Somatic mutations in early metazoan genes disrupt regulatory links between unicellular and multicellular genes in cancer

  1. Anna S Trigos
  2. Richard B Pearson
  3. Anthony T Papenfuss
  4. David L Goode  Is a corresponding author
  1. Peter MacCallum Cancer Centre, Australia
  2. The University of Melbourne, Australia
  3. Monash University, Australia
  4. The Walter & Eliza Hall Institute of Medical Research, Australia
11 figures and 7 additional files

Figures

Figure 1 with 9 supplements
Enrichment of CNAs and point mutations in EM genes.

(A) Fraction of amplified (left) and deleted (right) genes across phylostrata. EM genes are preferentially copy-number altered across tumor types, whereas MM genes are depleted. (B) Fraction of …

https://doi.org/10.7554/eLife.40947.003
Figure 1—figure supplement 1
Ratio of the number of missense and LoF mutations over synonymous mutations.

Known cancer genes (derived from the Cancer Census database) have a higher ratio that other genes, indicating that this metric captures genes more likely to be drivers. Only genes with missense or …

https://doi.org/10.7554/eLife.40947.004
Figure 1—figure supplement 2
Phylogenetic tree depicting gene phylostrata.

Genes assigned to earlier phylostrata (smaller numbers) are more ancient as they are across multiple phylogenetic groups of the tree of life, whereas genes assigned to later phylostrata (larger …

https://doi.org/10.7554/eLife.40947.005
Figure 1—figure supplement 3
Number of genes in each phylostratum.

Genes were assigned to phylostratum based on the age of the most ancient ancestor with an ortholog of the gene using phylostratigraphy. 38.80% (6719) of human genes are of UC origin (red), 45.84% …

https://doi.org/10.7554/eLife.40947.006
Figure 1—figure supplement 4
Enrichment in EM genes of recurrent CNAs and point mutations identified by Gistic and MutSig2CV.

(A) Fraction of amplified, (B) deleted and point mutated (C) genes across phylostrata. At least 3 of the top five most recurrently affected phylostrata by amplifications were EM in 26/29 tumor …

https://doi.org/10.7554/eLife.40947.007
Figure 1—figure supplement 5
Fraction of non-recurrent amplified genes.

We calculated the fraction of genes with non-recurrent amplifications of all genes with amplification. Although most mutated genes were non-recurrent across phylostrata, mutated MM genes are …

https://doi.org/10.7554/eLife.40947.008
Figure 1—figure supplement 6
Fraction of non-recurrent deleted genes.

We calculated the fraction of genes with non-recurrent deletions of all genes with deletions. Although most mutated genes were non-recurrent across phylostrata, mutated MM genes are particularly …

https://doi.org/10.7554/eLife.40947.009
Figure 1—figure supplement 7
Fraction of non-recurrent missense mutations.

We calculated the fraction of genes with non-recurrent missense mutations of all genes with missense mutations. Although most mutated genes were non-recurrent across phylostrata, mutated MM genes …

https://doi.org/10.7554/eLife.40947.010
Figure 1—figure supplement 8
Fraction of non-recurrent loss-of-function mutations.

We calculated the fraction of genes with non-recurrent loss-of-function mutations of all genes with loss-of-function mutations. Although most mutated genes were non-recurrent across phylostrata, …

https://doi.org/10.7554/eLife.40947.011
Figure 1—figure supplement 9
Presence of genes by phylostratum by fraction of aberrant chromosome.

The overall increasing trend across chromosomes and tumor types indicated that earlier genes tend to be located in more focal region of copy-number changes, whereas later genes tend to be located in …

https://doi.org/10.7554/eLife.40947.012
Figure 2 with 6 supplements
Point mutations in EM genes affect mostly regulators, whereas CNAs in EM genes affect downstream targets.

(A) Diagram of a GRN distinguishing regulator and target genes. The number of outgoing edges from a regulator corresponds to its out-degree, whereas the number of incoming edges to a target gene is …

https://doi.org/10.7554/eLife.40947.013
Figure 2—figure supplement 1
Distribution of out-degree of regulators in the GRN.

Genes in the upperquantile of the distribution (log10 out-degree greater or equal than 1) were selected as master regulators.

https://doi.org/10.7554/eLife.40947.014
Figure 2—figure supplement 2
Degree of genes across multiple molecular networks.

Although in protein-protein interaction networks (PahtwayCommons, Biogrid and WebIM) UC genes are the most highly connected, in the GRN EM genes are more likely to be hubs. The normalized degree was …

https://doi.org/10.7554/eLife.40947.015
Figure 2—figure supplement 3
Distribution of in-degree of target genes in the GRN.

EM genes tended to have a greater in-degree than UC and MM genes (Wilcoxon test p<2.2×10−16 in both cases), indicating these genes are highly regulated.

https://doi.org/10.7554/eLife.40947.016
Figure 2—figure supplement 4
Fraction of regulators and targets with mutations.

Genes with point mutations, especially LoF mutations, affect a higher fraction of regulators than genes with amplifications or deletions (Wilcoxon p for LoF mutations = 1.10×10−7, for missense …

https://doi.org/10.7554/eLife.40947.017
Figure 2—figure supplement 5
Fraction of mutated regulator and target genes by each mutation type.

Recurrent point mtuations were derived using MutSig2CV, and significant driver CNAs by Gistic. Point mutations affected a higher fraction of regulators (mean fraction altered = 0.44) than CNAs …

https://doi.org/10.7554/eLife.40947.018
Figure 2—figure supplement 6
Ratio of out-degree/in-degree (log2) of genes with mutations.

EM genes with point mutations held the strongest regulatory role across tumors (median ratio = 1.88), whereas UC and MM genes with point mutations did less so (median ratio = 1.33 and 0.27, …

https://doi.org/10.7554/eLife.40947.019
Figure 3 with 3 supplements
Point mutations in regulators affect UC-EM gene regulation.

(A) Classification of regulators by the age of their downstream targets. UC-t regulators mostly regulate UC genes, EM-t regulators EM genes, and UC/EM-i regulators are at the interface of UC and EM …

https://doi.org/10.7554/eLife.40947.020
Figure 3—figure supplement 1
Distribution of recurrent point mutations and CNAs identified by MutSig2CV and Gistic in regulators.

(A) (Lower panel) Percentage of UC, EM and MM target genes in regulators. (Upper panel) Distribution of recurrent point mutations (dark grey) and CNAs (light grey) across regulators. UC/EM-i …

https://doi.org/10.7554/eLife.40947.021
Figure 3—figure supplement 2
Distribution of CNAs and point mutations across recurrently mutated regulators.

Mutated UC-t and EM-t regulators were preferentially affected by CNAs (10/10 mutated UC-t regulators, 33/34 mutated EM-t regulators), whereas UC/EM-i regulators had the highest proportion of …

https://doi.org/10.7554/eLife.40947.022
Figure 3—figure supplement 3
Effect of point mutations in regulators on the expression of downstream genes.

Point mutations with a high impact (>5% differentially expressed downstream genes) are more likely to affect EM genes, rather than those with a limited downstream effect (<5%). The prevalence was …

https://doi.org/10.7554/eLife.40947.023
Figure 4 with 4 supplements
CNAs directly regulate the expression of UC and EM target genes.

(A) Fraction of downstream targets with CNAs in regulators. Targets of UC-t and EM-t regulators are more likely to be affected by CNAs than targets of UC/EM-i regulators. (B) Percentage of …

https://doi.org/10.7554/eLife.40947.024
Figure 4—figure supplement 1
Targets are more likely differentially expressed after CNAs than regulators.

We calculated the difference in the percentage of differentially expressed targets and regulators that were CNA. Values greater than 0 indicate a higher percentage of differentially expressed …

https://doi.org/10.7554/eLife.40947.025
Figure 4—figure supplement 2
Percentage of differentially expressed target genes with CNAs.

UC and EM target genes are more likely to be upregulated after amplifications and downregulated after deletions compared to younger, mammalian-specific genes. Jonckheere-Terpstra decreasing trend …

https://doi.org/10.7554/eLife.40947.026
Figure 4—figure supplement 3
The median fraction of targets with CNAs by regulator status.

Copy-number normal (CNN) regulators have a higher fraction of targets with CNAs than CNA regulators (Wilcoxon one-sided p-value=4.10×10−29).

https://doi.org/10.7554/eLife.40947.027
Figure 4—figure supplement 4
Difference in the fraction of downstream targets altered by CNAs when their regulators are CNN or CNA.

A higher fraction of CNA targets with a CNN regulator is observed for UC-t and EM-t regulators, but not for UC/EM-i regulators, regardless of regulator age. However, the opposite trend is more …

https://doi.org/10.7554/eLife.40947.028
Figure 5 with 7 supplements
UC/EM-i regulators are fundamental to tumor development and drug response.

(A) Fraction of known cancer drivers of each regulator class. While only 33% of regulators are UC/EM-i, 47% of cancer drivers are UC/EM-i regulators, indicating an enrichment of this regulator class …

https://doi.org/10.7554/eLife.40947.029
Figure 5—figure supplement 1
Number of driver and clonal point mutated regulators in 100 non-small-cell lung cancers from the TRACERx study (Jamal-Hanjani et al., 2017) and 50 breast cancers (Yates et al., 2015) obtained by Caravagna et al. (2018).

A high prevalence of UC/EM-i regulators as clonal drivers across lung cancer and breast cancer patients is observed, with a significant enrichment (p=1.16×10−11) of UC/EM-i regulators among clonal …

https://doi.org/10.7554/eLife.40947.030
Figure 5—figure supplement 2
Distribution of probabilities of dependency to regulators from the Avana CRISPR-Cas9 genome-scale knockout dataset.

The knockout of regulators generally reveals a low probability of dependency (the highest point of the distributions is close to 0). However, the knockout of a small number of regulators reveal a …

https://doi.org/10.7554/eLife.40947.031
Figure 5—figure supplement 3
Odds ratio of dependency of regulators in cancer cell lines after knockout at different probability cutoffs.

Regardless of the cutoff used, UC-t and UC/EM-i regulators are enriched (odds ratio >0) in regulators with high dependency.

https://doi.org/10.7554/eLife.40947.032
Figure 5—figure supplement 4
Correlation between cell line dependency and IC50 Distribution of correlations between cell line dependency to regulators and the IC50 (ln) of 250 drugs.

We selected correlations < −0.25 and with a p-value<0.05 as significant.

https://doi.org/10.7554/eLife.40947.033
Figure 5—figure supplement 5
Significant correlations between the probability of dependency to UC/EM-i regulators and the IC50 (ln) of drugs.

Only significant (p<0.05) Spearman correlations (<−0.25) were selected, which are cases where an increased dependency to the regulator was associated greater drug sensitivity at lower concentrations …

https://doi.org/10.7554/eLife.40947.034
Figure 5—figure supplement 6
Distribution of correlations of the probability of cell-line dependency to regulators and the IC50 (ln) of drugs.

The red line indicates the genes whose probability of dependency is significantly correlated (correlation <0.25 and p<0.05) with the IC50 of the drug.

https://doi.org/10.7554/eLife.40947.035
Figure 5—figure supplement 7
Correlation between drug sensitivity per cell-life tissue type and PPRC1 cell-line dependency.

Drug sensitivity to dactolisib, docetaxel, temsirolimus and YK-4–279 is correlated with PPRC1 dependency across cells lines of multiple tissue types. The red line indicates a strong correlation of …

https://doi.org/10.7554/eLife.40947.036
Appendix 1—figure 1
Results obtained using the TRRUST and RegNetwork databases pertaining to network composition and distribution of mutations in regulators and targets.

The top row corresponds to results obtained with TRRUST, and the bottom row those obtained with RegNetwork. The results are largely consistent with those obtained with the GRN from PathwayCommons.

https://doi.org/10.7554/eLife.40947.045
Appendix 1—figure 2
Results obtained using the TRRUST and RegNetwork databases related to the regulatory and target roles of genes with point mutations and CNAs.

In both the TRRUST (A) and RegNetwork (B), EM genes with point mutations have a stronger regulatory role (high out-degree/in-degree ratio) than UC and MM genes with point mutations. In contrast, …

https://doi.org/10.7554/eLife.40947.046
Appendix 1—figure 3
Concordance of regulator classification across databases.

The classification shown on the left corresponds to that obtained with GRN from PathwayCommons. Many regulators are only found in this databases (large sections of dark red). Although there is some …

https://doi.org/10.7554/eLife.40947.047
Appendix 1—figure 4
Results obtained using the TRRUST and RegNetwork databases pertaining to point mutations and CNAs in different classes of regulators.

The top row corresponds to results obtained with TRRUST, and the bottom row those obtained with RegNetwork. The results are largely consistent with those obtained with the GRN from PathwayCommons.

https://doi.org/10.7554/eLife.40947.048
Author response image 1
Author response image 2

Additional files

Supplementary file 1

Point mutated genes.

Genes marked as ‘Frequent’ were included in the analysis.

https://doi.org/10.7554/eLife.40947.037
Supplementary file 2

Ratio of out-degree and in-degree per age and mutation type.

https://doi.org/10.7554/eLife.40947.038
Supplementary file 3

Regulator classification.

https://doi.org/10.7554/eLife.40947.039
Supplementary file 4

Functional enrichment analysis of UC/EM-i regulators using gprofileR.

https://doi.org/10.7554/eLife.40947.040
Supplementary file 5

Significance of change of pathway activity levels after point mutation of UC/EM-i regulators.

https://doi.org/10.7554/eLife.40947.041
Supplementary file 6

Difference in median dependency between cell lines with regulator mutated and non-mutated.

https://doi.org/10.7554/eLife.40947.042
Transparent reporting form
https://doi.org/10.7554/eLife.40947.043

Download links