Skip to main page content
U.S. flag

An official website of the United States government

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2016 Sep;15(9):3045-57.
doi: 10.1074/mcp.M115.057729. Epub 2016 Jun 30.

Proteomic Screening and Lasso Regression Reveal Differential Signaling in Insulin and Insulin-like Growth Factor I (IGF1) Pathways

Cemal Erdem  1 , Alison M Nagle  2 , Angelo J Casa  3 , Beate C Litzenburger  3 , Yu-Fen Wang  3 , D Lansing Taylor  1 , Adrian V Lee  4 , Timothy R Lezon  5
Affiliations

Proteomic Screening and Lasso Regression Reveal Differential Signaling in Insulin and Insulin-like Growth Factor I (IGF1) Pathways

Cemal Erdem et al. Mol Cell Proteomics. 2016 Sep.

Abstract

Insulin and insulin-like growth factor I (IGF1) influence cancer risk and progression through poorly understood mechanisms. To better understand the roles of insulin and IGF1 signaling in breast cancer, we combined proteomic screening with computational network inference to uncover differences in IGF1 and insulin induced signaling. Using reverse phase protein array, we measured the levels of 134 proteins in 21 breast cancer cell lines stimulated with IGF1 or insulin for up to 48 h. We then constructed directed protein expression networks using three separate methods: (i) lasso regression, (ii) conventional matrix inversion, and (iii) entropy maximization. These networks, named here as the time translation models, were analyzed and the inferred interactions were ranked by differential magnitude to identify pathway differences. The two top candidates, chosen for experimental validation, were shown to regulate IGF1/insulin induced phosphorylation events. First, acetyl-CoA carboxylase (ACC) knock-down was shown to increase the level of mitogen-activated protein kinase (MAPK) phosphorylation. Second, stable knock-down of E-Cadherin increased the phospho-Akt protein levels. Both of the knock-down perturbations incurred phosphorylation responses stronger in IGF1 stimulated cells compared with insulin. Overall, the time-translation modeling coupled to wet-lab experiments has proven to be powerful in inferring differential interactions downstream of IGF1 and insulin signaling, in vitro.

PubMed Disclaimer

Figures

Fig. 1.
Fig. 1.
Three time-translation models are constructed. The overview of model construction steps is visualized. First, the data set was filtered, mean-centered, normalized, and partitioned into individual time point matrices. Three time-translation models were constructed for full (A) and leave-one-out analyses (B). (A) The construction steps of the three models. Construction of the TC model matrix using the joint covariance matrix, C (steps 1–4). Construction of the TM model matrix using entropy maximization (steps 1, 2, 5–7). Construction of TL model matrix using the lasso algorithm (steps 1, 8–10). Here, the glmnet package was called for each of the P protein expression vector, and the median matrix from an ensemble of R networks was selected as the model matrix (see Methods for details). These three matrices (steps 4, 7, 10) represent three models of time translation. (B) A leave-one-out cross validation (LOOCV) analysis was performed to check the validity of our approach. The information for test cell line (tan and red bars) was hidden and the model matrix was constructed via the scheme in (A). After the three time translation matrices are generated in either (A) or (B), they are used to calculate the expression profiles of the future time points using SFM condition data and corresponding T matrix.
Fig. 2.
Fig. 2.
Time-course (phospho)-protein expression profiles of 21 breast cancer cell lines in response to IGF1. The change in the (phospho)-protein levels upon IGF1 stimulation are shown. The values represent the difference in log2 expression levels (stimulated - serum starved). IGF1R phosphorylation is induced in all cell lines in response to IGF1. RAS/MAPK and PI3K/Akt cascades are regulated downstream of IGF1R signaling. Six columns are shown for each cell line, each column representing one time point (5, 10, 30 min, 6, 24, 48 h). Red color corresponds to up-regulation and blue color represents down-regulation of the proteins compared with serum starved cells. Row headings indicate proteins or phospho-proteins with phosphorylation sites indicated after the subscript p.
Fig. 3.
Fig. 3.
Performance analyses of the three models reveal that lasso models are accurate and robust. Three matrices were constructed for full (Fig. 1A) and LOOCV (Fig. 1B) models of each time translation. Correlation plots of data and prediction results from the three models for in-training (left) and leave-one-out cases (right) are shown. Black dots in the left panels represent data versus full model predictions, and black diamonds on the right represent the data versus LOOCV model predictions. The red lines are the least-square fits. The Pearson correlation coefficients are reported on each subpanel as R. Overall, the lasso models outperformed the other two methods. Modeling results of time translation for SFM to 10 min IGF1 stimulation in MCF7 cells shown as an example.
Fig. 4.
Fig. 4.
The lasso modeling inferred network of interactions in IGF1/Ins signaling pathways. The interactions inferred by the lasso model for SFM to 10 min stimulation are depicted. The purple edges are the top 20 exclusive interactions present in IGF1 model only. The blue edges are the interactions inferred only in the insulin model. The red edges represent the interactions present in both models whereas the magnitudes are stronger in the IGF1 model. The black edges are the ones with higher magnitudes in the Ins model. The yellow nodes are the two phospho-proteins taken as the outputs for the experimental validations.
Fig. 5.
Fig. 5.
The lasso model predictions are experimentally validated to show a differential effect of ACC knock-down on MAPK phosphorylation, and E-Cadherin knock-down on Akt phosphorylation levels. A, The representative Western blot for ACC knock-down in MCF7 cells at 5 and 10 min of stimulation, Scr = control cells, ACC = ACC-knock-down cells. The cells were stimulated with Vhc, IGF1, or Ins. The other experimental replicates are given in supplemental Fig. S14. B, The Western blot showing the change induced by stable E-Cadherin knock-down in T47D cells in response to 5 and 10 min IGF1 and Ins stimulation. Scr = control cells, CDH1 = E-Cadherin-knock-down cells. The cells were stimulated with Vhc, IGF1, or Ins. The efficiency of E-Cadherin knock-down is reported in supplemental Fig. S15. C, The quantification of Western blots given in (A), shown as relative values of pMAPK/MAPK normalized to Vhc treated Scr condition. D, The quantification of the Western blot in (b). Relative values of pAkt to total Akt, normalized by Vhc treated wild-type cells. The bars represent fold-changes from basal condition and results are shown as mean ± S.E. of two or three experimental replicates. The results are compared using unpaired, one-tailed two-sample t test, and p < 0.05 (*), p < 0.01 (**), p < 0.005 (***), nonsignificant (ns). The bars for each time point, from left-to-right: IGF1 stimulated WT cells, Ins stimulated WT cells, IGF1 stimulated knock-down cells, and Ins stimulated knock-down cells.
Fig. 6.
Fig. 6.
Links between IGF1/Ins signaling, ACC and E-Cadherin. A, The canonical IGF1R/InsR signaling pathways comprised of MAPK and Akt cascades. B, The canonical signaling including newly hypothesized interactions based on validated predictions of TT modeling approach. The adherens junction formation sequesters beta-catenin, leading to a repressive effect on Akt activation. When free in cytosol and translocate to nucleus, beta-catenin inhibits PTEN translation, leading to higher activation of Akt. Activated Akt in return phosphorylates GSK3b, and cause a decrease in beta-catenin degradation. Phosphorylated Akt also turns on NF-kB signaling and inhibits E-Cadherin translation. E-Cadherin driven cell-cell contacts also were shown to repress high affinity ligand binding to the receptors, leading to diminished downstream signaling. On the other hand, ACC1 and ACC2 control fatty acid synthesis cascade, but here shown to affect MAPK phosphorylation in response to IGF1 and insulin stimulation.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Casa A. J., Dearth R. K., Litzenburger B. C., Lee A. V., and Cui X. (2008) The type I insulin-like growth factor receptor pathway: a key player in cancer therapeutic resistance. Frontiers Biosci. 13, 3273–3287 - PubMed
    1. Sachdev D., and Yee D. (2001) The IGF system and breast cancer. Endocrine-Related Cancer 8, 197–209 - PubMed
    1. Boone D. N., and Lee A. V. (2012) Targeting the insulin-like growth factor receptor: developing biomarkers from gene expression profiling. Crit. Rev. Oncogenesis 17, 161–173 - PMC - PubMed
    1. Casa A., Litzenburger B., Dearth R., and Lee A. v. (2008) Insulin-like growth factor signaling in normal mammary gland development and breast cancer progression. Breast Cancer: Prognosis, Treatment and Prevention, CRC Press, Boca Raton FL, pp. 303–322, 2nd Ed Informa Healthcare, New York
    1. Liu J. L., and LeRoith D. (1999) Insulin-like growth factor I is essential for postnatal growth in response to growth hormone. Endocrinology 140, 5178–5184 - PubMed

Publication types