Integrins represent a family of transmembrane adhesion receptors, facilitating the adhesive connection between cells and their surrounding extracellular matrix (ECM) or neighboring cells (Takada et al., 2007; Barczyk et al., 2009; Kadry and Calderwood, 2020). They comprise a group of heterodimeric proteins generated by non-covalent association of an a- and a ß-subunit (Ginsberg, 2014). Both subunits are classified as type 1 transmembrane proteins, composed of a rather large extracellular domain and a relatively small transmembrane- and intracellular region (Calderwood, 2004; Ginsberg, 2014). The globular head domain creates a binding site for extracellular ligands while the short cytoplasmic tails interact with a cluster of associated proteins that ultimately connects to the cytoskeleton. In total there are 18 α-and eight ß-subunits, generating 24 different heterodimers, known to be expressed in humans (Calderwood, 2004). This variety in combinations allows integrins to interact with—and respond to a broad range of ligands, including insoluble ECM proteins, matricellular proteins, cell surface proteins, and soluble proteins (Alday-Parejo et al., 2019). Several recognition motifs for integrin-binding have been identified. The Arg-Gly-Asp (RGD) motif is recognized by eight different integrins and has been found in a plethora of molecules ranging from ECM proteins to growth factors to coats of microorganisms.

Integrin transmembrane receptors execute two core functions: they mediate adhesion of cells to the ECM or neighboring cells, and they engage in transduction of signals received from the microenvironment. Integrin-mediated cell adhesion is dynamic: flexibility in integrin conformation allows a balance between active (open; high affinity) and inactive (closed; low affinity) states. The active state is regulated by interaction of the intracellular adaptor proteins talin and kindlin with the ß-subunit cytoplasmic tail and is further stabilized by interaction with ligand at the extracellular integrin head domain (Calderwood, 2004; Sun et al., 2019). Moreover, firm cell adhesion requires integrins to cluster in cell adhesion complexes that connect to the cytoskeleton.

Integrin-mediated cell adhesion controls many aspects of cell behavior including survival, proliferation, metabolism, differentiation, as well as cell shape and motility (Huveneers and Danen, 2009). Several mechanisms of such outside-in signaling have been proposed. First, integrins allow cells to interact with the ECM in which soluble growth factors such as VEGF, TGFβ and many others are concentrated, modified, and presented to cells (Hynes, 2009). Second, integrins can directly bind and activate growth factors such that they can stimulate their cognate receptors, a process currently established for activation of TGFβ by αvβ6 and αvβ8 (Margadant and Sonnenberg, 2010). Third, integrin engagement and clustering can lead to local activation of receptors for soluble ligands such as EGF, PDGF, and others, often involving receptor crosstalk via Src family kinases (Ivaska and Heino, 2011; Brizzi et al., 2012). Fourth, the dynamic intracellular complex of adaptor and signaling proteins that couples integrins to the cytoskeleton allows 1) local signaling through GTPases and kinases and 2) sensing of- and responding to mechanical aspects of the microenvironment by mechanoresponsive interactions (Huveneers and Danen, 2009; Kechagia et al., 2019).

Dysregulation of integrin expression on cancer cells has been extensively studied in cell culture and animal models and shown to provide therapeutic opportunities for arresting tumor growth, reducing resistance to chemo-or radiotherapy, or attenuating invasion and metastasis. Studies using genetically engineered mouse models or using human tumor cells transplanted in immune deficient mice have extensively shown that deletion of integrins in cancer cells or preventing integrin function with blocking antibodies or peptides could interfere with tumor growth, metastasis, and resistance to chemo- or radiotherapy (Juliano and Varner, 1993; Danen, 2005; Desgrosellier and Cheresh, 2010; Hamidi and Ivaska, 2018; Cooper and Giancotti, 2019). For the large family of β1 integrins, dual roles have been identified in growth versus metastasis, indicating that caution is warranted for their application as therapeutic targets (Ramirez et al., 2011; Moran-Jones et al., 2012; Truong et al., 2014; Moritz et al., 2021). Integrins such as αvβ3, αvβ5, and α5β1, are not only expressed on tumor cells but are also induced on endothelial cells during the process of angiogenesis (Friedlander et al., 1995; Avraamides et al., 2008). These integrins have indeed been shown to serve as targets for anti-angiogenic therapies in cancer, although the mode of action of anti-angiogenic drugs targeting integrins remains enigmatic (Friedlander et al., 1995; Hynes, 2002; Alavi and Cheresh, 2008; Avraamides et al., 2008).

Recent studies have added a range of novel emerging cancer-related processes that require the participation of integrins, including the establishment of a pre-metastatic niche, epithelial-to-mesenchymal transition (EMT), metabolic rewiring, cancer cell stemness and dormancy (Barkan et al., 2010; Goel et al., 2014; Seguin et al., 2015; Ata and Antonescu, 2017; Ji et al., 2020; Park and Nam, 2020; Winkler et al., 2020; Coban et al., 2021). The involvement of integrin αvβ6 in activation of TGFβ was recently connected to SOX4 mediated cancer immune evasion: αvβ6 blocking antibodies could inhibit SOX4 expression and sensitize mouse models for triple negative breast cancer to T cell mediated killing in response to immune checkpoint inhibitors (Bagati et al., 2021). Integrin αvβ8, which can also activate TGFβ, represents a target expressed on immune cells for modulating anti-tumor immunity. I.e., αvβ8 blocking antibodies or specific depletion of integrin αvβ8 from the surface of CD4⁺CD25⁺ regulatory T cells could attenuate TGFβ mediated inhibition of CD8⁺ T cells and thereby restore tumor killing capacity of CD8⁺ T cells and synergizing with radio- or immune therapy (Dodagatta-Marri et al., 2021).

The expression of integrins on the cell-surface and their apparent role in several cancer related processes makes them appealing targets for the development of cancer therapies. However, despite the abundance of promising preclinical data, integrin targeting therapies in clinical studies have thus far largely failed to deliver. Notably, although not within the scope of this review, components of the integrin signaling complexes represent additional targets in cancer. For example, focal adhesion kinase (FAK) is overexpressed or activated in multiple cancers and supports tumor cell proliferation, migration, and therapy resistance. Small molecule inhibitors targeting FAK, such as defactinib, GSK2256098, VS-6063, and BI 853520, are currently being investigated in several clinical trials, mostly in combination with other agents (Mohanty et al., 2020; Dawson et al., 2021). Src is another interesting target associated with integrin signaling. Dasatinib, a Src inhibitor, showed efficacy when combined with docetaxel in castration-resistant prostate cancer patients (Araujo et al., 2012) (NCT00439270), and was more effective than imatinib in Pediatric Philadelphia Chromosome–Positive Acute Lymphoblastic Leukemia (Shen et al., 2020). On the other hand, dasatinib monotherapy failed to meet expectations in patients with recurrent glioblastoma (Lassman et al., 2015) or in patients with locally advanced or stage IV mucosal, acral, or vulvovaginal melanoma (Kalinsky et al., 2017). The challenges of targeting Src family proteins were recently reviewed by Martellucci and others (Martellucci et al., 2020). Integrins interact with many other cytoplasmic proteins, which are being investigated for their potential as therapeutic targets, however these have not yet been translated to the clinic (Cabodi et al., 2010; Bachmann et al., 2019).

In this review we focus on integrins as drug targets in cancer and discuss the apparent disagreement between preclinical and clinical results, we provide an overview of new encouraging preclinical findings and consider new opportunities to exploit the potential of integrin adhesion complexes in the effective treatment of cancer.

Thus far, antibodies have been the major type of anti-integrin therapeutics tested in clinical trials (Table 1). The exquisitely high specificity and corresponding low toxicity of these antibodies are most likely responsible for this high prevalence. A major limitation is a lack of knowledge with respect to expression of the target integrin in the tumor of the patient. Prior treatments may have affected integrin expression patterns in the tumor tissue. In addition, data on antibody pharmacology is generally lacking in the clinical studies. It is well known that targeting of therapeutics to the tumor tissue can be difficult due to poor vascularization, (Carmeliet and Jain, 2011), and this may be a significant problem for the relatively large antibody drugs. Hence, it is important to determine expression of the target integrin and establish actual reach of the integrin-targeting antibodies to the tumor tissue to relate these aspects to response rates in individual patients.

Many integrins show overlap in their ligand binding spectrum. I.e., key ECM proteins present in cancer tissues such as fibronectin, laminins and collagens can be recognized by more than one integrin (Danen, 2005). Hence, the effect of blocking one integrin may be compensated for by another integrin binding the same ligand. Patients entering experimental trials often present with a mix of primary and metastatic lesions at different stages. Integrin expression has been observed to differ between primary and metastatic lesions indicating that therapies may affect one but not the other stage. e.g., expression of integrin α2β1 was shown to promote tumor growth of a breast cancer cell line whereas α2β1 expression was attenuated once the breast cancer cells colonized the bone (Moritz et al., 2021). In fact, integrins have been shown in some cases to have opposing roles at different stages and repress rather than support disease progression and metastasis. While depletion of β1 integrins led to reduced outgrowth of primary tumors, it enhanced metastatic capacity in an orthotopic model using triple negative breast cancer cells (Truong et al., 2014). Deletion of β1 integrins also increased prostate cancer progression in a genetic mouse model (Moran-Jones et al., 2012). Likewise, specific deletion of one of the β1 integrins, α2β1, was demonstrated to inhibit tumor metastasis in mouse models for breast or prostate cancer (Ramirez et al., 2011; Moritz et al., 2021). Although similar examples are not described for the αv integrins targeted in clinical trials thus far, these findings suggest that therapeutic targeting of integrins may lead to complex responses in patients that may vary for individual patients.

Another mechanism that may underlie failure of anti-integrin drugs involves TEVs that have been implicated in tumor angiogenesis, immune evasion, and metastasis (Becker et al., 2016). Tumors produce more EVs with a different cargo composition (proteins, lipids and nucleic acids) as compared to normal tissues and these EVs can be derived from the tumor cells as well as other cell types in the tumor microenvironment. Integrins are expressed on TEVs, thus guiding their preference for organ colonization (Hoshino et al., 2015). As integrin expressing TEVs are released by various cancer types they may represent a common obstacle by sequestering integrin-targeting antibodies or peptides before these can reach their tumor target (Fedele et al., 2015; Hoshino et al., 2015; Singh et al., 2016; Carney et al., 2017; Krishn et al., 2019; Li et al., 2020). This concept has also been demonstrated for patients with inflammatory bowel disease where EVs expressing integrin α4β7 prevented vedolizumab from reaching α4β7 expressed on T cells, which may affect therapeutic efficacy (Domenis et al., 2020).

Integrin αvβ3 expression has been extensively associated with melanoma progression from an early radial growth phase to an invasive vertical growth and metastasis (Danen, 2005; Desgrosellier and Cheresh, 2010). Recently, differential expression of the integrins αvβ3 and αvβ6 has been observed in two subtypes of prostate cancer. Using patient derived tumor tissue and tumor bearing murine models, αvβ3 was found to be largely absent in prostate adenocarcinoma ADPrCa but significantly upregulated in the more malignant primary neuroendocrine prostatic cancer (NEPrCa) and its metastatic lesions in the lung (Quaglia et al., 2021). Combined with previous findings on the role of αvβ3 in the differentiation of ADPrCa to the aggressive NEPrCa, αVβ3 could have potential as a biomarker in the early detection of this malignant transition in prostate cancer (Quaglia et al., 2020; Quaglia et al., 2021). The expression of integrin αvβ5 has been suggested to represent a predictive biomarker for several cancer types amongst which, breast, hepatic, and gastric carcinomas (Bianchi-Smiraglia et al., 2013; Lin et al., 2018). Recently, elevated levels of αvβ5 have been detected in patients suffering from either glioblastoma or colorectal carcinoma (Zhang et al., 2019; Shi et al., 2021). For both types of cancer, the overexpression of αvβ5 was correlated with an unfavorable overall survival (Zhang et al., 2019; Shi et al., 2021). Integrin αvβ6 has been shown to represent an unfavorable prognostic marker in pancreatic cancer patients (Li et al., 2016). This integrin was recently found to be a promising serum biomarker for patients with pancreatic cancer. Based on the identification of αvβ6 in serum, chronic pancreatic (cP) patients could be distinguished from patients with pancreatic adenocarcinoma (PAC) and high serum levels of αvβ6 were associated with poor survival (Lenggenhager et al., 2021). Up to now, Carbohydrate antigen CA19-9 has been the only biomarker in use for PAC, yet its sensitivity and specificity failed to meet the expectations for use as conclusive diagnostic tool (Goonetilleke and Siriwardena, 2007). A study with a larger patient cohort will be needed to further assess the potential of αvβ6 alone or in combination with CA19-9 as a prognostic serum biomarker for PAC.

Metastasis in the bones is often lethal in patients with mammary tumors (Coleman, 2006; Wang et al., 2019). Therefore, finding a predictive biomarker is essential for the early recognition of potentially metastasizing tumors. Integrin α5β1 is known for its participation in tumor promoting processes like angiogenesis, proliferation and metastasis (Hamidi and Ivaska, 2018; Hou et al., 2020). In early-stage breast cancer patients, α5β1 expression in the primary tumor was recently associated with the presence of disseminated tumor cells in bone marrow aspirates and poor metastasis-free survival (Pantano et al., 2021). The same study showed that α5 gene silencing or pharmacological inhibition of α5β1 with volociximab attenuated bone colonization following intravenous injection of tumor cells in mice. Hence, stratification of breast cancer patients based on α5β1 expression may represent a way to exploit the potential of α5β1 antibodies, which have thus far not shown clinical benefit. Integrin α5β1 was also found to be upregulated in several gastrointestinal tumors where enhanced expression of ITGA5 corresponded with a poor prognosis (Zhu et al., 2021). Again, these findings warrant larger scale patient studies to explore the potential of α5β1 as a prognostic biomarker in solid tumors.

Liposomal vesicles have been used extensively to reduce the toxicity of conventional anti-cancer therapeutics in healthy tissues (Allen and Cullis, 2013). Low treatment efficacy with this approach is caused by ineffective reach of the tumor. The introduction of RGD peptides on the surface of liposomal like vesicles has generally enhanced both drug accumulation in the tumor and anti-tumor efficacy of the drug in mouse models (Fu et al., 2021; Gao et al., 2021; Gong et al., 2021; Khabazian et al., 2021). Additional adjustments were made to the vesicles to further improve their drug transporting characteristics (Figure 2). Sustained drug release of the liposomes was enhanced, making use of PEGylated positively charged lipids (Khabazian et al., 2021). The cationic liposomes decorated with the cRGD peptide were then able to deliver negatively charged siRNA into melanoma cells and effectively induce cell death (Khabazian et al., 2021). Alternatively, Gao et al. developed a double membrane vesicle (DMV), presenting not only the RGD peptide, but also lipopolysaccharides (LPS) (Gao et al., 2021). The association of LPS (normally exposed in the outer membrane of Gram-negative bacteria) with immune cells facilitated the transit of the vesicles from the vasculature into the tumor microenvironment where it could target melanoma cells and deliver therapeutics. Other αvβ3 targeting liposomal like formulations have shown a promising reduction in tumor growth for lung and hepatocellular carcinoma in in vivo models (Fu et al., 2021; Gong et al., 2021). Liposomes targeting other integrins are slowly emerging, although selective expression of these integrins in tumor tissue is less evident. Modification of the liposomal membrane with the α5β1 binding peptide PR_b, elevated the tumor specificity of the vesicle for pancreatic cancer cells (Shabana et al., 2021). The addition of a thermosensitive and biodegradable hydrogel in the formulation enabled sustained release of the combination treatment paclitaxel and gemcitabine and attenuated pancreatic tumor growth. Other liposomes presenting the integrin α2β1 binding ligand DGEA, were used to target breast cancer and effectively reduced tumor growth in vivo and enhanced the overall survival of the mice (Zhou et al., 2021).

The use of integrins to direct anti-cancer therapeutics has not been restricted to their application in liposomal drug transport. Integrins may represent targets for the development of novel tumor selective immunotherapies (Figure 3A, B). In mouse models for breast cancer and head and neck squamous cell carcinoma, it was shown that α6β4 is preferentially expressed on CSCs and represents a target for immunotherapies. Vaccination with dendritic cells pulsed with β4 peptide or adoptive transfer of T cells incubated with β4-CD3 bispecific antibodies, could induce T cell anti-tumor activity and inhibition of tumor growth and metastasis formation in the lungs of tumor bearing mouse models (Dobson et al., 2021). The application of covalent linking between an integrin binding peptide (mostly RGD) and an established anti-cancer therapeutic has also been explored (Figure 3D). This approach has led to reduced therapeutic-associated toxicity in healthy tissues (Cirillo and Giacomini, 2021). It will be interesting to compare toxicity profiles for this approach with those of liposomal encapsulations. Lastly, RGD peptides have also been incorporated in polydopamine (PDA) coatings to target photosensitizing agents such as gold nanostars leading to tumor specific cell death and limited adverse effects after near infrared activation of the drug (Li Y. et al., 2021) (Figure 3C).