Machine Learning Discoveries of ANTXR2-X Synergy in etc-1922159 Treated Colorectal Cancer Cells

1. Introduction

In the unpublished preprint Sinha [1], a frame work of a search engine was developed which can rank combinations of factors (genes/proteins) in a signaling pathway. Such combinations are of import due to the vast search space in which they exist and the difficulty to find them. The search engine facilitates in prioritizing the combinations as ranked biological hypotheses which the biologists might want to test in wet lab, to know if a synergistic combination is prevalent in a signaling pathway, in a direct or indirect manner. Interested readers are advised to go through unpublished preprints Sinha [1] and Sinha [2] for details regarding the search engine and the discoveries mentioned in there.

2. Materials and Methods

2.1. Combinatorial Search Problem and a Possible Solution

The issue of combinatorial search problem and a possible solution has been addressed in Sinha [3] and Sinha [2]. The details of the methodology of this manuscript have been explained in great detail in Sinha [3] & its application in Sinha [2]. Readers are requested to go through the same for gaining deeper insight into the working of the pipeline and its use of published data set generated after administration of ETC-1922159. In order to understand the significance of the solution proposed to the problem of combinatorial search that the biologists face in revealing unknown biological search problem, these works are of importance.

Briefly, from Sinha [2], the pipleline works by computing sensitivity indicies for each of these unique combinations and then vectorising these indices to connote and form discriminative feature vector for each combination. Since each combination is unique, the training and the test data are same. In the training data, the combinations are arranged and ranks from 1 to n are assigned. The ranking algorithm then learns the patterns from these combinations/sensitivity index vectors. Next the learned model is used to rank the test data by generating the ranking score for each of the unique combination. Sorting these shuffled scores of test data leads to prioritization of the combinations. Joachims [4] show an example of applying learned model to training data (same as the test data) in https://www.cs.cornell.edu/people/tj/svm_light/svm_rank.html. Note that these combinations are now ranked and give the biologists a chance to narrow down their focus on crucial biological hypotheses in the form of combinations which the biologists might want to test. Analogous to the webpage search engine, where the click of a button for a few key-words leads to a ranked list of web links, the pipeline uses sensitivity indices as an indicator of the strength of the influence of factors or their combinations, as a criteria to rank the combinations.

3. Results & Discussion

3.1. Anthrax Toxin Receptor Related Synergies

3.1.1. ANTXR2 - Collagen Cross Family Analysis

Anthrax toxin receptor ANTRX is known to capture the Bacillus anthracis toxin and form the cause of the anthrax disease. Regulatory mechanism of the ANTXR1 has been demostrated essential component in the fibrosis processes in fibroproliferative diseases. Loss of ANTXR1 (a.k.a TEM8) in fibroblasts leads to increased rates of synthesis of fiber-forming collagens, resulting in progressive fibrosis in skin and other organs Besschetnova et al. [5]. TEM8 interacts with the cleaved C5 domain of collagen 3(VI) Nanda et al. [6]. Hotchkiss et al. [7] also indicate the interaction of TEM8 and collagens. Bell et al. [8] indicate that a recombinant portion of ANTXR2 (a.k.a CMG2) was found to bind collagen type IV and laminin, suggesting a potential role in basement membrane matrix synthesis and assembly. Bürgi et al. [9] show that CMG2/ANTXR2 regulates extracellular collagen VI which accumulates in hyaline fibromatosis syndrome. A distinctive early childhood-onset disorder, systemic hyalinosis, is characterized by mutations in the anthrax toxin receptor 2 gene (ANTRX2) as shown by Shieh et al. [10]. Not much is known about the behaviour of ANTRX2 with collagens in colorectal cancer. In CRC cells treated with ETC-1922159 these were up regulated. Our search engine was able to rank the 2nd order combinations between these two to see if there is a possible existing synergy based on the already explored pathological functionality in the above works. At in silico level we found possible high numerical valued ranks pointed to some of the combinations of ANTXR2 and collagen family.

On the left side of Table 1 we found ANTXR2 to be up regulated w.r.t COL-5A3/28A1 (probably COL-7A1 also). These are reflected in rankings of 2006 (laplace) and 2217 (rbf) for COL5A3-ANTXR2; 2259 (linear) and 2296 (rbf) for COL28A1-ANTXR2 and probably 2119 (linear) and 1690 (rbf) for COL7A1-ANTXR2. On the right side, we found, COL-7A1/28A1 up regulated w.r.t ANTXR2. These are reflected in rankings of 1722 (laplace) and 2121 (rbf) for COL7A1-ANTXR2 and 1731 (linear) and 2362 (rbf) for COL28A1-ANTXR2. Table 2 shows the derived influences which can be represented graphically, with the following influences - • ANTXR2 w.r.t COL with COL5A3 $->$ ANTXR2 and COL28A1 $->$ ANTXR2; and • COL w.r.t ANTXR2 with COL7A1 $<-$ ANTXR2 and COL28A1 $<-$ ANTXR2.

3.1.2. ANTXR2 - Integrin Cross Family Analysis

To define whether there is a possible functional cooperation between TEM8/ANTXR1 and integrins to accomplish cell spreading, Werner et al. [11] sought to disrupt the function of collagen binding integrins, which mostly belong to $\beta$1 family of integrins, with $\beta$1 integrin-blocking antibodies. Their experiments suggest that the participation of $\beta$1 integrins can be excluded in TEM8-mediated cell spreading on collagen in primary fibroblasts. However, collectively, they indicate that the participation of these integrins in TEM8 spreading on collagen vary with cell type. The cytoplasmic domain of ANTXR1 affects binding of the protective antigen which is similar to integrin I domains that convert between open and closed conformations that bind ligand with high and low affinities, respectively. These findings are observed by Go et al. [12]. Scobie et al. [13] observe similar behaviour. Abnormal clustering of TEM8/ANTXR1 with integrin $\beta$1 and vascular endothelial growth factor receptor 2 (VEGFR2) occurs in endothelial cells within cutaneous infantile hemangiomas, the most common vascular anomaly in childhood Besschetnova et al. [5]. These findings suggest the possibility to synergy between ANTXR2 and Integrin family. In CRC cells treated with ETC-1922159, ANTXR2 and integrin families were up regulated. The search engine alloted high numerical valued ranks to some of the 2^nd order combinations of ANTXR2 and integrin family members thus pointing to possible synergy in CRC cells. Table 3 shows the rankings of ANTRX2 along with integrin family members.

On the left side, we found ANTRX2 to be up regulated w.r.t ITG-A2/A3/B1BP1/B5/B8. These are reflected in the rankings of 2261 (laplace) and 2444 (rbf) for ITGA2-ANTXR2; 2027 (laplace) and 2134 (linear) for ITGA3-ANTXR2; 2444 (laplace) and 2128 (rbf) for ITGB1BP1-ANTXR2; 1860 (linear)and 2315 (rbf) for ITGB5-ANTXR2; and 2354 (linear)and 2136 (rbf) for ITGB8-ANTXR2. On the right side, we found ITGB6 to be up regulated w.r.t ANTRX2. These are reflected in the rankings of 2000 (linear) and 1896 (rbf) for ITGB6-ANTXR2; Table 4 shows the derived influences which can be represented graphically, with the following influences - • ANTXR2 w.r.t ITG with ITG-A2/A3 $->$ ANTXR2 and ITG-B1BP1/B5/B8 $->$ ANTXR2; and • ITG w.r.t ANTXR2 with ITGB6 $<-$ ANTXR2.

3.1.3. ANTXR2 - MMP Cross Family Analysis

Compromised interactions between TEM8/ANTXR1-deficient endothelial and fibroblastic cells cause dramatic reduction in the activity of the matrix-degrading enzyme MMP2 Besschetnova et al. [5]. They observe experimentally that loss of MMP2 activity requires loss of TEM8/ANTXR1 function in both endothelial and fibroblastic cells. Matrix metalloproteinases (MMP) are members of the metzincin group of proteases which share the conserved zinc-binding motif in their catalytic active site Löffek et al. [14]. these enzymes are capable of degrading all kinds of extracellular matrix proteins, but also can process a number of bioactive molecules as well as play a major role in cell behaviors such as proliferation, migration, differention, apoptosis and host defense, Wikipedia contributors [15]. In CRC cells treated with ETC-1922159, ANTXR2 and integrin families were up regulated. The search engine alloted high numerical valued ranks to some of the 2^nd order combinations of ANTXR2 and integrin family members thus pointing to possible synergy in CRC cells. Table 5 shows the rankings of ANTRX2 along with MMP family members.

On the left side, we found ANTXR2 to be upregulated w.r.t MMP28. This is reflected in rankings of 2468 (laplace) and 1765 (linear) for MMP28-ANTXR2. On the right side we found MMP-1/15/28 up regulated w.r.t ANTXR2. These are reflected in rankings of 2009 (linear) and 2142 (rbf) for MMP1-ANTXR2; 2219 (linear) and 1926 (rbf) for MMP15-ANTXR2; and 1857 (linear) and 2092 (rbf) for MMP28-ANTXR2. Table 6 shows the derived influences which can be represented graphically, with the following influences - • ANTXR2 w.r.t MMP with ANTXR2 $<-$ MMP28 and • MMP w.r.t ANTXR2 with MMP-1/15/28 $<-$ ANTXR2.

3.1.4. ANTXR2 - WNT Cross Family Analysis

Abrami et al. [16] show that LRP6 can indeed form a complex with ATRs (anthrax toxin receptors), and that this interaction plays a role both in Wnt signalling and in anthrax toxin endocytosis. Through the ATR-LRP6 interaction, adhesion to the extracellular matrix could locally control Wnt signalling. The authors demonstrated that physical and functional interaction between CMG2/ANTXR2 and LRP6 also raised the possibility that the complex clinical manifestation of Systemic Hyalinosis might be due in part to defects in Wnt signalling. Fluorescence microscopy and biochemical analyses showed that LRP6 enables toxin internalization by interacting at the cell surface with PA receptors TEM8/ATR and/or CMG2/ANTXR2 to form a multi-component complex that enters cells upon PA binding (Wei et al. [17]). Verma et al. [18] postulate that the developmentally controlled expression of TEM8 modulates endothelial cell response to canonical Wnt signaling to regulate vessel patterning and density. These findings definitely indicate the synergy of ANTRX with Wnts. In CRC cells treated with ETC-1922159, ANTXR2 and WNT families were up regulated. The search engine alloted high numerical valued ranks to some of the 2^nd order combinations of ANTXR2 and WNT family members thus pointing to possible synergy in CRC cells. Table 7 shows the rankings of ANTRX2 along with WNT family members.

On the left side, we found ANTXR2 to be upregulated w.r.t WNT4. This is reflected in rankings of 1833 (linear) and 2341 (rbf) for WNT4-ANTXR2. Table 8 shows the derived influences which can be represented graphically, with the following influences - • ANTXR2 w.r.t WNT with ANTXR2 $<-$ WNT4. This synergistic upregulation of the WNT4 with ANTXR2 might indicate possible control over the signalling in CRC cells treated with ETC-1922159.

3.1.5. ANTXR2 - TNF Cross Family Analysis

The author could not find much about TNF-ANTXR2 combinations in pathological cases in existing literature, however, Lee et al. [19] report the "both LeTx and EdTx markedly inhibited LPS-induced transcription of tumour necrosis factor alpha (TNF-$\alpha$), interleukin (IL)-1$\beta$, and IL-6 in J774A.1 cells. In contrast, EdTx synergised with LPS to increase the transcription of IL-6 and IL-8 in HAECs. We showed that HAECs are suitable for anthrax toxin research and express higher levels of the two anthrax toxin receptors - tumour endothelial marker 8 (TEM8/ANTXR1) and capillary morphogenesis protein 2 (CMG2/ANTXR2) - than do J774A.1 cells". The high expression of the ANTXR-1/2 is shown, however, the possible synergy between ANTXR and TNFs in not shown. Our search engine pointed to some of the combinations in CRC cells treated with ETC-1922159 treatment. In Table 9, on the left we found ANTXR2 to be up regulated w.r.t TNF-AIP1/RSF10B. These are reflected in rankings of 1769 (linear) and 1946 (rbf) for TNFAIP1-ANTXR2 and 2278 (linear) and 2218 (rbf) for TNFRSF10B-ANTXR2. On the right we found, TNF-RSF10A/RSF10D/RSF12A/RSF14 was up regulated w.r.t ANTXR2. These are reflected in rankings of 2260 (laplace) and 2377 (rbf) for TNFRSF10A-ANTXR2, 2258 (laplace) and 2363 (rbf) for TNFRSF10D-ANTXR2, 2190 (laplace) and 2061 (rbf) for TNFRSF12A-ANTXR2 and 2370(laplace) and 1777 (linear) for TNFRSF14-ANTXR2.

Table 10 shows the derived influences which can be represented graphically, with the following influences - • ANTXR2 w.r.t TNF with ANTXR2 $<-$ TNFAIP1 and ANTXR2 $<-$ TNFRSF10B and • TNF w.r.t ANTXR2 with TNFRSF10A $<-$ ANTXR2; TNFRSF10D $<-$ ANTXR2; TNFRSF12A $<-$ ANTXR2 and TNFRSF14 $<-$ ANTXR2. This synergistic upregulation of the TNF with ANTXR2 might indicate possible control over the signalling in CRC cells treated with ETC-1922159.

3.1.6. ANTXR2 - IL Cross Family Analysis

The author could not find much about TNF-ANTXR2 combinations in pathological cases in existing literature, however, Lee et al. [19] report the "both LeTx and EdTx markedly inhibited LPS-induced transcription of tumour necrosis factor alpha (TNF-$\alpha$), interleukin (IL)-1$\beta$, and IL-6 in J774A.1 cells. In contrast, EdTx synergised with LPS to increase the transcription of IL-6 and IL-8 in HAECs. We showed that HAECs are suitable for anthrax toxin research and express higher levels of the two anthrax toxin receptors - tumour endothelial marker 8 (TEM8/ANTXR1) and capillary morphogenesis protein 2 (CMG2/ANTXR2) – than do J774A.1 cells". The high expression of the ANTXR-1/2 is shown, however, the possible synergy between ANTXR and IL in not shown. Our search engine pointed to some of the combinations in CRC cells treated with ETC-1922159 treatment. In Table 11, on the left we found ANTXR2 to be up regulated w.r.t IL-1RN/6ST/17C/17REL. These are reflected in rankings of 1914 (linear) and 1894 (rbf) for IL1RN-ANTXR2; 1944 (laplace), 2219 (linear) and 1914 (rbf) for IL6ST-ANTXR2; 1832 (laplace) and 2334 (linear) for IL17C-ANTXR2 and 1889 (linear) and 2303 (rbf) for IL17REL-ANTXR2. On the right we found, IL-1A/1B/6ST/17C was up regulated w.r.t ANTXR2. These are reflected in rankings of 2356 (linear) and 1859 (rbf) for IL1A-ANTXR2; 1780 (linear) and 1865 (rbf) for IL6ST-ANTXR2; 1924 (laplace) and 1901 (rbf) for IL15RA-ANTXR2; and 2121 (linear) and 2437 (rbf) for IL17C-ANTXR2.

Table 12 shows the derived influences which can be represented graphically, with the following influences - • ANTXR2 w.r.t IL with ANTXR2 $<-$ IL1RN; ANTXR2 $<-$ IL6ST; ANTXR2 $<-$ IL17C and ANTXR2 $<-$ IL17REL; and • IL w.r.t ANTXR2 with IL1A $<-$ ANTXR2; IL1B $<-$ ANTXR2; IL6ST $<-$ ANTXR2; and IL17C $<-$ ANTXR2.