Current preclinical anticancer drug testing carries a low predictive value, since only 5% of compounds showing efficacy in in vitro tests are licensed following clinical trials [1
]. Therefore, improved predictive in vitro methods are urgently needed. We developed a human tumor leiomyoma–derived matrix “Myogel” to better mimic the human tumor microenvironment (TME) [2
]. Myogel provides a three-dimensional (3D) culture environment for in vitro cancer studies, preserving the soluble factors present in the human TME by including, for example, soluble cytokines and growth factors [3
]. Myogel proteome differs greatly from mouse sarcoma–derived Matrigel [3
]. We previously demonstrated that 66% of the Myogel protein content differs from Matrigel [3
]. Yet, these matrices share several proteins such as laminin, type IV collagen, heparan sulfate proteoglycans, nidogen, and epidermal growth factor [3
]. Carcinoma cells show a markedly higher migration and invasion ability on Myogel compared to Matrigel [2
Head and neck squamous cell carcinoma (HNSCC) represents the sixth most common cancer worldwide, with a relatively low survival rate of around 50% [5
]. Treatment approaches include surgery combined with radio-, chemo-, or epidermal growth factor receptor- (EGFR-) targeted therapy relying on cetuximab (Erbitux). EGFR is frequently overexpressed in HNSCC [6
] and its inhibitors have shown clear anticancer effects in both in vitro and in vivo preclinical studies [7
]. However, most EGFR inhibitors have yielded disappointing results in HNSCC clinical trials and there are no clinically approved predictive biomarkers for EGFR inhibitor response [12
]. We propose that the presence of the human tumor matrix, as a physiologically relevant extracellular matrix for human cancer cells, could improve the predictability of anticancer drug testing in preclinical studies of HNSCC.
In this study, using a high-throughput drug screening method, we tested 12 HNSCC cell lines cultured on plastic, mouse sarcoma–derived Matrigel, or Myogel using 19 anticancer compounds targeting EGFR, MEK, and PI3K/mTOR. We compared the efficacy of the anticancer drugs tested in vitro against the patients´ response rates in 19 monotherapy clinical trials. We identified different drug effects between Myogel and the other testing conditions, and demonstrated that Myogel in in vitro drug screening replicates the best results from clinical trials.
Traditionally, new anticancer compounds are tested on cancer cell lines cultured on top of plastic wells, followed by testing using xenograft mouse models prior to clinical trials [1
]. However, it is now clear that this strategy remains unsuccessful given the very low predictive value of the drugs’ efficacy in clinical trials [1
]. We, thus, hypothesized that this failure is caused by missing the human physiologically relevant TME. Thus, we developed the first human-derived TME-mimicking matrix, “Myogel”, in which 66% of its protein content differs from the mouse tumor–derived Matrigel [2
]. Thus far, we have demonstrated that a human tumor matrix induces the invasion of carcinoma cells more than mouse sarcoma–derived matrix “Matrigel” [2
]. In this study, we compared traditional 2D plastic drug screening method with 2D and 3D Matrigel and Myogel. Combining data from monotherapy clinical trial results for HNSCC to our in vitro drug testing results, we demonstrated that Myogel improved the reliability of in vitro drug testing.
In solid tumors, cancer cells are influenced by TME containing hundreds of effectors, such as stromal and immune cells, their cytokines, growth factors, and structural matrix proteins [2
]. Clearly, all of these factors interfere with the chemosensitivity of cancer against anticancer drugs [20
]. Due to the unknown but presumably very large number of such proteins, it is quite difficult to precisely analyze the factors interfering in drug screening. Unfortunately, this also applies to our drug testing using Myogel representing the human TME.
Currently, cancer research is consistently moving towards 3D cell culture models for in vitro studies [20
]. Thus, several matrices have been applied, such as Matrigel, collagen, and chemically defined synthetic hydrogels [21
]. All of these matrices miss a broad spectrum of both human- and/or tumor-derived structural or signaling molecules. We, thus, hypothesized that a human tumor–derived matrix would provide a more appropriate TME for the cells and, therefore, a more in vivo–relevant drug response.
Cell number measurement using imaging technique is not the optimal way to compare cell density in different matrices due to alternated cell morphology (spheroid in Matrigel and stellate-shaped in Myogel) and 3D growth (Figures S3 and S4
). In this study, we used the CellTiter-Glo luminescent assay to determine the number of viable cells in culture plates. Even though some Matrigel wells showed low cell density under microscope, the luminescent assay readouts indicated similar or even larger number of viable cells compared to 2D and Myogel wells (Figure S4
). This was because the cells on Matrigel are tightly packed together forming spheroid structures. We also did not find correlation between alternated cell growth in different matrices and the drug response.
The EGFR inhibitors were considered promising anticancer drugs for HNSCC. However, only modest response rates have been found [22
]. Among all EGFR inhibitors, only cetuximab has been approved by the Food and Drug Administration (FDA) for HNSCC. Unfortunately, its impact has remained quite limited, and even responsive patients rapidly developed drug resistance [6
]. Other EGFR inhibitors, such as gefitinib, have failed in phase III clinical trials despite their promising results in preclinical in vitro and in vivo animal studies [6
]. Interestingly, in our setting, both plastic and Matrigel yielded higher response rates than the clinical outcomes of the EGFR inhibitors. This might explain why so many EGFR-targeted drugs successfully passed preclinical testing and subsequently failed during clinical trials [1
]. Yet, EGFR inhibitors tested using Myogel wells exhibited similar response rates to the 16 monotherapy clinical trial results (with a total of 1258 patients).
In order to understand the mechanism behind the variable cell responses to EGFR and MEK inhibitors in different culturing conditions, we studied the protein expression of EGFR, ERK1/2, and pERK1/2 in five cell lines which revealed the most differential EGFR- and MEK-inhibitor effects between Myogel and the other testing conditions. After repeating the experiments three times, we could not see any clear difference in the protein levels of EGFR, ERK1/2, and pERK1/2 in these culture conditions. Therefore, there was no correlation between the protein levels and responses to EGFR and MEK inhibitors (Figure S5
). Our in vitro results are in line with the in vivo results, which showed that EGFR protein expression and its gene copy number have failed as a predictive biomarker in numerous studies [12
]. Additionally, previous in vitro study on HNSCC cell lines showed that EGFR amplification or overexpression was only weakly associated with EGFR inhibitor response [23
]. While many studies report EGFR overexpression in HNSCC, supporting data are inconsistent and limited due to the large variation in antibodies, the lack of controls, and testing only at the RNA level [24
]. One study revealed that EGFR protein overexpression more commonly occurs in established HNSCC cell lines (n
= 14) than in clinical samples (n
= 55) [25
]. Clinical HNSCC samples (n
= 55) did not overexpress EGFR at the protein level compared to healthy mucosa (n
= 46) [25
Several genomic alterations in HNSCC affect the PI3K/AKT/mTOR pathway activation [26
], which plays an important role in cancer initiation and progression. mTOR inhibitors have shown promising anti-tumor activity in preclinical studies and early stage clinical trials in HNSCC [27
]. Based on two phase II clinical trials, temsirolimus showed promising tumor shrinkage, but this was associated with no objective response [15
]. Our in vitro results, relying on a DSS value of 5 as the cut-off point, did not predict patient outcome in clinical trials across all testing conditions. However, the majority of the tested cell lines yielded a low DSS value, close to the cut-off point of 5, which raises questions about the reliability of that score as a marker for an objective response. In one study, the authors only highlighted DSS values of less than 10 as non-responders [28
]. If the cut-off point is increased to DSS > 10, the results more closely mirror patient responses. The selection of the most reliable response cut-off point is crucial and small changes in it could greatly induce the drug response rates, particularly when the DSS values are close to the cut-off point. Additionally, here we used only monotherapy clinical trials; those patients typically resistant to traditional treatment. This renders the comparison to the in vitro results relatively less than ideal. However, we excluded combination therapy trials, since separating the drug effect from other treatments (radiation or chemotherapy) would be impossible.
Another mTOR inhibitor, sirolimus, has thus far been studied in only one monotherapy HNSCC clinical trial among 16 patients. It showed an objective response rate of 25% and one complete patient response [19
]. Although our in vitro study revealed a much higher response rate for sirolimus, further clinical trials are needed to interpret the in vitro results.
Clearly, those drugs which target receptor activities, such as EGFR, are more greatly affected by the nature of the extracellular environment than those that target cytosolic enzymes, such as mTOR. This could explain Myogel’s ability to reveal the real response rate for EGFR antibodies better than for mTOR inhibitors.
We predicted that a 3D culture would provide more reliable drug testing results than 2D monolayers. However, in contrast, 2D Myogel- and Matrigel-coated wells yielded rather similar results to 3D cultures for most of the drugs tested. Thus, our data suggest that a 2D-coated culture is suitable for drug testing purposes as long as the culture contains critical elements of the human TME.
In conclusion, since the human tumor matrix improved the predictability of the in vitro anticancer drug testing of HNSCC cell lines, we argue that using it would reduce the number of false-positive preclinical results, the cost of drug development, and the unnecessary suffering of cancer patients.