In a recent publication, Rees et al.1 reported a new computational tool capable of elucidating small molecule mechanisms of action by correlating the sensitivity patterns of 481 compounds with 18,543 basal transcript levels across a large panel of human cancer cell lines (CCLs) and identifying differential gene expression patterns characterized by specific outlier transcripts. Overall, they measured the responses of 823 unique CCLs out of 947 previously characterized as part of the Cancer Cell Line Encyclopedia (CCLE) project.2 The data and correlation methods have been made publicly available through the Cancer Therapeutics Response Portal (CTRP; http://portals.broadinstitute.org/ctrp/ ).1
The authors indicated that hematopoietic CCLs were generally more sensitive to many small molecule perturbagens than were CCLs from solid tumors and they observed that outlier transcripts corresponding to putative biomarker genes for hematopoietic CCLs showed significant correlation with many compounds.1 They speculated that these correlations reflected the general sensitivity of hematopoietic CCLs, rather than a mechanistic connection of biomarker genes to compounds.
In particular, Rees et al.1 noted that an outlier transcript, TNFRSF12A, was the transcript most correlated with sensitivity to 36 small molecules enriched for several classes of drugs—including microtubule modulators, chromatin modifiers, topoisomerase inhibitors and nucleotide analogs—where low expression correlated with hematopoietic CCL sensitivity (z scores > 15.857; where r is the Pearson expression-sensitivity correlation coefficient and z is Fisher’s z(r)-scored Pearson expression-sensitivity correlation coefficient).
We recently demonstrated that activation of interacting transcriptional effectors of the Hippo signaling cascade, transcriptional co-activator with PDZ-binding motif (TAZ) encoded by the WWTR1 gene and TEA domain transcription factor 1 (TEAD1) involved in the transduction of cytoskeletal, adhesive and mechanical cues,3 contributed to the proteasome inhibitor-resistant phenotype of a multiple myeloma cell line.4 This was associated with elevated expression of TAZ-TEAD1 target genes (e.g., ITGB5 and CRIM1),5 as well as the tight junction protein gene TJP1 (also known as zonula occludens 1, ZO-1).4
This molecular phenotype was shared by a subset of multiple myeloma patients who developed relapsed/refractory disease while on proteasome inhibitor-based therapy. Additionally, Connectivity Map analysis predicted that translation inhibitors may overcome the drug-resistant state;6,7 we confirmed this prediction by showing that omacetaxine mepesuccinate (homoharringtonine)—the first translation inhibitor to be approved by the United States Food and Drug Administration—displayed potent cytotoxicity on the proteasome inhibitor-resistant multiple myeloma cells.4,8
In an attempt to obtain a generalizable biomarker of homoharringtonine sensitivity, I queried the CTRP using the CellMinerCDB web application (https://discover.nci.nih.gov/cellminercdb/ ).9 The RPL3 gene encoding ribosomal protein L3 of the large 60S ribosomal subunit bound by homoharringtonine was identified as a potential candidate (https://www.drugbank.ca/drugs/DB04865#targets );10,11 both hematopoietic and non-hematopoietic CCLs with increased RPL3 expression exhibit enhanced sensitivity to the drug (Fig. 1). This is an attractive possibility for further evaluation since up-regulation of ribosome biogenesis is frequently associated with tumor progression and adverse clinical outcomes.12,13
Unexpectedly, I also discovered that low TNFRSF12A expression correlated with sensitivity to homoharringtonine (93rd ranked compound out of 481; z score = 12.47) (Fig. 2A). Additionally, I found that low TNFRSF12A expression correlated with sensitivity to another translation inhibitor, the silvestrol analog CR-1-31B (45th ranked compound out of 481; z score = 15.46) which inhibits cap-dependent translation initiation by targeting the eIF4A RNA helicase (Fig. 2B).14,15 I subsequently learned that TNFRSF12A is a direct target of TEAD transcription factors complexed with TAZ or its paralog YAP1.16 Notably, TNFRSF12A is coexpressed (top 20 gene neighbors) with WWTR1/TAZ, YAP1, TEAD1, ITGB5, CRIM1 and TJP1 as well as other YAP/TAZ direct targets (e.g., CYR61)16 across all CCLE CCLs (Fig. S1).
Hematopoietic CCLs generally expressed lower levels of these transcripts than non-hematopoietic CCLs (Fig. 3, Fig. S1). Furthermore, DAVID analysis of the overlapping outlier transcripts for homoharringtonine and CR-1-31B sensitivity (homoharringtonine, z score > 9; CR-1-31B, z score > 11) (Fig. 4) revealed that the most significantly enriched cluster comprises YAP/TAZ annotation terms (Table S1),17 whereas oPOSSUM-3 transcription factor binding site motif discovery identified TEAD1 as the top-ranked overrepresented motif in the set of overlapping genes (Table S1).18
Further inspection of the list of 481 compounds in the CTRP uncovered a third translation inhibitor, the silvestrol analog, SR-II-138A. It turned out that SR-II-138A is among the 36 compounds most highly associated with low TNFRSF12A expression (24th ranked compound out of 481; z score = 16.56) (see Supplementary Data Set 7 of Rees et al.1). I interpret these observations to indicate that the generally increased sensitivity of hematopoietic CCLs to several classes of drugs compared to non-hematopoietic CCLs is due to lower basal levels of activation of TAZ-TEAD mechanotransduction pathways.3,19 Translation inhibitors, which appear to be a class of drugs that can be included in this category, may thus demonstrate therapeutic efficacy against hematologic malignancies, in front-line as well as in certain relapsed/refractory settings.4,8
Supporting information
Supplementary material for this article is available at https://doi.org/10.14218/ERHM.2017.00022 .
Fig. S1
Heat maps of the top 20 neighbors of TNFRSF12A, YAP1, WWTR1/TAZ, TEAD1, TJP1, ITGB5 and CRIM1 across all CCLs in the CCLE.2
Dark red indicates the highest expression levels. Dark blue indicates the lowest expression levels. TNFRSF12A is indicated with a red asterisk and YAP1, WWTR1/TAZ, TEAD1, TJP1, ITGB5 and CRIM1 are indicated with blue asterisks on the right side of the figure. Note that the expression levels of TNFRSF12A, YAP1, WWTR1/TAZ, TEAD1, TJP1, ITGB5 and CRIM1 are generally lower in hematopoietic CCLs (samples labeled “HAEMATOPOIETIC_AND_LYMPHOID_TISSUE”) than in non-hematopoietic CCLs.
(PDF)
Table S1
The complete lists of outlier transcripts for homoharringtonine and CR-1-31B sensitivity.
(XLSX)
Abbreviations
- CCLE:
Cancer Cell Line Encyclopedia
- CCLs:
cancer cell lines
- CTRP:
Cancer Therapeutics Response Portal
- r:
Pearson expression-sensitivity correlation coefficient
- TAZ:
transcriptional co-activator with PDZ-binding motif
- TEAD1:
TEA domain transcription factor 1
- TJP1:
tight junction protein 1 (also known as zonula occludens 1)
- z:
Fisher’s z(r)-scored Pearson expression-sensitivity correlation coefficient.
Declarations
Acknowledgement
I thank Irene Riz for critical comments on the manuscript. The author gratefully acknowledges financial support from the Katzen Cancer Research Center at the GW Cancer Center and a generous philanthropic gift from Marc Cohen. RGH was supported in part by NIH grant R21AI113806.
Conflict of interest
The author has no conflict of interest related to this publication.
Authors’ contributions
Manuscript writing (RGH).