Background
Over 90% of drug development fails despite the use of state-of-the-art screening and optimization strategies [
1]. This percentage is even higher for anti-cancer drugs, which fail in advancing to clinical use because of problems with efficacy or toxicity in 97% of cases [
2]. Thus, more stringent and advanced screening methods are required to verify the therapeutic potential of new drugs.
Here, we exploit an innovative selection strategy, developed by our laboratory, to screen genetic libraries for molecules able to modulate cancer cell invasiveness directly in vivo (herein referred to as CanSel, for Cancer Selection). CanSel is based on the delivery of an arrayed library of cDNAs cloned in Adeno-Associated Vectors (AAV) in a relevant animal model, followed by the induction of a selective stimulus that eliminates most of the transduced cells. Should any of the transgenes exert a protective effect against the selecting stimulus, it will result enriched. This in vivo approach overcomes the major limitations of cellular screening, as shown in our previous work, where we exploited it to discover novel molecules protecting from cardiac ischemia [
3,
4], and enhancing the engraftment of mesenchymal stromal cells for tissue repair [
5]. Here, we adapted this strategy to screen for the most potent proteins inhibiting cancer cell invasiveness.
Traditional cancer treatments mainly inhibit the proliferation of cancer cells, using either radiotherapy or chemotherapy. Monoclonal antibodies, targeting cancer-specific receptors, and immunotherapy, including chimeric antigen receptor (CAR) T cells, have significantly prolonged the survival of patients with aggressive cancers. However, a limited number of patients respond to these treatments and most of them become resistant, often experiencing refractory disease within a year [
6‐
8]. Therapeutic strategies aiming at interfering with the invasiveness of cancer cells, which eventually leads to their dissemination and patient death, are still missing.
A growing body of evidence indicates that both cellular and extracellular elements modify tumor development, dissemination and response to therapy [
9]. Mechanical properties of the tumor extracellular matrix (ECM), such as stiffness and viscosity, also influence oncogene activation, cell morphology and migration, as well as the efficacy of existing therapies [
9,
10]. However, therapeutic interventions specifically targeting cancer ECM still fall short of expectations [
11].
Thus, we screened for secreted proteins able to shape the extracellular milieu to inhibit cancer cell invasiveness. Being secreted, the selected proteins could be delivered exogenously as biotherapeutics, avoiding the hurdles of gene therapy. In this way, we identified EMI domain-containing protein 2 (EMID2) as the best hit in dampening cancer cell invasiveness. We proved that EMID2 overexpression at the site of primary tumors reduces its growth and dissemination at distant sites, using mouse models of orthotopic and genetic tumors, relevant for human cancer. We also showed that high levels of EMID2 are protective in highly aggressive human cancers.
Discussion
Existing cancer therapies fail to hamper tumor cell invasiveness. Here we took advantage of CanSel, an innovative in vivo selection strategy [
3], to identify the most potent proteins inhibiting cancer cell invasiveness, tumor growth and dissemination.
The top-ranked protein identified by CanSel was EMID2, which has the potential to make a significant impact in cancer therapy, as (i) it reduces tumor growth and dissemination in in vivo models of lung and pancreatic cancer, (ii) it modifies the composition of cancer ECM, reducing cell invasiveness, and (iii) its high expression levels are associated with good prognosis in aggressive human cancers.
The power of CanSel in identifying and ranking factors with anti-invasive properties relies on the screening of thousands of molecules in a totally unbiased manner, in the absence of any prior knowledge on their function. Gene ontology analysis on both enriched and lost transgenes indicated that out of > 1000 transgenes included in the AAV library, the top 60 encoded mainly for proteins associated with ECM (fibulins, galectins and collagens), angiogenesis (members of the Vascular Endothelial Growth Factor family) and immune response (interleukins, chemokines). These processes all pertain to extracellular milieu, which is consistently altered in aggressive cancers and, therefore, it represents an attractive target for therapy [
35].
EMID2 stood out as the most potent factor in both inhibiting cell invasiveness in vitro and reducing tumor growth and dissemination in vivo. After initial comparison of the 10 top hits, we selected the best performing 4 proteins for in vivo validation. More specifically, we assessed their capacity to modify the shape of the invasive front. While in control tumors, the leading edge contained collagen and fibronectin bundles, perpendicularly aligned with the tumor border, which steers cancer cell escape from the primary tumor [
36‐
38], EMID2-overexpressing tumors had a round border, with few infiltrating branches.
EMID2 belongs to the EDEN superfamily, which includes members sharing an EMI domain [
22,
39]. Two additional members of this family, Emilin1 and Emilin2, resulted enriched in their respective AAV9-pool, and scored high in the final ranking (33rd and 77th position, respectively), confirming a relevant role of these ECM proteins in inhibiting tumor invasiveness [
40,
41]. While the function of Emilins has been extensively investigated in other fields, scant information is available for EMID2, which confirms the power of CanSel in selecting novel molecules. Two studies have associated
EMID2 to pathological and developmental processes characterized by abnormal ECM, such as asthma, nasal polyps and collagen fibrillogenesis in the cornea [
42,
43].
Our data show that EMID2 overexpression normalizes ECM composition, reducing the density of collagen I and fibronectin, which are known to support tumor growth and invasiveness [
44,
45]. In advanced tumors, TGFβ stimulates the activation of CAFs, leading to ECM remodeling [
46], which in turn primes tumor cells for distal dissemination [
47]. Consistent with previous data on Emilin1 [
23], we observed a significant reduction of active TGFβ in EMID2-treated cells and tumors, with reduced CAF activation.
ECM stiffness is generally higher in tumor than in normal tissues, which favors cancer cell dissemination and reduces drug and immune cell penetration in multiple cancer types, including breast, liver, lung and pancreatic cancer [
48,
49]. EMID2 reduces ECM stiffness both in vitro and in vivo, resulting in reduced nuclear YAP.
Most importantly, overexpression of EMID2 at the site of primary tumors in the pancreas, significantly reduced the metastatic burden in the lung, paving the way to its therapeutic use in disseminated human cancers, in combination with standard therapies. Consistently, high expression of EMID2 is a favorable prognostic factor for most aggressive human cancers. The anti-cancer activity of EMID2 is consistent with high level of methylation of its gene in both LUAD and PAAD cells compared to the corresponding healthy tissues, as detected using the (
http://www.bioinfo-zs.com/smartapp/) and the MethMarkerDB! (
https://methmarkerdb.hzau.edu.cn) databases.
Our initial idea of using a library of secreted proteins has obvious therapeutic implications. Being a secreted protein, EMID2 could be administered as a recombinant protein, avoiding the hurdles related to the use of viral vectors and exploiting the production platforms developed for other collagenous proteins, already used for human therapy [
50‐
52].
To what extent the therapeutic activity of EMID2 could act synergistically with other treatments (i.e. immunotherapy, integrin inhibitors, anti-angiogenic therapy) remains an outstanding issue, which deserves further investigation.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.