0) and survival. This mechanism has been proposed to mediate the resistance to chemotherapeutic agents that target dividing cells.

The first studies into the influence of DYRK1B in cancer suggested a role in the survival of cancer cells, with a stronger DYRK1B expression in CRC samples than in normal tissue [95]. Several studies extended this finding to other tumor types, included liposarcoma [96], rhabdomyosarcoma [97], osteosarcoma [98], lung [99], breast [100], ovary [101] and PDAC [68,102,103]. Indeed, a differential expression analysis using TCGA data finds *DYRK1B* to be overexpressed in several tumor types, including bladder (BLCA); breast (BRCA); kidney (KICH, KIRC and KIRP); liver (LIHC); prostate (PRAD); thyroid (THCA) and uterus (UCEC) (Table S1). Furthermore, we confirmed previous reports on the amplification of the *DYRK1B* genomic region (19q13.2) in ovarian cancer [104,105] and PDAC [102,106] (Table S1). The amplification of this region with coherent DYRK1B overexpression was observed in other tumor types (Table S1), suggesting that they may underlie the increase in *DYRK1B* expression, although this may also be provoked by transcriptional activation due to changes in the transcriptional profiles of tumor cells [107–111].

The functional interaction of DYRK1B with signaling pathways involved in cancer cell proliferation has been explored, assessing both the fluctuations in DYRK1B expression upon the perturbation of growth pathways and the output provoked by DYRK1B depletion in cancer cell lines. Several findings point to an antagonistic role of DYRK1B and MAPK signaling, with an increase in DYRK1B in response to inhibitors of the MAPK kinase (MEK) in CRC and melanoma cell lines [36,95] and a reduction following the mitogen activation of the RAS-MEK-extracellular signal-regulated kinase (ERK) pathway in skeletal myoblasts [112]. The cross-talk between DYRK1B and the MAPK pathway was further explored in ovarian cancer and NSCLC cell lines, where DYRK1B knockdown increased c-RAF and ERK activation [107]. This DYRK1B-MAPK cross-talk might be even more complex, since DYRK1B is an ERK substrate at a residue that potentiates DYRK1B activity [36], and accordingly, oncogenic KRAS mutants act as positive modulators of DYRK1B activity [113,114]. The RAS-DYRK1B axis was proposed to participate in both autocrine and paracrine Hh signaling in PDAC [114], although the role of DYRK1B in the regulation of Hh signaling in cancer remains controversial, as it has been attributed opposite functions within this signaling pathway [61,114,115]. Finally, there also appears to be cross-talk between DYRK1B and the mammalian target of rapamycin (mTOR) pathway, with DYRK1B expression upregulated upon mTOR inhibition [109] and mTOR/AKT activation induced by DYRK1B within the Hh signaling pathway in pancreatic and ovarian cancer cells [115].

Like DYRK1A, DYRK1B phosphorylates several cell cycle regulators, like Cyclin D1, p21, p27 and Lin52 [52,54,57,59] (Table 1). In this context, DYRK1B overexpression may help maintain a reversible quiescent state or inhibit cancer cell proliferation [116–118], while DYRK1B reduction can drive cell cycle entry in quiescence (by reducing the DYRK1B expression in PDAC or by DYRK1B inhibition in CRC cell lines) [119]. By contrast, the depletion of DYRK1B in HPV E7-expressing keratinocytes interferes with the induction of the S-phase promoted by E7 [120]. In addition, the depletion or inhibition of DYRK1B enhances the DNA damage, apoptosis and sensitivity to reactive oxygen species (ROS) or chemotherapeutic drugs targeting proliferating cells [15,104,121–123], as well as the sensitivity to compounds that target pathways favoring proliferation in cell lines of different tumor origins, such as mTOR and MEK inhibitors [107,109] (Figure 2).

A protumorigenic role for DYRK1B has been proven in cellular models of ovarian and pancreatic cancers. Thus, DYRK1B knockdown negatively affects different aspects of ovarian cancer cell malignancy, including viability, proliferative potential and migratory capacity [1,124,125]. Likewise, DYRK1B knockdown negatively affects PDAC cell proliferation, migration and invasion [68,102], whereas a treatment of PANC1 xenografts with a DYRK1B inhibitor impairs tumor growth [103]. In summary, and in contrast to the controversial role of DYRK1A in cancer, clear oncogenic facets have been attributed to DYRK1B, acting as a prosurvival factor that could help cancer cells survive in suboptimal growth conditions and preventing chemotherapeutic-induced DNA damage and apoptosis.

#### **6. DYRK2**

DYRK2 is a class II DYRK that has been more intensely studied in terms of its involvement in the events associated with tumor progression. The biochemistry and biology of DYRK2 was covered in recent reviews [126,127], and thus, here, we will center on the activity of this kinase in the context of tumor biology.

#### *6.1. Altered DYRK2 Expression in Cancer*

The first hints that DYRK2 may influence carcinogenesis were derived from a genomic analysis and differential gene expression studies, highlighting *DYRK2* overexpression in association with the amplification of its genomic locus in esophageal and lung adenocarcinomas [128,129], GIST [130], gastric adenocarcinoma [131] and liposarcoma [132]. Additional evidence for the involvement of DYRK2 in cancers came from a germline-somatic association study of genetic alterations in multiple cohorts of breast cancer patients [133]. Moreover, the upregulation of DYRK2 was described in triple-negative breast cancer (TNBC) and multiple myeloma [134]. Conversely, DYRK2 was downregulated in lung adenocarcinoma and squamous cell carcinoma [135], diffuse large B-cell lymphoma [136], CRC [137], hepatocellular carcinoma (HCC) [138,139] and high-grade glioma [140]. Our analysis of TCGA data found *DYRK2* to be overexpressed in eight tumor cohorts: bladder (BLCA), breast (BRCA), esophagus

(ESCA), kidney (KIRC and KIRP), liver (LIHC), lung (LUAD and LUSC) and stomach (STAD) (Table S1). An analysis of CNAs indicated that *DYRK2* upregulation might be associated with gene amplification in BLCA, BRCA, LUAD, LUSC and STAD (Table S1). Moreover, a significant *DYRK2* overexpression in *DYRK2*-amplified tumors was also observed in HNSCC, ovary (OV), melanoma (SKCM) and sarcoma (SARC) (Table S1). Notably, no correlation was observed between the DYRK2 protein and mRNA levels in breast cancer tissue when compared with healthy tissue [141]. A similar discrepancy between the DYRK2 protein and mRNA was also detected in liver and lung cancers between a published protein data analysis and our TCGA analysis of mRNA changes (Table S1). Hence, post-transcriptional mechanisms may play a crucial role in determining the levels of the DYRK2 protein in tumor cells, and these might explain, at least in part, the conflicting data obtained in relation to breast cancer (see below).

Besides the alterations to the DYRK2 expression, it has been proposed that this kinase may represent a prognostic marker for different types of cancer, based on a correlation analysis between the gene/protein expressions and distinct clinical features like the degree of malignancy, relapse, response to chemotherapy or patient survival. Thus, higher DYRK2 levels were positively correlated with a more favorable prognosis and better response to chemotherapy in lung and bladder cancer patients [142–144] and with better survival in patients with CRC liver metastases [145]. Likewise, a weaker DYRK2 expression was associated with a worse prognosis in ovarian serous adenocarcinoma [146], CRC [137,145], HCC [138,139], glioma [140] and non-Hodgkin's lymphoma [136] patients. The situation in breast cancer is less clear, with conflicting results. As mentioned above, such discordance may be due to the use of mRNA or protein to assess the DYRK2 expression. As such, DYRK2 protein levels were shown to inversely correlate with tumor invasiveness [56], and enhanced 10-year disease-free survival was evident in DYRK2-positive breast cancer patients when compared to DYRK2-negative patients [147], while a stronger DYRK2 mRNA expression was associated with a worse prognosis in another study of breast cancer patients [148]. Apart from this, and in general, it appears that the weaker the expression of DYRK2, the worse the prognosis. This model is consistent with experimental data when DYRK2 levels are manipulated in carcinoma cell lines (ovary, CRC and HCC) that are then used as xenografts in mice, whereby DYRK2 gene silencing confers an enhanced proliferative capacity and metastatic potential in vivo [139,145,146]. However, again, some discordant phenotypes have been described in vivo when studying DYRK2-depleted breast cancer cell lines.

A few works have explored the mechanisms underlying the reduction of DYRK2 levels in tumor cells. The Kruppel-like factor 4 transcription factor has been shown to repress DYRK2 expression, acting directly on the DYRK2 promoter in chronic myeloid leukemia (CML) cell lines and mouse models, thereby favoring tumor progression [149]. Moreover, the DNA-methyltransferase 1-dependent methylation of the *DYRK2* promoter provokes transcriptional downregulation that may influence DYRK2 expression in CRC cells [150]. DYRK2 protein levels are also modulated by several E3 ubiquitin ligases, including seven in absentia homolog 2 (SIAH2) and MDM2 [9,151]. Interestingly, the DDR protein kinase ATM is involved in this process by phosphorylating DYRK2 and, thus, preventing DYRK2 degradation mediated by MDM2 [9,151]. This relationship could contribute to the ability of these E3 ubiquitin ligases to promote survival in states of hypoxia and in the face of DNA-damaged stress, respectively, by suppressing the proapoptotic activities of DYRK2. In particular, mutual regulation has been described for SIAH2 and DYRK2 [151]; indeed, an increase in the SIAH2 protein has been observed in lung cancer tissue and linked to DYRK2 downregulation [135]. Besides the alterations to the cellular levels of DYRK2, changes in the substrate selectivity have been seen in relation to Snail in ovarian cancer, with DYRK2 phosphorylation prevented by the prior p38-mediated phosphorylation of Snail [152].

#### *6.2. The Molecular Mechanisms Underlying the Role of DYRK2 in Cancer Cells*

Several clues have been obtained regarding the putative molecular mechanisms responsible for DYRK2-mediated tumor development/progression. Thus, DYRK2 activity appears to affect crucial processes like the cell cycle, DDR, epithelial-to-mesenchymal transition (EMT), the xenobiotic response system and cellular proteostasis [127]. The activity of DYRK2 has often been linked to its ability to negatively regulate the stability of its target proteins—in particular, through its interaction with the UBR5/EDD-DNA damage-binding protein 1 (DDB1)-DDB1- and cullin 4-associated factor homolog 1 (DCAF1/VPRBP) (EDVP) E3 ubiquitin ligase complex [38] (Figure 5A). The relevance of this interaction is highlighted by the alterations in the assembly of the EDVP complex detected in the analysis of certain DYRK2 mutants found in cancer samples [153]. In addition, it is worth noting that many of the proteins that are degraded following DYRK2 phosphorylation are targets of the tumor suppressor F-box/WD repeat-containing protein 7 (FBXW7) (Figure 5A), suggesting possible cross-talk with E3 ubiquitin ligase protein complexes made up of this factor. As previously mentioned, DYRK2 and SIAH2 cellular levels inversely correlate [151], further supporting a regulatory cross-talk between DYRK2 and several E3 ubiquitin ligases. Moreover, phosphorylation of the 19S subunit PMSC4/Rpt3 [148] might also contribute to the DYRK2-dependent modulation of protein accumulation.

**Figure 5.** Cancer-associated activities of DYRK2. (**A**) DYRK2 substrates are associated with different aspects of tumorigenesis, including proliferation (Myc, c-Jun, centrosomal protein of 110 kDa (CP110) and katanin: [38,56]); transformation (TERT: [154]); invasiveness (Snail: [141]); signaling (mTOR, Notch and Gli2/3: [34,64,155]) or the xenobiotic response system (pregnane X receptor (PXR): [156]). Red lines mark those substrates that are degraded when DYRK2 associates with the multicomponent E3 ubiquitin ligase EDVP. The rest (black lines) are all targets of FBXW7, considered to be a tumor suppressor [157]. (**B**) DYRK2 has been proposed as both a protumorigenic factor, as well as a tumor suppressor, in breast cancer. On the one hand, reduced levels of DYRK2 enhance the accumulation of mitogenic transcription factors like c-Jun and Myc, as well as the epithelial-to-mesenchymal transition (EMT)-promoting factor Snail, which is correlated with tumor progression and more aggressiveness. On the other hand, DYRK2 phosphorylates and positively regulates the 26S proteasome, promoting triple-negative breast cancer (TNBC) cell growth, and thus, DYRK2 depletion or inhibition impairs tumor growth in vivo.

In conjunction with the reduced expression of DYRK2 in tumor samples, DYRK2 depletion promotes the proliferation of cell lines originating from distinct tumor types, including breast, lymphoma, osteosarcoma, CRC and HCC [56,136–139,145,158], suggesting that DYRK2 may acts as a brake on proliferation. In this regard, DYRK2 phosphorylates the oncogenic pro-proliferative transcription factors c-Jun and Myc, increasing their rate of degradation [56]. Indeed, DYRK2 levels negatively correlated with c-Jun/Myc levels in breast tumor tissues [56] (Figure 5B). Other DYRK2 targets associated with cell cycle regulation are the centrosomal proteins katanin p60 and CP110/CCP110 and the telomerase TERT (Figure 5A), although no specific link between these proteins and DYRK2-dependent tumorigenic processes has as yet been proposed [38,154,159].

Besides the cell cycle, DYRK2 also regulates cell factors involved in other processes crucial for tumor progression, such as apoptosis or DDR. The interaction between DYRK2 and the E3 ubiquitin ligase RNF8 was proposed to influence DYRK2 recruitment to the DNA repair machinery [160], and the phosphorylation of p53 by DYRK2 promotes apoptosis in response to DNA damage, with ATM acting upstream by increasing the DYRK2 nuclear accumulation [9,19]. The modulation of p53 and Myc was also proposed as a DYRK2-mediated mechanism in leukemia stem cells and CML cell lines [149]. The putative regulatory activity during DDR and/or the ability of DYRK2 to increase components of the xenobiotic response system, such as the PXR/NR1I2 nuclear receptor [156] (Figure 5A), may contribute to enhance the resistance to chemotherapy drugs observed upon DYRK2 silencing [64,138,146]. A reduction in DYRK2 has also been linked to the enhanced migration and invasion of breast, glioma and ovary cancer cell lines [56,64,140,141,146]. In this regard, DYRK2 phosphorylates the EMT transcription factor Snail, priming it for ubiquitination-mediated degradation [141] (Figure 5A), which provides additional evidence that DYRK2 prevents the activation of aggressive phenotypes in breast and ovarian cancer cells.

To date, the most controversial role for DYRK2 associated with tumors is in breast cancer (Figure 5B). Based on results from xenograft experiments using MCF-7 cells, a tumor-suppressor role was first proposed given that DYRK2 silencing favored tumor growth [56]. The enhanced expression of direct DYRK2 targets like c-Jun or Myc, and/or other proteins like CDK14, could account for this phenotype [56,158]. Similarly, DYRK2 silencing increased the invasion, metastasis [141] and breast cancer cell stemness [161]. Conversely, using a clustered regularly interspaced short palindromic repeats (CRISPR)-based approach to generate DYRK2-knock out MDA-MB-468 breast cancer cells, DYRK2 was seen to promote breast cancer cell proliferation and tumor growth in xenografts. This effect could be mediated by the DYRK2-dependent phosphorylation of the proteasomal 19S subunit PMSC4/Rpt3 [148] (Figure 5B). In this context, two DYRK2 inhibitors, the natural drug curcumin and the small-molecule LDN192960, impaired cell proliferation and invasion and induced apoptosis in multiple myeloma and TNBC cell lines [134,162]. Whether these contradictory results arise from the use of cell lines with different responsiveness to estrogen/progesterone and/or an "addiction" to proteasome activity must be further explored. In any case, the DYRK2-associated stratification of breast tumors should be properly studied before designing any DYRK2-targeting therapeutic approach.

#### **7. DYRK3 and DYRK4**

The contribution of DYRK3 and DYRK4 to tumorigenesis is less clear, with very little evidence for the participation of DYRK3 and almost no evidence for that of DYRK4. This lack of information also mirrors the limited knowledge of the biological activities of these two family members.

DYRK3 was initially described as a kinase involved in erythroid development [24,163], although its most relevant activity described to date is the ability to regulate phase-transition during mitosis, thereby mediating the formation of multiple liquid-unmixed compartments such as stress granules, an essential process for proper mitotic division [16,17]. The association of DYRK3 with the mTORC1 pathway was established through the ability of DYRK3 to phosphorylate PRAS40, thereby promoting mTORC1 activity [16].

Our analysis of the TCGA data did not reveal any specific trend for DYRK3, which is under-expressed in breast (BRCA), kidney (KIHC), lung (LUAD and LUSC), prostate (PRAD) and thyroid (THCA) tumor cohorts and overexpressed in colon (COADREAD), HNSCC, kidney (KIRC) and stomach (STAD) cancer tissues (Table S1). Likewise, no particular trend can be found in the literature. Thus, *DYRK3* mRNA was found significantly increased in highly invasive NSCLC cell lines compared with low invasive lines [164], while a strong DYRK3 expression was positively correlated with survival in glioma patients [165]. Moreover, DYRK3 was proposed as a specific early-stage tumor driver in gastric cancer [166]. Finally, a reduction in the DYRK3 protein was recently described in HCC biopsies relative to normal tissue, and low DYRK3 levels were associated with a poor prognosis in this type of cancer [167]. In addition, manipulating the DYRK3 expression in HCC cells demonstrated an

inverse correlation with proliferation rates both in vitro and in tumor xenograft models, as well as with the metastatic potential of the tumor cells, further evidencing that DYRK3 fulfills a tumor-suppressor role in this type of cancer [167]. Indeed, a regulatory axis was proposed that involves the ATF4 transcription factor and its coactivator NCOA3 as a direct DYRK3 substrate, regulating the expression of key metabolic enzymes in the purine synthesis pathway that are relevant to HCC progression. However, whether this role for DYRK3 can be extrapolated to other tumors remains to be confirmed.

DYRK4 is the DYRK family member associated with the least significant alterations in the TCGA cohorts analyzed. We found that it was downregulated in lung (LUAD), prostate (PRAD) and stomach (STAD) cohorts and with different patterns of expression in the three kidney cohorts: overexpressed in KIRC and KIRP and downregulated in KIHC (Table S1). Interestingly, a recent high-throughput screen on 313 kinase-deficient cell lines revealed that DYRK4 knockout cells were among the most sensitive to agents that produce DNA damage [168], suggesting that DYRK4 might merit further exploration as a putative target to enhance chemotherapy toxicity on cancer cells.

#### **8. DYRK Inhibitors as Antitumor Therapies**

Chemical compounds that bind and functionally block protein kinases have been studied extensively and employed as antitumor agents, both in research and in clinical trials [83]. Although the role of DYRK family members in tumorigenesis and tumor progression has not been fully elucidated, pharmacological inhibitors of DYRK kinases have been tested in laboratories for their antimalignant activity, and a few of them are already undergoing clinical trials.

In the case of DYRK1A, the search for both naturally occurring and synthetic inhibitors has been extensive given that DYRK1A may be a potential pharmacological target not only in cancer but, also, in neurodegenerative diseases (reviewed in [43]), DS [169–172] and diabetes (reviewed in [173]). DYRK1A inhibitors have been comprehensively reviewed elsewhere [174–176], and so, we will only refer to the orally bioavailable archetypic DYRK1A inhibitors in tumor contexts. For instance, the anticancer properties of green tea and its derivatives have been proven in many animal models, a product that contains the natural DYRK1A inhibitor Epigallocatechin-3-gallate (EGCG). However, EGCG can potentially target many different intracellular pathways [177], making it difficult to assign particular effects to DYRK1A inhibition. Additionally, the ß-carboline alkaloid harmine selectively inhibits DYRK1A and—albeit, less efficiently—other members of the family [178,179], and it has been reported to have cytotoxic effects on cancer cell lines [66,85,180,181] and antitumor effects in vivo in glioma and in PDAC xenograft experiments [66,68], as well as synergistic effects with other chemotherapeutic agents [79,80,182]. However, the neurotoxic effects of harmine due to the targeting of monoamine oxidase A rule against its use in humans. Therefore, the search for harmine derivatives with enhanced antitumor activity and reduced neurotoxic effects has been intense in recent years [183–185]. Finally, the synthetic DYRK1A inhibitor INDY, proven to modulate the phenotypic effects of DYRK1A overexpression in vivo [186], has been shown to improve the response of ovarian cancer spheroids to carboplatin [79].

Compounds targeting DYRK1B, with either restricted or broad specificity, have been used as research tools, and they display toxicity towards several types of cancer cells or they promote the cell cycle re-entry of quiescent tumor cells (reviewed in [94]). The latter would enhance the effectiveness of other antiproliferative drugs in combinatorial approaches. For instance, the DYRK1B inhibitor AZ191 [52] increases the anticancer effects of doxorubicin in liposarcoma cell lines [96] or sensitizes the PDAC cell lines to mTOR inhibition [115]. However, AZ191 has been also shown to counteract the antitumor effects of the lysosome inhibitor Bafilomycin A1 in HCC cell lines [111]. For DYRK2, experimental data on the antitumor effects of the natural DYRK2 inhibitor curcumin and of the synthetic compound LDN192960 was obtained in both in vitro and in vivo models of TNBC and multiple myeloma, supporting the hypothesis that DYRK2 is a promising pharmaceutical target in these malignancies [134,162]. Finally, better understanding the role of DYRKs in tumor cells has proven valuable by helping to identify combinatorial therapeutic approaches, as in the cases of the DYRK1B

inhibitors that enhance the inhibitory efficiency of MEK and mTOR [107,109,187] or DYRK2 inhibition sensitizing MDA-MB-468 cells to the proteasome inhibitor bortezomib [148].

Most kinase inhibitors lack complete specificity [178,188], a potentially negative property that might be exploited in multitargeting strategies, which become a familiar situation in antitumor therapies. Interestingly, the only inhibitors of the DYRK family members currently being screened in clinical trials were identified as inhibitors of other protein kinases. In particular, compound CX-4945 was initially identified as a casein kinase 2 inhibitor, but it was subsequently shown to be a potent DYRK1A and DYRK1B inhibitor [171], and it is currently in phase I and II clinical studies for medulloblastoma, cholangiocarcinoma and basal cell carcinoma (NCT02128282, NCT03904862 and NCT03897036). Recently, OTS167, a chemical initially described as a maternal embryonic leucine zipper kinase inhibitor, has been proven to have potent anti-DYRK1A activity [189]. OTS167 is currently being assessed in clinical trials for the treatment of advanced breast cancer and TNBC (phase I) and for multiple types of leukemia, including AML and advanced CML (phase II: NCT02795520). Finally, two other DYRK inhibitors have been assessed in clinical trials for non-neoplastic disorders: GSK-626616 [16] completed a phase I clinical trial to evaluate its action on anemia (NCT00443170), and lorecivivint, a potent CLK2 inhibitor that also inhibits DYRK1A [190], is being studied in a phase II trial for the treatment of moderate-to-severe symptomatic osteoarthritis (NCT03706521). Thus, they could be repurposed in trials for the treatment of specific cancer types.

#### **9. Conclusions**

In the last decade, more experimental evidence indicates that DYRK protein kinases are a novel class of "kinase-of-interest" in cancer. However, this evidence mostly comes from studies exploring DYRK expressions in tumor tissues and/or the phenotypic changes triggered by manipulating the DYRK protein in cancer cell lines. These data not only provide a partial and confusing picture of the influence of DYRKs in tumor initiation and progression, but also, they highlight the many questions that still need to be addressed. In particular, it remains unclear which molecular pathways are regulated by DYRKs in different tumor types and which of them selectively trigger cells to engage in neoplastic transformation or enhance the malignant phenotype of tumor cells. Resolving these issues will not only help understand the biology behind the activity of these kinases, but also, it will provide a basis for the rational design of therapeutic approaches based on inhibitors. In this regard, while incomplete, the currently available data provides precious information on which forthcoming therapeutic approaches may be based. Therefore, the tumor types in which downregulation of the DYRK kinase has been associated with increased tumor growth and/or invasiveness should not be considered for trials with DYRK inhibitors. Conversely, inhibitors targeting DYRK family members that are known to favor the tumorigenesis of specific tumor types should be considered for such trials. Nevertheless, putative side effects due to the inhibition of members that are essential to maintaining cellular homeostasis in normal cells, such as the dosage-sensitive DYRK1A or DYRK1B, should be carefully monitored. In this context, engineering drugs to increase their specificity, exclusively targeting proliferating cells, would be worthwhile. Finally, and considering the differential and sometimes opposite roles of distinct DYRK kinases in tumor progression, selectivity towards a specific member of the family is crucial and, at the same time, very challenging, particularly given the strong structural similarity of the catalytic domain. Smart solutions might include an allosteric drug design or other additional efforts to increase compound selectivity.

To conclude, many important advances in understanding how the dysregulation of DYRK protein kinases is associated to pathological phenotypes in humans have been made in recent years—in particular, in terms of the involvement in DYRK cancers. Still, many secrets behind the oncogenic or protective potential of DYRK kinases remain to be revealed, and we anticipate that the field will continue to grow for the foreseeable future.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/12/8/2106/s1: Table S1. Differential expression of DYRK genes in TCGA tumor samples.

**Author Contributions:** C.R.-P. and N.L.-B. analyzed the TCGA data. J.B. created the figures and Table S1. S.D.L.L. created Table 1. J.B. and S.D.L.L. wrote the manuscript with input from C.F. All the authors have read and approved the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** J.B. is an FPI predoctoral fellow (BES-2014-069983). De la Luna's lab is supported by grants from the Spanish Ministry of Science and Innovation (BFU2016-76141-P, AEI/FEDER), the AGAUR grant from Secretaria d'Universitats i Recerca del Departament d'Empresa i Coneixement de la Generalitat de Catalunya (SGR14/674) and the CIBER de Enfermedades Raras. We thank La Marato of TV3 for its support to our research in cancer. We also acknowledge the support of the Spanish Ministry of Science and Innovation to the EMBL partnership, the Centro de Excelencia Severo Ochoa and the support of the CERCA Programme/Generalitat de Catalunya.

**Acknowledgments:** We are grateful to all members of Susana de la Luna's laboratory for their helpful discussions, and we thank Mark Sefton for English language editing. The authors acknowledge the efforts of the DYRK community and apologize to the investigators whose works do not appear in this review and should have been included.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **The Relevance of the SH2 Domain for c-Src Functionality in Triple-Negative Breast Cancer Cells**

**Víctor Mayoral-Varo 1,† , María Pilar Sánchez-Bailón 1,2,† , Annarica Calcabrini 1,3,† , Marta García-Hernández <sup>4</sup> , Valerio Frezza <sup>4</sup> , María Elena Martín 4 , Víctor M. González <sup>4</sup> and Jorge Martín-Pérez 1,5,\***


**Simple Summary:** Triple-Negative breast cancers (TNBC) have not specific therapeutic targets and are considered the most aggressive mammary tumors. c-Src controls several cellular processes: proliferation, differentiation, survival, motility, and angiogenesis. Alteration of c-Src functionality, by increasing its expression and/or its kinase activity, is associated to progression and metastasis of tumors in mammary gland, pancreas, colon, brain, and lung. However, c-Src tyrosine kinase inhibitors alone are not fully clinically effective, suggesting that c-Src adapter SH2/SH3 domains may be important. We questioned whether the SH2-c-Src domain is relevant for tumorigenicity of TNBC SUM159 and MDA-MB-231 human cell lines. Conditional expression of SH2 and SH3 inactivating mutants in these TNBC cells, or transfection of aptamers directed to SH2, allowed us to show that this domain is required for their tumorigenesis. Therefore, the SH2-c-Src domain could be a promising therapeutic target that, combined with c-Src kinase inhibitors, may represent a novel therapeutic strategy for TNBC patients.

**Abstract:** The role of Src family kinases (SFKs) in human tumors has been always associated with tyrosine kinase activity and much less attention has been given to the SH2 and SH3 adapter domains. Here, we studied the role of the c-Src-SH2 domain in triple-negative breast cancer (TNBC). To this end, SUM159PT and MDA-MB-231 human cell lines were employed as model systems. These cells conditionally expressed, under tetracycline control (Tet-On system), a c-Src variant with pointinactivating mutation of the SH2 adapter domain (R175L). The expression of this mutant reduced the self-renewal capability of the enriched population of breast cancer stem cells (BCSCs), demonstrating the importance of the SH2 adapter domain of c-Src in the mammary gland carcinogenesis. In addition, the analysis of anchorage-independent growth, proliferation, migration, and invasiveness, all processes associated with tumorigenesis, showed that the SH2 domain of c-Src plays a very relevant role in their regulation. Furthermore, the transfection of two different aptamers directed to SH2-c-Src in both SUM159PT and MDA-MB-231 cells induced inhibition of their proliferation, migration, and invasiveness, strengthening the hypothesis that this domain is highly involved in TNBC tumorigenesis. Therefore, the SH2 domain of c-Src could be a promising therapeutic target and combined treatments with inhibitors of c-Src kinase enzymatic activity may represent a new therapeutic strategy for patients with TNBC, whose prognosis is currently very negative.

**Citation:** Mayoral-Varo, V.; Sánchez-Bailón, M.P.; Calcabrini, A.; García-Hernández, M.; Frezza, V.; Martín, M.E.; González, V.M.; Martín-Pérez, J. The Relevance of the SH2 Domain for c-Src Functionality in Triple-Negative Breast Cancer Cells. *Cancers* **2021**, *13*, 462. https:// doi.org/10.3390/cancers13030462

Academic Editor: Francisco M. Vega Received: 30 December 2020 Accepted: 19 January 2021 Published: 26 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

**Keywords:** triple-negative breast cancer (TNBC); c-Src; SH2 domain; inactivating point mutation; aptamers

#### **1. Introduction**

The Src family of non-receptor tyrosine kinases (SFKs) is composed of nine members, and it has a modular structure, containing the SH2 and SH3 (Src homology domains 2 and 3), which are involved in protein-protein interactions with tyrosine phosphorylated proteins or with proteins containing proline rich sequences, respectively [1,2]. These two domains are also present in many other adapter and regulatory proteins, and facilitate the formation of intracellular signaling complexes [3]. While the activity of c-Src, the prototype of the SFKs, is mainly modulated by phosphorylation, the SH2 and SH3 domains are also required for the conformational changes associated with its cellular distribution, kinase activity, and cell functionality [4–7]. c-Src plays a key regulatory role of many cellular processes, including proliferation, differentiation, survival, motility, and angiogenesis. Therefore, alteration of c-Src functionality, by increasing its expression and/or its kinase activity, has been associated to progression and metastasis of tumors in the mammary gland, pancreas, colon, and lung [1,5–8].

Breast tumors are diverse, and they have been classified according to their genetic and histological characteristics [9,10]. Among them, the basal triple negative breast cancer (TNBC) does not express estrogen and progesterone receptors (ER−, PR−), nor overexpresses HER2, and usually has an inactive p53 mutant [10–12]. TNBCs do not have specific therapeutic targets and are considered the most aggressive mammary tumors, with a tendency to metastasize mainly into the lung, brain, and bone [13,14].

Experimentally, inhibitors of c-Src kinase activity, or its suppression, block proliferation, survival, migration, and invasion, as well as tumorigenesis in vivo [15–17]. However, inhibitors of c-Src tyrosine kinase activity alone do not appear to be fully clinically effective [8,18], suggesting that its adapter domains may play an important role for c-Src functionality in tumorigenesis. In this context, expression of c-Src with point-inactivating mutations at either SH2 or SH3 domains, which conferred stimulation of its kinase activity, blocks the prolactin-induce activation of Jak2 in MCF7 [19]. In mouse SYF fibroblasts, expression of c-Src-R175*L* prevents Fak auto-phosphorylation (pY397), malignant transformation, motility defects, and focal adhesion formation, indicating the relevance of the SH2 domain of c-Src [20]. The SH2 domain of c-Src interacts with the pY397-Fak facilitating the open conformation of c-Src that activates its kinase activity and, in turn, protects pY397- Fak from phosphatases [21,22]. In addition, c-Src phosphorylates Fak on several tyrosine residues, thus promoting cellular signaling and tumor progression [6,23,24].

Small molecules, such as inhibitory peptides and non-peptides, have been used to block the SH3/SH2 domains of c-Src [25–29] with a relative success. Aptamers are single stranded oligonucleotides (DNA or RNA) that bind to proteins with high affinity and specificity, blocking their functionality. They have been used for diagnosis and therapy in several infectious, inflammation, vascular diseases, as well as in other pathologies including breast cancer [30–32].

Here, we analyzed the role of the adapter domain of c-Src in the in vitro tumorigenic properties of SUM159PT (from now on SUM159) and MDA-MB-231 TNBC cell lines. We found that the conditional expression of c-Src variants with suppression of SH2 functionality caused profound effects on the behavior of these triple negative cell lines. Consistently, two different aptamers directed to SH2-c-Src inhibited proliferation, migration, and invasiveness of both SUM159 and MDA-MB-231 cells. Thus, the SH2-c-Src domain appears to play a crucial role in TNBC tumorigenesis.

#### **2. Results**

#### *2.1. c-Src Variants of the SH2 Adapter Domain*

In the studies presented here we used two different triple negative breast cancer (TNBC) cell lines, SUM159 and MDA-MB-231. Although SUM159 and MDA-MB-231 are both Basal-Mesenchymal TNBC cell lines with a spindle phenotype, they show differences in deleted and mutated genes. Furthermore, previously published data from the laboratory using both SUM159 and MDA-MB-231 cells showed that they differ in some signaling responses [16]. All together, we can conclude that even if both are representing TNBC cells, their cellular behavior could diverge.

To analyze the role of the SH2 adapter-domain of c-Src in the in vitro tumorigenic properties of SUM159 and MDA-MB-231 cell lines, we conditionally expressed (Tet-On system) chicken c-Src variants with point mutations inactivating this domain (Figure 1A). It should be pointed out that chicken c-Src could replace human c-Src functionality [16], as they have more than 94% identity at the amino acid sequence [33]. Nevertheless, the EC10 mouse monoclonal antibody (Millipore, no. 05-185) specifically recognizes chicken c-Src, making it possible to determine by Western blot (WB) the expression of c-Src variants in the presence of the endogenous human c-Src of SUM159 and MDA-MB-231 cells.

We used the R175*L* point mutation (SH2) affecting intra- and inter-molecular c-Src interactions, preventing c-Src from being in the close configuration, consequently, c-Src-R175*L* has a constitutive tyrosine-kinase activity [34]. The c-Src-SH2-SH3 double variant (c-Src-W118*A*-R175*L*) [19], which also has a constitutive tyrosine-kinase activity, was also employed to test for the impact of the SH3 domain in SH2 functionality. To avoid undesirable effects due to the insertion of the c-Src variants in the genome of the cells upon transfection, each cell line was made of a pool of, at least, three positive independent clones. Since we are studying the role of SH2 domain in TNBCs SUM159 and MDA-MB-231, we expressed the wild type c-Src (c-Src-wt) to compare the results obtained with the SH2 mutant. As observed in Figure 1B, the chicken c-Src variants were induced by the addition of Doxy to culture media (0.2 µg/mL, 72 h) at similar levels in SUM159 and MDA-MB-231. Analyses of the degree of activation of c-Src chicken variants by determining the autophosphorylation at Y418 showed that the expression of c-Src-wt slightly increased the levels of phosphorylation at Y418 in both cell lines. In contrast and consistent with the scheme presented in Figure 1A, expression of both c-Src-R175*L* and c-Src-W118*A*/R175*L* showed a high degree of phosphorylation at Y418, which agrees with the stimulated tyrosine kinase activity of these mutants (Figure 1B).

SFKs expression is associated with a different outcome in breast cancer patients [35]. Thus, we decided to determine the protein levels of several SFKs members in both cell models by WB. Expression of c-Src and Fyn was higher in SUM159 than in MDA-MB-231 cells, while the contrary was observed for c-Yes. Regarding Lyn protein, SUM159 cells showed higher Lyn A levels than MDA-MB-231, whereas the opposite situation was observed for Lyn B expression (Figure 1C). Expression of c-Src variants did not alter that of endogenous Src-kinases (data not shown). Therefore, we used two TNBC cell lines that exhibit a different expression pattern of SFKs, which better represent the variability of TNBC.

**Figure 1.** c-Src variants and expression of Src kinases in SUM159 and MDA-MB-231 cells. (**A**) Schematic design of c-Src and the variants employed in this study, which were conditionally expressed (Tet-On system) upon addition of doxycycline (Doxy, 0.2 µg/mL) to the cell culture. The R175*L* mutation inhibits both intra- and inter-molecular interactions of the SH2 domain of c-Src. The W118*A*/R175*L* double mutation inhibits both the SH2 and SH3 domains. (**B**) Induction of chicken c-Src variants by Doxy was detected by Western blot (WB) using the EC10 mouse monoclonal antibody that specifically recognizes chicken c-Src and β-Actin as a loading control. (**C**) Comparative expression of Src kinases in SUM159 versus MDA-MB-231 cells determined by WB. Actin was used as a loading control and the ratio of kinase/actin in SUM159 was considered as 1.

#### *2.2. SH2 Domain of c-Src Is Important for In Vitro Breast Cancer Stem Cell-Renewal*

Within a tumor, there is a small portion of the tumor-mass (1–2%) derived from the stem-cell population that by mutations acquired tumorigenic properties. These breast cancer stem cells (BCSCs) are slow dividing and capable of regenerating a tumor upon transplantation in nude mice [35]. To determine the self-renewal capacity of the enriched population of BCSCs, we performed the mammosphere formation assay during three generations to define, at the third generation, the sphere formation efficiency (SFE, see Materials and Methods) [36–38]. We observed that SFE increased in all SUM159 and MDA-MB-231 cells expressing c-Src variants, indicating the enrichment of BCSCs population (Figures S1 and S2). Nevertheless, it should be noticed that SUM159 contained higher numbers of mammospheres than MDA-MB-231. We then analyzed the effect of c-Src variants described above (Figure 1A) on the BCSCs renewal ability of each cell line SUM159 or MDA-MB-231 expressing c-Src-mutants and compared it with the c-Src-wt. The functionality of SH2 and SH3 adapter domains appeared to be relevant for SFE, as both

c-Src-R175*L* and c-Src-W11A/R175L significantly reduced the self-renewal of the enriched population of MDA-MB-231 and SUM159-BCSCs (Figure 2A). start a new page without indent 4.6cm

**Figure 2.** Role of c-Src variants in the self-renewal of breast cancer stem cells (BCSCs) derived from SUM159 and MDA-MB-231. The evaluation of self-renewal was determined by the sphere formation efficiency (SFE) in the enriched subpopulation of BCSCs derived from SUM159 and MDA-MB-231 (**A**). The SFE was measured at the third generation of mammospheres (see Materials and Methods). Each experiment was measured in triplicates (*n* = 3) and repeated three times. Results were expressed as a percentage of the mean ± standard deviation (SD). The statistical significance is referred to those obtained from c-Src-wt, \* *p* < 0.5, \*\* *p* < 0.01. Quantitative analyses of stem cell markers ALDH1, NANOG, and Oct3/4 expression by WB in SUM159 and MDA-MB-231 cells conditionally expressing c-Src variants (**B**). Results represented data obtained from three independent WB (*n* = 3) using β-Actin as a loading control expressed as a percentage of the mean ± SD, and were referred to those obtained from cells expressing c-Src-wt considered as 1. The statistical significance is referred to cells expressing c-Src-wt, \* *p* < 0.5, \*\* *p* < 0.01.

We have also found that the induction of SrcDN (c-Src-K295*M*/Y527*F*, which is devoid of catalytic activity but with functional SH2 and SH3 domains), a functional mirror image of c-Src-W118*A*/R175*L*, significantly reduced SFE in both SUM159 and MDA-MB-231. Furthermore, the endogenous c-Src function is required for SFE in MDA-MB-231 cells, as its conditional suppression [16] inhibited SFE (Figure S3).

Altogether, the findings indicate that the three domains are necessary for self-renewal, as the alteration of only one of them reduced the sphere formation ability of both TNBC cells.

Consistent with the reduction of the enriched population of BCSCs induced by the expression of c-Src-R175L and c-Src-W118A/R175L c-Src mutants in both SUM159 and MDA-MB-231 cells, we analyzed the expression of ALDH1 by WB, a stem cell marker [36–39]. The results showed that in the MDA-MB-231 disruption of the functionality of c-Src SH2 domain by R175*L* mutation inhibited the ALDH1 level as compared to c-Src-wt (Figure 2B and Figure S4). Disruption of the functionality of c-Src SH2 and SH3 domains by the double mutant c-Src-W118*A*/R175*L* clearly reduced ALDH1 levels in either SUM159 or MDA-MB-231 (Figure 2B and Figure S4). Expression of NANOG and Oct3/4 was reduced in SUM159 cells expressing c-Src-R175L and c-Src-W118A/R175L mutants as compared to the expression of c-Src-wt (Figure 2B and Figure S4). In contrast, in MDA-MB-231, only the induction of c-Src-W118A/R175L mutant reduced the levels of Oct3/4, while none of the c-Src mutants altered the expression of NANOG (Figure 2B and Figure S4). Furthermore, the levels of NANOG and Oct3/4 were reduced in SUM159 and MDA-MB-231 cells following SrcDN expression or suppression of the endogenous c-Src (Figure S3).

Collectively, these results indicate that the SH2-c-Src domain is relevant for renewal of the enriched population of BCSCs in SUM159 and MDA-MB-231 cells.

#### *2.3. Role of Adapter Domains in Anchorage-Independent Growth*

Anchorage-independent growth correlates with cellular tumorigenic and metastatic potential, a typical feature of in vivo TNBC aggressive phenotype. Thus, we analyzed the role of the SH2 adapter domain of c-Src in this event by determining cellular growth in soft agar. In SUM159 cells, induction of the c-Src-R175L mutant did not alter the colony formation as compared to the wild type. In contrast, mutation of the SH2 and SH3 domains together appeared to be relevant for colony formation in soft agar, as expression of c-Src-W118*A*/R175*L* significantly reduced the number of colonies in the agar (Figure 3A and Figure S5). In MDA-MB-231 cells, Doxy induction of either c-Src-R175*L* or c-Src-W118*A*/R175*L* mutants significantly inhibited soft-agar colony formation, as compared to c-Src-wt (Figure 3A and Figure S5). Concurrently, these results support the role of the SH2-c-Src domain in anchorage-independent growth.

**Figure 3.** Modulation of anchorage-independent growth and cell proliferation by expression of c-Src variants in SUM159 and MDA-MB-231 cells. (**A**) Colony formation in soft-agar was employed to determine the effect of c-Src variants expression in SUM159 and MDA-MB-231 cells in anchorage-independent growth (see Materials and Methods). Each experiment was measured in triplicates (*n* = 3) and repeated three times. Results were expressed as the mean ± SD of the number of colonies/plate, \* *p* < 0.5, \*\* *p* < 0.01. (**B**) Cell proliferation of SUM159 and MDA-MB-231 expressing c-Src variants was determined by the trypan blue exclusion assay (see Materials and Methods), \* *p* < 0.5, \*\* *p* < 0.01, \*\*\* *p* < 0.001. (**C**) Quantitative analyses of cell proliferation markers Myc, cyclin D1, and p27 expression by WB in SUM159 and MDA-MB-231 cells conditionally expressing c-Src variants. Results represented data obtained from three independent WB (*n* = 3) using either β-Actin, α-Tubulin, or GAPDH as a loading control expressed as a percentage of the mean ± SD, and were referred to those obtained from cells expressing c-Src-wt considered as 1. The statistical significance is referred to cells expressing c-Src-wt, \* *p* < 0.5, \*\*\* *p* < 0.001.

#### *2.4. c-Src-SH2 Domain Modulates Cellular Proliferation*

Several data show the relevance of c-Src in cell proliferation and survival [1]. Therefore, we evaluated the effect of c-Src variants in SUM159 and MDA-MB-231 cell proliferation. In SUM159, induction of the c-Src-R175L and c-Src-W118*A*/R175L significantly reduced cell proliferation compared to c-Src-wt (Figure 3B). Similar results were obtained in MDA-MB-231 cells. Nevertheless, the reduction of cell proliferation observed in SUM159 was higher than in MDA-MB-231, which was modest (Figure 3B). When proliferation was analyzed considering cells without Doxy-induction as the control (Figure S6), since c-Srcwt expression did not alter proliferation, we observed similar results to those obtained considering c-Src-wt as the control. Cell cycle analyses by propidium iodide showed that in these mutants the number of cells in "Sub-G1" increased, and "G2-M" was reduced (Figure S7), which may help in understanding the proliferation differences. Analyses of cell cycle by pulse/chase with BrdU and propidium iodide in SUM159-Tet-On-c-Src-W118A/R175L showed an increased number of cells in the G1 phase upon induction of this mutant versus control (Doxy), a reduction of "S", as well as in "G2/M". MTT analyses of these cells showed a 45% reduction of metabolic active cells, which may be related to the number of viable and proliferating cells [40]. These results agree with those observed here (Figure 3B). Together, these results suggest that both c-Src-R175L and c-Src-W118A/R175L variants reduced proliferation with no signs of toxic effects.

We then analyzed by WB the levels of Myc, cyclin D1, and p27kip1 cell cycle makers (Figure 3C). The functionality of SH2-c-Src domain appeared relevant as induction of c-Src-R175L and c-Src-W118*A*/R175L in SUM159 reduced Myc expression. In contrast, in MDA-MB-231, no significant variations were detected upon induction of c-Src variants (Figure 3C). Interestingly, in SUM159 cells, expression of c-Src-R175*L* and c-Src-W118A/R715L variants highly induced cyclin D1 (Figure 3C). In MDA-MB-231 cells, while c-Src-R175L did not modify cyclin D1 levels, the double mutant c-Src-W118A/R175L significantly reduced them. When the data were analyzed considering cells without Doxy induction as the control, the results showed the same tendency (Figure S6). The cyclin D1 gene regulation is complex and it varies between cell lines and experimental conditions. In rodent cells, it has been reported that Myc induces D1 in some cases while in others, Myc does not induce or even repress D1 (for review see [41]). Therefore, the downregulation of Myc observed in SUM159 cells may contribute to the upregulation of cyclin D1.

Furthermore, regarding the cell cycle inhibitor p27Kip1, induction of both c-Src-R175*L* and c-Src-W118*A*/R175*L* mutants significantly increased its expression in SUM159 and in MDA-MB-231 (Figure 3C and Figure S6).

These results showed that SUM159 cells seemed to be more sensitive than MDA-MB-231 to the alteration of SH2-SH3 adapter domain functionality, as demonstrated by the effects induced by the expression of mutants on cell proliferation and cell cycle marker levels. Therefore, the kinase activity has an essential role in cell proliferation as the SH2 domain just partially modulates cell proliferation.

#### *2.5. Regulation of Cellular Migration and Invasion by c-Src Adapter Domains*

Cell migration is one of the essential steps of the metastatic cascade. We analyzed by wound-healing assays the effect of c-Src variants expression in SUM159 and MDA-MB-231 cell migration. In SUM159, Doxy induction of either c-Src -R175L or -W118A/R175L did not significantly alter migration as compared to c-Src-wt expression, and as shown by the lack of significant difference in the remaining wound-healing area after 13 h of migration between the control and the two variants (Figure 4A and Figure S9). When the analyses were made considering unstimulated c-Src variants (Doxy) as the control, it was observed that c-Src-wt inhibited cell migration in SUM159, while induced it in MDA-MB-231 (Figures S8–S10). Nevertheless, the results showed the same tendency as observed when the c-Src-wt expression was used as the control. However, looking at the recording videos of migration, we observed abnormal movements of SUM159 cells expressing the c-Src-W118A/R175L variant. Thus, we further analyzed the migration pattern of individual

cells at the migration border by tracking the path of single cells. We found that only the expression of c-Src-W118*A*/R175*L* variant in SUM159 cells caused random migration, as the ratio of Euclidean/Accumulated distances was significantly reduced, while the velocity of cell migration was increased, as compared to the control (Figure 4A and Figure S9). Consequently, SUM159 cells expressing the c-Src-W118A/R175L variant did not close the wound-healing area more or faster than cells expressing c-Src-wt or c-Src-R175L (Figure 4A). Indeed, they were moving randomly, not all in the direction to close the wound area, and at higher velocity than the other cells. In MDA-MB-231 cells, mutations affecting the SH2 domain functionality (R175L and W118*A*/R175L) reduced migration compared to c-Src-wt (Figure 4A and Figure S9). However, none of these mutations caused random migration in MDA-MB-231 cells.

The oncogenic potential of c-Src in tumor cells is pleiotropic and controls cytoskeletallinked events, such as extracellular matrix-adhesion, migration, and invasion. We previously showed that the catalytic activity of this proto-oncogene is involved in invasion and migration [15]. Now, we analyzed whether the SH2 adapter domain of c-Src is involved in the regulation of SUM159 and MDA-MB-231 cellular invasion. In SUM159, expression of c-Src-R175*L* variant inhibited the number of invading cells, while the double SH2/SH3 mutant c-Src-W118A/R175L, was unable to modify the invasiveness of SUM159 cells, as compared to the c-Src-wt (Figure 4B). In contrast to SUM159, expression of c-Src -R175L and -W118A/R175L significantly inhibited cell invasion in MDA-MB-231 in comparison to c-Src-wt (Figure 4B). When the analyses were made considering unstimulated c-Src variants (Doxy) as the control, it was observed that c-Src-wt did not alter invasiveness in SUM159, while induced it in MDA-MB-231 (Figure S8). Nevertheless, the results showed the same tendency as observed when the c-Src-wt expression was used as the control.

The growth factor or integrin stimulation induces Fak autophosphorylation on Y397, generating a high affinity binding site for the Src SH2 domain [20,21], which in turn phosphorylates Fak at several tyrosine residues allowing the activation of multiple signaling pathways [42]. The association of Src with Fak controls the turnover of focal adhesion complexes, which are involved in cell motility, migration, and invasion [39]. These processes involve the dynamic control of protein associated with focal adhesion complex, among them, Fak, Paxillin, Caveolin 1, etc., due to, at least in part, their phosphorylation/activation [16,42,43]. We then analyzed the degree of activation/phosphorylation of these proteins involved in cell migration and invasion by WB. In SUM159, expression of c-Src-W118A/R175L increased Caveolin 1 levels (Figure S11A), while only c-Src-R175*L* increased pY14-Caveolin 1. Paxillin expression remained constant upon induction of c-Src variants, albeit pY118-Paxillin/Paxillin augmented in c-Src-R175L expressing SUM159 cells (Figure S11A). Fak expression was constant, whereas pY397 diminished upon induction of c-Src-R175L and c-Src-W118A/R175L compared to c-Src-wt. Phosphorylation of Fak at Y576 was significantly reduced in cells expressing c-Src-R175L, while it was increased by c-Src-W118A/R175L, as compared to c-Src-wt (Figure S11A). In MDA-MB-231, expression of Caveolin 1 remained constant for all c-src variants expressing cells (Figure S11B). The mutants c-Src-R175L and c-Src-W118A/R175L have an open conformation, consequently, they highly increased the activation of Caveolin 1 (Figure S11B). Expression of c-Src mutants did not alter the levels of Paxillin protein, while the activation of Paxillin (ratio p118Y-Paxillin/Paxillin) was surprisingly inhibited in MDA-MB-231 expressing c-Src -R175L and -W118A/R175L, as compared to c-Src-wt (Figure S11B). Fak protein levels were unaltered in any of the MDA-MB-231 cell lines expressing c-Src variants (Figure S11B). Conversely, Fak autophosphorylation at Y397 was increased by expression of W118A/R175L. The specific activity of Fak (pY576-Fak/Fak) was not augmented in all c-Src mutants expressed in MDA-MB-231 (Figure S11B) as compared to c-Src-wt. When the results of WBs were analyzed considering non-induced conditions for c-Src variants as the control (Doxy conditions), the results showed the same tendency (Figures S12 and S13). When all the c-Src variants expressing SUM159 and MDA-MB-231 cells were analyzed together in a single WB, the results showed that expression of total c-Src was higher in SUM159 than in MDA-MB-231,

supporting the data from Figure 1B, as it was observed for total Fak. Changes in pY397-Fak were not evident in any of the two cell lines. In contrast, phosphorylation of Fak by c-Src at Y576 increased upon expression of the mutants in both cell lines. However, results from triplicate experiments (Figures S11–S13) showed inhibition of pY576-Fak in SUM159 expressingc-Src-R175L. Possibly, this discrepancy is due to the fact that the analysis of Figure S14 represents a single sample, while data from Figures S11 and S12 represented the average of three different samples. Similarly, while the levels of total Akt practically unchanged in SUM159 and MDA-MB-231 cells due to the expression of c-Src variants, the extent of pS473-Akt was increased upon induction of all the c-Src variants (Figure S14).

We have also analyzed the effect of c-Src mutants in the cellular distribution of pY14- Caveolin 1 and pY418-Src in both SUM159 and MDA-MB-231 cells by confocal microscopy. In c-Src-wt or in c-Src-R175*L* expressing SUM159 cells, pY14-Caveolin 1 and pY418-Src co-localized at the focal adhesion sites (Figure S15), whereas in those expressing c-Src-W118*A*/R175*L*, distribution of pY418-Src did not fully co-localize with pY14-Caveolin 1 at focal adhesion sites (Figure S15 basal layer, and Figure S15 upper layer). However, if we focus at higher levels, we find that in both mutant expressing SUM159 cells pY14-Caveolin and pY418-Src decorated caveolae (spherical structures within the cellular cytoplasm) (Figure S15).

Confocal microscopy analyses at the basal layer of MDA-MB-231 overexpressing c-Srwt and -R175*L* showed a distribution of pY14-Caveolin 1/pY418-Src at the adhesion areas. The intracellular accumulation of pY418-Src was also detected at perinuclear areas. In cells expressing c-Src-W118A/R175L, the focal adhesion was not clearly displayed (Figure S15). The distribution of pY14-Caveolin 1/pY418-Src in MDA-MB-231-c-Src-R175L at the upper layer showed their co-localization at the perinuclear region, and in the cytoplasm where they decorated some vacuolar structures, as observed in SUM159. Similar to SUM159, in MDA-MB-231-c-Src-W118*A*/R175*L*, the vacuolar structures were also observed but to a much lesser extent (Figure S15).

**Figure 4.** Effect of c-Src variants expression in SUM159 and MDA-MB-231cellular migration and invasion. Migration of cells was analyzed by wound-healing assays, and tracking analyses (*n* = 3), as described in Materials and Methods in SUM159 and MDA-MB-231 cells (**A**). Additionally, tracking analyses of W118A/R175L to determine the cell migration of at least 100 individual cells determine the Accumulated distance, Euclidean, and the Velocity of migration to evaluate the ration of Euclidean/Accumulated distance that define random migration. (**B**) The capability of cells to migrate through a layer of Matrigel was employed to determine cell invasion (see Materials and Methods) in both SUM159 and MDA-MB-231 conditionally expressing c-Src variants. The control value is similar to that in Figure 2. Results of three independent experiments (*n* = 3) were expressed as the mean ± SD. The statistical significance is referred to cells expressing c-Src-wt, \* *p* < 0.5, \*\* *p* < 0.01, \*\*\* *p* < 0.001.

#### *2.6. SH2-c-Src Directed Aptamers Reduced Cell Proliferation*

The 14F and 17F aptamers directed to SH2-c-Src were designed and selected as described (Materials and Methods, and in the extended methods in Supplementary Information). The analyses of aptamers have been referred to the aptamer control (containing 38xAG, see Material and Methods), as in preliminary studies no differences were detected in cell proliferation of either SUM159 or MDA-MB-231 cells when compared to the transfection of the control (38xAG) and mock (empty transfection) [40]. We then determined the IC50 concentration for each aptamer in SUM159 and MDA-MB-231 cells by considering the total number of cells after the treatment with different concentrations of the aptamers (0, 25, 50, 200, and 500 nM), taking as 0% the concentration of aptamers with no effect and as 100% the effect at the maximal concentration of the aptamers for each cell line (see Materials and Methods). The results were graphically represented (Figure 5A). The IC50 values for 14F (117.1 and 141.7 nM) and 17F (94.3 and 97.0 nM) for SUM159 and MD-MB-231, respectively, were slightly different. Nevertheless, we decided to perform all the experiments at 100 nM for each aptamer at both cell lines. We determined the effect of SH2-c-Src directed aptamers (100 nM) in cell proliferation using the trypan blue exclusion method, allowing us to evaluate the total number of cells, the dead cells, and the living cells 72 h after transfection. The results showed (Figure 5B) that in both cell lines the number of dead cells was similar for all aptamers control (aptamer control), 14F and 17F, indicating that at this concentration they were not cytotoxic (Figure 5B). The 14F and 17F aptamers reduced the total cell number, as well as the number of viable cells in SUM159, while in MDA-MB-231 only the 14F caused a significant reduction of total and viable cells, as there is no significant reduction in viable cells in the MDA-MB-231 cell line with the 17F aptamer.

Then, we analyzed the levels of Myc, cyclin D1, and p27Kip1 in SUM159 and MDA-MB-231 cells treated with 14F and 17F aptamers. As compared to the aptamer control considered as 1, these aptamers reduced the expression of Myc and cyclin D1, while they increased those of p27Kip1 in both SUM159 and MDA-MB-231 cells (Figure S18).

**Figure 5.** Dose/response of SH2-c-Src directed aptamers in SUM159 and MDA-MB-231 cells and evaluation of cell proliferation. (**A**) Control aptamer (38xAG) and SH2-c-Src directed aptamers 14F and 17F were transfected at 25, 50, 200, and 500 nM to SUM159 and MDA-MB-231and the IC50 values were determined as described in Material and Methods (*n* = 3). (**B**) Control and SH2-c-Src directed aptamers 14F and 17F were transfected at 100 nM to either SUM159 or MDA-MB-231 and, 72 h later, the number of total, dead, and viable cells were determined by the trypan blue exclusion method (see Materials and Methods). Each experiment was measured in triplicates (*n* = 3) and repeated three times. Results were expressed as a percentage of the mean ± SD, \* *p* < 0.5, \*\* *p* < 0.01, \*\*\* *p* < 0.001.

#### *2.7. Role of 14F and 17F in Cell Migration and Invasion*

1

As cell migration and invasion are steps required for TNBC metastasis, we evaluated the relevance of the SH2 -c-Src domain in the regulation of these events by blocking its functionality in SUM159 and MDA-MB-231. We first observed that MDA-MB-231 cells migrated significantly less than SUM159, as the wound area after 13 h of migration was bigger. In either of the cell lines, the aptamers significantly reduced migration, as both 14F and 17F had a bigger wound area at the end of the experiments compared to the control (Figure 6A, Figures S16 and S17).

**Figure 6.** Role of SH2-c-Src directed aptamers in SUM159 and MDA-MB-231 cells to evaluate cellular migration and invasion. Wild type SUM159 and MDA-MB-231 cells were transfected with either control, 14F or 17F aptamers at 100 nM and cell migration (**A**) and invasion (**B**) were determined as described in Materials and Methods. Each experiment was measured in triplicates (*n* = 3) and repeated three times. Results were expressed as a percentage of the mean ± SD, \* *p* < 0.5, \*\* *p* < 0.01.

As for cell invasion, while only the 17F aptamer clearly reduced invasiveness in SUM159 cells, both 14F and 17F inhibited cell invasion of MDA-MB-231 cells (Figure 6B). We analyzed the effects of 14F and 17F aptamers on the expression of c-Src, pY418- Src/C-Src, Fak, pY397-Fak, Caveolin 1, pY14-Caveolin 1/Caveolin 1, Paxillin, and pY118- Paxillin/Paxillin in both SUM159 and MDA-MB-231 cells by WB considering the aptamer control as 1. In SUM159 and MDA-MB-231 cells, expression of c-Src and the activated form were unmodified, as also observed for Fak and auto-phosphorylated (Figure S18). In contrast, Caveolin 1 levels were reduced by aptamer 14F and increased by aptamer 17F, whereas pY14-Caveolin 1/Caveolin 1 was reduced by both aptamers in SUM159, but not in MDA-MB-231. Paxillin levels were unaltered by either of these aptamers, although the pY118-Paxillin/Paxillin ratio was significantly reduced by both 14F and 17F in both SUM159 and MDA-MB-231 cells (Figure S18). These results are different to those observed in Figure S11, as employing aptamers were blocking the functionality of the SH2 domain (Figure S18), while through the other experimental approach the induction of expression of c-Src variants was achieved.

Considering that migration and invasion are required for the metastatic process, these results support the experiments carried out with the conditional expression of c-Src variants and indicated that the SH2 domain is relevant for c-Src functionality.

#### **3. Discussion**

The SFKs control many signaling pathways involved in the regulation of several cellular processes. Thus, the deregulation of their functionality is associated with tumors, including breast cancer [1,7,44]. Here, we studied the relevance of the SH2 adapter domain of c-Src in two TNBC cell lines SUM159 and MDA-MB-231. Several scientific reports showed that, even though the two cell lines share several common characteristics, as both are considered Basal-Mesenchymal TNBC cell lines with a spindle phenotype, SUM159 (primary breast adenocarcinoma) has mutations in HRAS and PIK3CA (https://web. expasy.org/cellosaurus/CVCL\_5423), while MDA-MB-231 (pleural effusion) has deletions in p14ARF, p16, and CDKN2, and mutations in the KRAS, BRAF, and TERT promoter (https://web.expasy.org/cellosaurus/CVCL\_0062), supporting the heterogeneity observed in TNBC [12,45–48].

Our analyses of SFKs expression in SUM159 and MDA-MB-231 cells showed that c-Src, Fyn, and Lyn A were expressed at higher levels in SUM159 than in MDA-MB-231, while the opposite occurred for Lyn B and Yes.

To determine the relevance of the SH2-c-Src domain we followed two independent and complementary approaches, the conditional expression of c-Src variants with inactivating point mutations affecting SH2 functionality (R175L, W118A/R175L), and the transfection of two different aptamers directed to the SH2-c-Src domain, that interacted and then blocked the SH2-c-Src function in SUM159 and MDA-MB-231

It is well established that c-Src increased its expression/activity as the tumor progresses [1,5,7]. Our results also indicated that the c-Src played a relevant role in the renewal of the enriched population of BCSCs. The role of Src in the maintenance of these cells was previously observed in MCF7 [38], and here in SUM159 and MDA-MB-231 cells by the conditional expression of the dominant negative form of c-Src (SrcDN), as well as in MDA-MB-231 that conditionally have c-Src suppressed. Indirectly, expression of miR205 in SUM159, which inhibits SFKs members expression, suppressed SUM159 BCSCs renewal and stem cell markers [37]. Moreover, selective inhibitors of SFKs tyrosine-kinase activity also have an inhibitory effect in BCSCs renewal. In SUM159, resistant to paclitaxel, Dasatinib causes epithelial differentiation and enhances sensitization to paclitaxel, and a combination of both compounds reduces stem cell renewal and synergizes to diminish the viability of paclitaxel-resistant SUM159 cells [49]. In high-grade serous ovarian cancer cells, co-treatment of Saracatinib (SFKs inhibitor, AZD0530) and selumetinib (MEK inhibitor, AZD6244) reduced SFE and ALDH1 positive cells and, in vivo the loss of tumor-initiating cells following serial tumor xenografting [50]. Our results showed that the inactivating mutation of the SH2 and SH3 domains R175*L* and W118*A*/R175*L* significantly reduced BCSCs renewal in both cell lines, supporting the relevance of the adapter domains in the renewal of the tumor initiating cells. In primary PDAC cultures, established from patient-derived xenografts with Dasatinib or PP2 reduced the clonogenic, self-renewal, and tumor-initiating capacity of PaCSCs, which we attribute to the downregulation of key signaling factors such as p-FAK, p-ERK1-2, and p-AKT [51].

The anchorage-independent growth, which characterized tumor cells, showed the discrepancy between SUM159 and MDA-MB-231 cells. While in SUM159 cells only the suppression of both SH2 and SH3 functionality reduced it, in MDA-MB-231 cells, both c-Src-R175L and c-Src-W188A/R175L diminished colony formation in soft-agar as compared to c-Src-wt. Suppression of endogenous c-Src in MDA-MB-231 cells significantly reduced anchorage-independent growth [16]. Likewise, specifically silencing c-Src and not Yes or Fyn inhibited soft-agar colony formation in MDA-MB-231, MDA-MB-436, and SKBR3 [52]. Furthermore, SFKs catalytic activity inhibition or stable transfection of catalytically inactive c-Src into MDAMB-468 and MCF7 reduced the colony formation ability [53]. In addition, inhibition of SFKs expression by miR205 in SUM159 significantly diminished the anchorageindependent growth [37]. Inhibitors of SFKs catalytic activity such as Dasatinib inhibits soft-agar colony formation in BxPC3 and PANC1 pancreatic cancer cells [54]. Therefore, the

SH2-c-Src domain appeared relevant for MDA-MB-231 breast cancer cells in the anchorageindependent growth.

Cell proliferation was also influenced by SH2 functionality, as it was reduced upon induced expression of c-Src-R175L and c-Src-W118A/R175L variants as compared to c-Src-wt, as it was Myc expression in SUM159. On the contrary, we observed that in MDA-MB-231, inhibition of cell proliferation was not linked to alteration in the levels of Myc. Inhibition of SFKs tyrosine-kinase activity clearly blocks cell proliferation in MDA-MB-231 cells [15]. Interestingly, while Myc expression was reduced, cyclin D1 increased in response to the expression of c-Src-R175L and c-src-W118A/R175L in SUM159 cells. In this context, in rodent cells, it has been reported that Myc induces D1 in some cases, while in others, Myc does not induce or even repress D1, supporting the concept of the complex regulation of cyclin D1 gene (for review see [41]). Thus, the downregulation of Myc observed in SUM159 cells may contribute to the upregulation of cyclin D1.

Numerous studies show that the catalytic activity of SFKs is important for migration and invasion of tumor cells [6,15,16,22,55–57]. Induction of Fak autophosphorylation by the growth factor or by integrins facilitates its interaction with the SH2 domain of c-Src, opening c-Src conformation and, consequently, increasing its tyrosine kinase activity. Then, c-Src phosphorylates Fak at other sides, and facilitates the interaction/activation of other signaling molecules [1,7,20,21]. The complex Src/Fak phosphorylates/activates several focal adhesion proteins involved in migration and invasion [16,39,42,43]. In addition, our results showed that the SH2-c-Src domain was also relevant for modulation of invasion in both TNBC cell lines. Interestingly, they also revealed in SUM159 cells that altering the functionality of both SH2 and SH3 c-Src domains (c-Src-W118A/R175L variant) caused migration to occur in a random manner, as cells had high motility but they did not close the wound healing area. In HT1080 fibrosarcoma, overexpression of PEAK1 kinase, which is phosphorylated/activated at Y665 by SFKs, causes random migration and elevates cell invasion [58]. Consistently, in SUM159 cells, increased migration was associated with the activation of focal adhesion proteins caveolin1, paxillin, and Fak which were increased by overexpression of c-Src-W118A/R175L variant. In contrast, the c-Src-R175*L* reduced pY576-Fak/Fak. Confocal-microscopy analyses of the cellular distribution of activated Src (pY418-Src) and caveolin 1 (pY14-Caveolin 1) in SUM159 cells expressing the c-Src-R175L mutant at the basal layer showed their colocalization at the focal adhesion sites, as observed in SUM159-c-Src-wt. Interestingly, at the upper layer of analyses both c-Src -R175*L* and -W118A/R175L variants showed co-decoration of caveolae-like structures as compared to c-Src-wt. As mentioned above, SUM159 and MDA-MB-231 cells though they share a good number of common properties, also show some differences. Nevertheless, results showed that the SH2-c-Src domain played an important functional role in SUM159 and MDA-MB-231 TNBC cells.

To support these data, we approached this study by a different and complementary method. We designed two different aptamers directed to interact with the SH2-c-Src domain. The results showed that at a dose around the IC50 concentration both aptamers significantly inhibited the proliferation of SUM159 and MDA-MB-231, without inducing apoptosis, as the number of dead cells was unaltered. In agreement with these observations, the expression of Myc and cyclin D1 were reduced, while p27Kip1 levels were augmented. As observed for cells expressing c-Src-R175*L*, these aptamers inhibited migration and invasion in both TNBC cells. The aptamers design to bind to MNK1, which controls the eIF4E function by phosphorylation, significantly inhibits proliferation and migration of MDA-MB-231 [30]. The AS1411 aptamer directed to nucleolin induces blc-2 mRNA instability, reduces cell growth by causing cytotoxicity in MCF7 and MDA-MB-231 [59]. This aptamer has been tested in different tumors including glioma, renal cell carcinoma [31].

The results obtained indicated that the functionality of SH2-c-Src domain is important for the potential tumorigenicity of SUM159 and MDA-MB-231 cells as the inactivating point-mutation of this domain inhibited the biological functions required for the TNBC

cell. Similarly, aptamers directed to the SH2-c-Src domain also significantly reduced the performance of these TNBC cells.

#### **4. Materials and Methods**

#### *4.1. Antibodies and Reagents*

Table S1 contains the antibody information. The chemical reagents and enzymes used were of analytical grade and purchased from Thermo-Fisher (Waltham, MA, USA), Roche (Basel, Switzerland), Corning (Merck, Darmstadt, Germany**)**, PeproTech (London, UK), PAA Laboratories GmbH (Cölbe, Germany), Bio-Rad (Hercules, CA, USA), GE Healthcare and Sigma-Aldrich/Merck (Merck, Darmstadt, Germany).

#### *4.2. Cell Lines and Culture*

MDA-MB-231 (HTB-26) was from ATCC, and SUM159PT (CVCL-5423) [60] was provided by Dr. G. Dontu [61]. Cell lines were mycoplasma free and authenticated by the short-tandem-repeat analysis (GenePrintR 10 System from Promega (Madison, WI, USA), and GeneMapper v3.7 STR profile analysis software, Life Technologies, Carlsbad, CA, USA) (see Supplementary Information). Profiles were checked against public databases ATCC and DSMZ. MDA-MB-231 was maintained in DMEM, 5% FCS, 2 mM glutamine, 100 IU/mL penicillin, and 100 µg/mL streptomycin. SUM159 was cultured in Ham's F12, 5% FCS, 5 µg/mL insulin, 1 µg/mL hydrocortisone, 2 mM glutamine, 100 IU/mL penicillin, and 100 µg/mL streptomycin.

Generation of SUM159PT-Tet-On-c-Src- and MDA-MB-231-Tet-On-c-Src -wt, -R175L, - and -W118A/R175L, was carried out as described [16], and grown in the presence of 3 µg/mL blasticidin, 100 µg/mL zeocin to maintain the plasmid selection of cells expressing c-Src-wt and c-Src-R175L or with 3 µg/mL blasticidin, 3 µg/mL hygromycin for the selection of cell expressing c-Src-W118A/R175L.

The wild type (wt) and c-Src variants used in these experiments were from a chicken origin [1,4,19]. The BLASTp comparative analysis of c-Src protein sequences between *Homo Sapiens* (Protein Accession Number: P12931.3) and *Gallus-Gallus* (Protein Accession: P00523.4) resulted in over 94% identity at the amino acid sequences [33]. Since the EC10 mouse monoclonal antibody (Millipore, #05-185) specifically recognizes chicken c-Src, it was possible to determine c-Src variants in the presence of the endogenous human c-Src of SUM159 and MDA-MB-231 cells.

#### *4.3. Mammosphere Cultures*

Single cell suspensions of adherent cultures were plated in 6-well ultralow attachment plates (Falcon, Corning Life Science, Merck, Darmstadt, Germany) at 2 × 10<sup>3</sup> cells/well. Mammosphere cultures were maintained in serum-free DMEM/F12 media (1:1), B27 (1:50), EGF (20 ng/mL) and bFGF (20 ng/mL), insulin (5 µg/mL), hydrocortisone (5 µg/mL). After 10 days, cells were pipetted up and down to eliminate cellular aggregates and mammospheres (sphere-like structures with diameter ≥ 50 µm) were clearly detected by the optical phase contrast microscope (Nikon-Eclipse TS100, 4× magnification). Cultures were then trypsinized to induce mammosphere dissociation to single cells, which were seeded again for mammosphere formation. The experiment ended at the third generation of mammosphere formation. Sphere forming efficiency (SFE) was then calculated as the number of spheres formed per number of seeded cells and expressed as % means ± SD, as described [37,38].

#### *4.4. Anchorage-Independent Growth*

Cells were resuspended in a warmed solution of 0.3% agarose in a complete medium and seeded at 10<sup>5</sup> cells/60 mm dishes with a bottom layer of 0.5% agarose. Cells were re-fed every 72 h with a complete medium (300 µL/dish). At the 10-day growth, plates were stained with 0.5 mL of 0.005% crystal violet/water for 1 h and colonies with diameter ≥ 0.1 mm from 4–5 fields/plate were counted, as described [37].

#### *4.5. Cell Proliferation*

Cell proliferation was evaluated by counting viable cells performing a Trypan blue (Sigma-Aldrich) exclusion assay. Cells were seeded at 3 × 10<sup>5</sup> cells/60 mm dishes, 72 h later they were trypsinized, cells were pelleted and resuspended in a culture medium, mixed with a 0.4% Trypan blue/PBS solution (1:1), loaded on a hemocytometer, and Trypan blue-negative (viable cells) and Trypan blue-positive cells (dead cells) were counted.

#### *4.6. Cell Migration*

Cells were seeded in a complete medium in a 6-well plate and grown to confluence. The monolayer was scratched with a 200 µL micropipette tip, and washed with a fresh medium to remove floating cells. A complete medium was added to the cultures, and photomicrographs were taken every 30 min with a Microscope Cell Observer Z1 system (Carl Zeiss AG) equipped with a controlled environment chamber and Camera Cascade 1 k to monitor the wound closure. Migration was quantified using the wound-healing tool ImageJ, as described [15,37]. Tracking of cell migration was carried out in 100 cells/assay using the "manual-tracking" from the ImageJ program together with the "chemotaxis and migration tool".

#### *4.7. Invasion Assay*

Invasiveness was determined as described [15]. Briefly, cells were seeded in a serumfree medium on the upper chamber of cell culture inserts of 24-well plates (8 µm-pore PET membranes, BD) coated with MatrigelTM (5 × 104/well/200 µL). The lower chamber was filled with 600 µL of 20% FBS; 22 h later, after removing the cells on top of the inserts, those on the lower surface were fixed with methanol, nuclei stained with DAPI, and mounted on slides with a Prolong antifade-reagent. Filters were observed with a Plan 20×/0.50 objective of an axiophot fluorescence microscope (Zeiss, Oberkochen, Germany) equipped with an Olympus DP70 digital camera. DAPI-stained nuclei were counted.

#### *4.8. Western Blot Analysis*

Cell lysates preparation and Western Blot (WB) analyses were carried out, as previously described [37]. Briefly, cells were lysed at 4 ◦C with a lysis buffer (10 mM Tris–HCl (pH 7.6), 50 mM NaCl, 30 mM sodium pyrophosphate, 5 mM EDTA, 5 mM EGTA, 0.1% SDS, 1% Triton X-100, 50 mM NaF, 0.1 mM Na3VO4, 1 mM PMSF, 1 mM benzamidine, 1 mM iodoacetamide, and 1 mM phenantroline). Cell lysates were obtained by centrifugation at 21,380× *g* for 30 min at 4 ◦C; the protein concentration in the supernatant was determined by the BCA protein assay (Pierce, Rockford, IL, USA), and lysates were adjusted to equivalent concentrations with a lysis buffer. Aliquots of 30 µg of total cell lysates were then separated on SDS–PAGE. Proteins were transferred to PVDF membranes that were blocked 1 h at room temperature with 5% non-fat milk in TTBS (TBS with 0.05% Tween-20) or 5% BSA in TTBS for phosphoproteins. Incubation with primary specific antibodies was carried out overnight at 4 ◦C, and horseradish peroxidase-conjugated secondary antibodies in a blocking solution for 1 h at room temperature. Immunoreactive bands were visualized by the ECL kit.

#### *4.9. Aptamers Design and Selection*

The SH2 and SH3 domains of c-Src cloned into GST [62] and expressed in *E. coli* were purified from the soluble fraction by glutathione-resin affinity chromatography (Genescript, Piscataway, NJ, USA) as described [30]. Aptamers selection, cloning and sequencing, and secondary ssDNA structure prediction, as well as an enzyme-linked oligonucleotide assay (ELONA) methodology was previously described [30]. The aptamers employed were: 1. Control containing 38xAG, as described [30]; ApSH2.F14: GCGGATGAAGACTGGTGTAGACAATGGATACTCCCGCCACCTCCTCCCCCG CCCC-CCCGCCCTAAATACGAGCAAC; ApSH2.17F: GCGGATGAAGACTGGTGTGCGGTGGT

GGGTTGGGTGGGTGGGTTTGCGGGTTGCGTTGGCCCTAAATACGAGCAAC. Please see the extended method in the Supplementary Information for details.

#### *4.10. Aptamers Transfection and IC50*

SUM159 and MDA-MB-231 cells were seeded in 24 multi-well plates (10<sup>4</sup> cells/well/ 500 µL) in their corresponding culture media without antibiotics; 24 h later, cells were washed twice with their serum and antibiotic-free corresponding media. Then, the cells were incubated in 400 µL of culture media without antibiotics and 100 µL of the transfection mixture: 1. 0.25, 0.5, 2, or 5 µL of each aptamer, corresponding to 25, 50, 200, or 500 nM, in 49.75, 49.5, 48, or 45 µL of culture media; 2. 1.25 µL of DharmaFECT-4 (Thermo-Scientific) in 48.75 µL culture of media, following the manufacturer's manual; 8 h later, the cells were extensively washed with a culture media without antibiotics, then incubated for an additional 40 h and then analyzed as previously described in the cell proliferation assay section. The IC50 for each aptamer was determined considering the total number of SUM159 or MDA-MB-231 cells after the treatment with different concentrations of the aptamers. To this end, the untreated cells were considered as 0% (without effect) and 100% the effect at the maximal concentration of the aptamers for each cell line. Then, the obtained data were graphically represented employing the mathematical formula for the logarithmic trendline calculated with Excel to obtain the IC50 for each aptamer. Then, the IC50 values of each aptamer were used to determine their effects in cell proliferation, migration, and invasion of SUM159 and MDA-MB-231 cells, as previously described.

#### *4.11. Statistical Analyses*

Mean values, standard deviation, and statistical significance between data from the two different experimental conditions (±Doxy) were determined by the two-tail Student *t*-test. Data were normalized to the activity of the c-Src-wt variant for each cellular assay.

#### **5. Conclusions**

Our results conclude that the SH2-c-Src domain functionality is relevant for the potential tumorigenicity of SUM159 and MDA-MB-231, as the inducible expression of c-Src with the unfunctional SH2-c-Src domain inhibited the renewal of the enriched BCSCs cells, as well as other relevant functionalities in these TNBC cells. Similarly, the aptamers directed to the SH2-c-Src domain also significantly reduced the performance of these TNBC cells. Therefore, using a combination of SH2-c-Src functional inhibitors with those directed to the tyrosine kinase activity should be able to fully block the c-Src functionality and, consequently, could be therapeutically effective in the breast cancer treatment. Furthermore, as c-Src is also involved in other types of tumors (pancreas, colorectal, lung, etc.) [1,7,18,43,63,64], our results could eventually be extrapolated to these other pathologies.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2072 -6694/13/3/462/s1. 1. Supplementary Materials and Methods: 1. Purification and selection of aptamers for the SH2 domain of c-Src (extended method); 2. Immunofluorescence by lasers-canning confocal microscopy; 2. Supplementary Figures: Figure S1: Analyses of sphere formation efficiency (SFE); Figure S2: Images of mammospheres of SUM159 and MDA-MB-231 cells expressing c-Src variants; Figure S3: Effect of conditional expression of SrcDN in SUM159 and MDA-MB-231 cells or suppression of endogenous c-Src in MDA-MB-231 on SFE; Figure S4: Western blot analyses of ALDH1, NANOG, and Oct3-4 from mammospheres of SUM159 and MDA-MB-231 cells expressing c-Src variants; Figure S5: Soft-agar colonies from SUM159 and MDA-MB-231 cells expressing c-Src variants; Figure S6: Analyses of cell proliferation and Myc, cyclin D1, and p27Kip1 in Sum159 and MDA-MB-231 cells expressing c-Src variants, considering—Doxy conditions as the control; Figure S7: Cell cycle analyses of SUM159 and MDA-MB-231 cells expressing c-Src variants; Figure S8: Migration and invasion data of SUM159 and MDA-MB-231 cells expressing c-Src variants; Figure S9: Kinetics curves of wound-healing assays of SUM159 and MDA-MB-231 cells expressing c-Src variants; Figure S10: Representative images of wound-healing assays of SUM159 and MDA-MB-231 cells expressing c-Src variants; Figure S11: Expression and activation of Caveolin 1, Paxillin, and Fak

in SUM159 and MDA-MB-231 cells expressing c-Src variants. Referred to c-Src-wt as the control; Figure S12: Expression and activation of Caveolin 1, Paxillin, and Fak in SUM159 and MDA-MB-231 cells expressing c-Src variants. Referred to—Doxy as the control; Figure S13: Representative WB analyses of expression and activation of Caveolin 1, Paxillin, and Fak from the total cell extract from SUM159 and MDA-MB-231 cells expressing c-Src variants; Figure S14: Western Blot analyses of total c-Src, Fak, phosphorylated pY397-Fak, pY576-Fak, and Akt and phosphorylated pS473-Akt from SUM159 and MDA-MB-231 cells expressing c-Src variants; Figure S15: Confocal scanning microscopy analyses of pY14-Caveolin 1 and pY418-Src localization in SUM159 and MDA-MB-231 cells expressing c-Src variants; Figure S16: Kinetics curves of wound-healing assays of SUM159 and MDA-MB-231 cells treated with aptamers 14F and 17F; Figure S17: Representative images of wound-healing assays of SUM159 and MDA-MB-231 cells treated with aptamers 14F and 17F; Figure S18: Analyses by WB of expression of proliferation markers Myc, Cyclin D1, p27Kip1, and of migration Caveolin 1, Paxillin, and Fak in SUM159 and MDA-MB-231 cells treated with aptamers; Table S1: Detailed information for antibodies used in this work. Authentication of SUM159PT and MDA-MB-231 cell lines by short-tandem-repeat analyses. Uncropped gels of Figures 1–4, S3, S7, S8, S9, S12, S13.

**Author Contributions:** Conceptualization, J.M.-P., V.M.-V., A.C., and M.P.S.-B.; methodology, J.M.-P., V.M.-V., A.C., M.P.S.-B., M.E.M., and V.M.G.; software: J.M.-P., V.M.-V., A.C., M.P.S.-B., M.E.M., and V.M.G.; validation: J.M.-P., V.M.-V., M.P.S.-B.; A.C., M.E.M., and V.M.G.; investigation: J.M.-P., V.M.-V., A.C., M.P.S.-B., M.G.-H., V.F., M.E.M., and V.M.G.; resources: J.M.-P.; writing—original draft preparation, J.M.-P.; writing—review and editing: J.M.-P., V.M.-V., A.C., M.P.S.-B., M.E.M., and V.M.G.; funding acquisition, J.M.-P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work and the salary of Víctor Mayoral-Varo have been supported by grant number SAF2016–75991-R (MINECO, AEI/FEDER, UE) to Jorge Martín-Pérez.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available in Cancers-1076808 and in its Supplementary Materials.

**Acknowledgments:** We are thankful to M. Izquierdo, J. León, I. Palmero, and L. Molero for their comments and support, and to Javier Pérez for his comments and suggestions in art graphics. J.M.-P. is a member of the GEICAM (Grupo Español de Investigación en Cáncer de Mama) and IdiPaz. We acknowledge support for the publication fee by the CSIC Open Access Publication Support Initiative through its Unit for Information Resources for Research (URICI).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

