EMT, Stemness, and Drug Resistance in Biological Context: A 3D Tumor Tissue/In Silico Platform for Analysis of Combinatorial Treatment in NSCLC with Aggressive KRAS-Biomarker Signatures

Abstract

Epithelial-to-mesenchymal transition (EMT) is discussed to be centrally involved in invasion, stemness, and drug resistance. Experimental models to evaluate this process in its biological complexity are limited. To shed light on EMT impact and test drug response more reliably, we use a lung tumor test system based on a decellularized intestinal matrix showing more in vivo-like proliferation levels and enhanced expression of clinical markers and carcinogenesis-related genes. In our models, we found evidence for a correlation of EMT with drug resistance in primary and secondary resistant cells harboring KRAS^G12C or EGFR mutations, which was simulated in silico based on an optimized signaling network topology. Notably, drug resistance did not correlate with EMT status in KRAS-mutated patient-derived xenograft (PDX) cell lines, and drug efficacy was not affected by EMT induction via TGF-β. To investigate further determinants of drug response, we tested several drugs in combination with a KRAS^G12C inhibitor in KRAS^G12C mutant HCC44 models, which, besides EMT, display mutations in P53, LKB1, KEAP1, and high c-MYC expression. We identified an aurora-kinase A (AURKA) inhibitor as the most promising candidate. In our network, AURKA is a centrally linked hub to EMT, proliferation, apoptosis, LKB1, and c-MYC. This exemplifies our systemic analysis approach for clinical translation of biomarker signatures.

Keywords: 3D lung tumor tissue models; EMT; KRAS biomarker signatures; boolean in silico models; drug resistance; invasion; stemness; targeted combination therapy.

Figure 1

A preclinical tissue tumor model reflecting EMT, invasion, clinical and biological markers. (A) 3D tumor model based on porcine jejunum. (B) Dynamic cell culture in a flow bioreactor. (C) HCC827 lung cancer cells on SISmuc (Pan-cytokeratin, green) and (D) HCC827 cells on SISmuc + TGF-β1 in a flow bioreactor, arrows: vimentin-positive cells (red) invading across the basement membrane the collagen matrix. DAB staining of TTF1, CK7, and SPP1 on paraffin sections: adenocarcinoma of the lung (E–G), HCC827 cells on SISmuc under static culture conditions (H–J), inserts: HCC827 in 2D, A549 cells on SISmuc under static culture conditions (K–M), inserts: A549 in 2D. Scale bar in (D,G) = 100 µm for (C–M); scale bar in m = 100 µm for (h–m).

Figure 2

EMT correlates with stemness, is inducible by TGF-β1, and is inversely correlated with Mucin-1 expression, independent of its location. Three different 3D tumor models (HCC827, H358, HCC44) with and without TGF-β1 (2 ng/mL) treatment are stained for different markers of EMT and stemness: pan-cytokeratin (PCK, green), E-cadherin (E-CAD, light blue), Mucin-1 (MUC-1, yellow), vimentin (VIM, red), and CD44 (purple). Cell nuclei are counterstained with DAPI (blue). Scale bar = 50 µm for all images.

Figure 3

EMT correlates to some extent with invasion, but the intrinsic invasion of HCC44 exceeds TGF-β1-induced invasion in H358 and HCC827. (A) Collagen IV (red) immunofluorescence staining with DAPI (blue) counterstaining of HCC44 tumor models or H358 and HCC827 3D models treated with 2 ng/mL TGF-β1. Invasive cells are indicated with white arrowheads. HCC827 gefitinib-resistant subclones A2 and A3 display a similar degree of invasion as parental HCC827 stimulated with TGF-β1. Scale bar = 100 µm (upper panel); 50 µm (lower panel); n = 4. (B) Quantitative evaluation of invasive cells; n = 4. Significance determined with unpaired t-tests versus H358 or HCC827, respectively. *: p ≤ 0.05, **: p ≤ 0.01, ***: p ≤ 0.001.

Figure 4

EMT, dedifferentiation, and stemness correlate with primary as well as secondary resistance. (A) Proliferation indices and (B) normalized cell numbers of H358 and HCC44 cells on 3D SISmuc tumor models with or without 1 µM ARS-1620 treatment; n = 4. Fold increase in apoptosis over untreated control evaluated by M30 ELISAs of (C) H358 and (D) HCC44 cells in 3D after the treatment with indicated concentrations of ARS-1620. Red line indicates the baseline apoptosis of the DMSO control. Triangles (▼) represent values from single biological replicates; n = 2. (E) Fold increase in apoptosis over untreated control (red line) and (F) proliferation indices in HCC827, HCCresA2 and HCCresA3 cells in 2D and 3D after treatment with 1 µM gefitinib; 4 ≤ n ≤ 13. (G) Immunohistochemistry staining of CD44 and immunofluorescence staining of EMT markers pan-cytokeratin (PCK, green), vimentin (VIM, red), E-cadherin (E-CAD, green), β-catenin (β-CAT, red), and Mucin-1 (MUC-1, green) of HCC827, HCCresA2, and HCCresA3; scale bar = 100 µm. Reflection electron microscopy (REM): white arrows indicate the elongated shape of resistant cells; scale bar = 50 µm. Significance determined with unpaired t-tests. **: p ≤ 0.01, ***: p ≤ 0.001.

Figure 5

In silico signaling network topology used for H358 and HCC44 therapy simulations. (A) The network of protein–protein interactions (rounded rectangles) and cellular responses (hexagons) integrates common co-mutations of KRAS. Interactions: arrows activating interactions, blunted arrows inhibitory interactions; white rectangles: interacting nodes, gray rectangles: assumed to be constant, yellow rectangles: constant nodes tested experimentally, orange rectangles: KRASmt node. (All coding also applies to the following simulations) (B) SQUAD calculates the activity changes and responses for every node in the network in detail. H358 and HCC44 treated with ARS-1620, with H358 being a responder to KRAS inhibition seen by induction of apoptosis and slight reduction in proliferation, which is not the case in simulations for HCC44 cells. (a) Untreated H358, (b) H358 treated with ARS-1620, (c) untreated HCC44, and (d) HCC44 treated with ARS-1620. Only the readout of interesting nodes of our network is shown. However, all nodes of the network as given in (A) are simulated and available in their trajectories so that novel drugs, as well as the detailed response of the whole network, can be studied.

Figure 6

EMT is more a marker than a maker of resistance. (A,C) H&E and immunofluorescence stainings of pan-cytokeratin (green) and vimentin (red) of 3D tumor models with H358 (A) and HCC827 cells (C) treated with 2 ng/mL TGF-β1 and 1 µM ARS-1620 or 1 µM gefitinib, respectively; n = 4. Scale bars = 100 µm. Proliferation indices and fold increase in apoptosis over untreated control of H358 (B) and HCC827 (D) cells in 3D after the treatment with 2 ng/mL TGF-β1 and 1 µM ARS-1620 (H358) or 1 µM gefitinib (HCC827). Red line indicates the baseline apoptosis of the corresponding controls. Significance determined with unpaired t-tests. *: p ≤ 0.05, **: p ≤ 0.01, ***: p ≤ 0.001; 4 ≤ n ≤ 6.

Figure 7

Neither EMT status nor CD44 expression correlates in PDX cell lines with drug response in 2D and 3D. (A) Pan-cytokeratin (PCK, green), vimentin (VIM, red), E-cadherin (light blue), and CD44 (purple) immunofluorescence staining of KRAS^G12C-mutated PDX-derived lung cancer cells in 2D (n = 2) and 3D (n = 2). Scale bar = 100 µm. (B) CellTiter-Glo viability assay of PDX-derived lung cancer cells after treatment with increasing concentrations of ARS-1620. Calculated IC₅₀ values for 2D cultures are indicated. The picture shows the IC₅₀ curve of one representative experiment of two independent assays; n = 2. (C) MTT-assay of 3D SISmuc tumor models seeded with the PDX-derived cell lines and treated with 1 µM ARS-1620 for 72 h. Significance determined with unpaired t-tests. *: p ≤ 0.05, ***: p ≤ 0.001; n = 4.

Figure 8

Combination of ARS1620 with AURKA inhibitor alisertib as the most effective combination in 3D HCC44 tumor models. (A) MTT assays of HCC44 tumor models treated with 1 mM metformin or 5 µM SHP099, erdafitinib, gefitinib, crizotinib, or alisertib. Drugs were tested either in monotherapies or in combination with 1 µM ARS-1620. Triangles (▼) represent values from single biological replicates; n ≥ 2. (B) Relative cell numbers and (C) proliferation indices of HCC44 cells in 3D after the treatment for 72 h with 1 µM ARS-1620, 5 µM alisertib and the combination of both inhibitors; n = 4. (D) In silico combination therapy simulations of (a) HCC44 treated with crizotinib and ARS-1620 and (b) HCC44 treated with alisertib and ARS-1620. Color code for different readout parameters is given on the right side of the figure. Significance determined with unpaired t-tests. *: p ≤ 0.05, **: p ≤ 0.01, ***: p ≤ 0.001.