FOXO3a and β-catenin co-localization: double trouble in colon cancer?

Metastatic colorectal cancer is a largely incurable disease with a pressing need for targeted therapies. A new study sheds light on a surprising interaction between FOXO3a and β-catenin in metastatic colorectal cancer, suggesting new therapeutic avenues for agents targeting the PI3K-AKT pathway (pages 892–901).

Metastatic colorectal cancer is diagnosed annually in an estimated 240,000 individuals worldwide and has a five year survival rate of just 11% (ref. 1). The phosphoinositide 3-kinase (PI3K)-AKT and WNT–β-catenin signaling pathways both have prominent roles in tumorigenesis and disease progression in colorectal cancer2,3. Forkhead box O3a (FOXO3a) is a transcriptional regulator downstream of PI3K-AKT signaling that is normally regarded as a tumor suppressor, as it promotes cell cycle arrest and apoptosis4. FOXO3a is sequestered in an inactive form in the cytoplasm when phosphorylated by AKT; inhibition of AKT signaling releases FOXO3a to translocate to the nucleus, where it regulates gene expression4 (Fig. 1).

Figure 1: Concomitant localization of FOXO3a and β-catenin in the nucleus mediate progression and metastasis in colorectal cancer upon PI3K or AKT inhibition.

Untreated, primary colorectal cancer cells have heterogeneous levels of β-catenin (β-cat; far left). In active proliferating cells, PI3K-AKT activation results in phosphorylation of FOXO3a, restricting its localization to the cytoplasm. Treatment of colon carcinoma cells with selective PI3K or AKT inhibitors leads to translocation of FOXO3a to the nucleus. In this context, tumor cells with low nuclear β-catenin undergo apoptosis (partially responding carcinoma), whereas those with high nuclear protein levels of both β-catenin and FOXO3a become resistant to induction of apoptosis. In these resistant cancer cells, nuclear FOXO3a and β-catenin act together to regulate a defined set of target genes mediating cell scattering and metastasis. Although the tumor partially responds to treatment, the resistant cell population disseminates and contributes to treatment-induced metastasis (metastatic carcinoma). Concurrent treatment of these cells with AKT or PI3K inhibitors and WNT pathway inhibitors, such as a tankyrase inhibitor, causes enhanced apoptosis and impairs metastasis (responding carcinoma). RTK, receptor tyrosine kinase.

In this issue of Nature Medicine, Tenbaum et al.5 show surprising findings that suggest a previously unappreciated role for nuclear FOXO3a in promoting metastasis in the context of nuclear β-catenin accumulation, a common occurrence in colon cancers. Applying AKT inhibitors to tumor cells with concomitant high nuclear expression of FOXO3a and β-catenin resulted in resistance to apoptosis and promotion of the metastatic program (Fig. 1). The findings have important implications—a new cooperative interaction of these two fundamental pathways in the nucleus in mediating metastasis, along with clinical relevance for patients with colorectal cancer and dual elevation of FOXO3a and β-catenin, as they may respond poorly to PI3K-AKT inhibitors. These findings suggest a diagnostic strategy that may help identify nonresponsive individuals.

A notable strength of the study is the use of a combination of relevant cell line models and tumor samples with associated clinical outcome. By analyzing tumor samples, the authors found that nuclear FOXO3a and β-catenin were frequently concurrently expressed in late-stage cancers and that this was associated with poor prognostic outcome. Using cell line models, Tenbaum et al.5 showed that concurrent expression of ectopic activated FOXO3a together with β-catenin promoted cell scattering in vitro and metastasis in vivo. This seemed to be due to an altered transcriptional program consisting of induction of genes involved in cell adhesion and scattering. Increased β-catenin protein amounts promoted resistance to apoptosis induced by AKT and PI3K inhibitors; moreover, inhibition of AKT signaling promoted scattering and metastasis of xenografted colon tumor cells with elevated β-catenin. Lastly, patient-derived sphere cultures and tumor xenografts showed differential sensitivity to an AKT inhibitor that was inversely correlated to their intrinsic nuclear β-catenin levels5. The authors therefore concluded that it will be important to investigate the subcellular localization of β-catenin and FOXO3a in tumor samples before administering AKT inhibitors to patients with colorectal cancer, as in the context of active Wnt signaling such treatment may induce malignant behavior of tumor cells.

The study adds to previous studies showing that the Wnt and PI3K-AKT pathways interact at multiple levels. For example, AKT phosphorylation of glycogen synthase kinase-3β is essential in regulating β-catenin degradation and blocking Wnt signaling6. Moreover, transcriptional upregulation of the kinase SGK by T cell factor–lymphoid enhancer factor–β-catenin complexes results in SGK-directed phosphorylation and retention of inactive FOXO3a in the cytoplasm7. In these examples, the pathways have antagonistic functions—the observation of direct crosstalk between β-catenin and FOXO3a in the nucleus and a cooperative role in triggering metastasis is therefore an unexpected aspect of the complex interplay between these pathways. It remains to be seen whether this effect is specific for PI3K-AKT signaling and unique to colorectal cancer.

A recent study suggests that elevated β-catenin activation may not be a general resistance factor for other targeted therapies that regulate FOXO3a localization in different tumor types; in particular, endogenous β-catenin is required for BRAF (an oncoprotein mutated in melanoma) inhibitor–induced apoptosis of melanoma cells harboring V600E missense mutations in the gene encoding BRAF. High nuclear β-catenin protein levels synergize with a BRAF inhibitor to induce apoptosis and decrease tumor growth, suggesting high β-catenin amounts may enhance the efficacy of BRAF-targeted therapy. Thus, the role of β-catenin in apoptosis may be context or tissue dependent. In addition, FOXO3a levels are also known to be regulated by kinases other than AKT, notably IKK and ERK4,9. In fact, ERK phosphorylation of FOXO3a, albeit on different sites than those regulated by AKT, drives tumorigenesis by triggering degradation of FOXO3a via a mechanism involving the E3 ubiquitin ligase MDM2 in breast cancer cell lines9. Therefore, blocking the BRAF-MEK-ERK axis with selective inhibitors could conceivably act analogously to AKT inhibition in terms of elevating levels of nuclear FOXO3a protein in colorectal cancer, a hypothesis that could be readily tested using the models described by Tenbaum et al.5.

The findings of Tenbaum et al.5 have clear clinical implications because they suggest that high nuclear FOXO3a and β-catenin amounts may predict resistance to PI3K-AKT–targeted therapies in colorectal cancer. Inhibitors of this pathway are the subject of intensive clinical investigation, and preclinical studies with the PI3K and mammalian target of rapamycin inhibitor GDC-0980 have suggested relatively modest activity in colorectal cancer cell lines compared to breast or prostate lines10. The authors may have provided a clue to this observation, showing that simultaneous accumulation of β-catenin and FOXO3a not only confers poor prognosis but also seems to be predictive of a poor efficacy for AKT inhibitors. Individuals with low nuclear β-catenin protein amounts may then be the most likely to benefit from therapeutic intervention with PI3K-AKT inhibitors. Alternatively, those with high nuclear β-catenin and FOXO3a protein may require concurrent treatment with selective inhibitors of Wnt–β-catenin signaling. Indeed, the authors showed that treatment of resistant colorectal cancer cells with the tankyrase inhibitor XAV-939, which reduces nuclear β-catenin protein levels and transcriptional activity11, resulted in increased PI3K or AKT inhibitor–mediated apoptosis5. However, tankyrase inhibitors have not yet entered clinical development, and dual targeting of these signaling pathways may still be a distant goal.

This study also highlights the importance of understanding interpatient tumor heterogeneity to guide administration of rationally targeted therapies. The lack of cell and tumor models that closely mimic relevant human cancers has long been recognized as a major hurdle in translating basic cancer research to the clinic. By using patient-derived tumors to assess the response to drug treatment in individual tumors, the authors have pushed the limits of this type of analysis by first testing the in vitro sensitivity of patient-derived colorectal cancer sphere cultures and then extending these findings to in vivo tumor explants5. The results suggest the possibility that such sphere cultures could be used immediately after surgical resection to evaluate biomarkers and assess efficacy of candidate drugs with great accuracy in real time to individualize treatment regimens. Alternatively, pharmacodynamic effects on FOXO3a nuclear levels and apoptosis could be directly assessed in pre- and post-treatment tumor biopsies, an approach several of the investigators in the current study have successfully employed with other therapeutics12. Though operational and technical hurdles remain, the approach of Tenbaum et al.5 clearly signals a promising trend in translational research.

The work by Tenbaum et al.5 suggests several additional avenues of scientific inquiry. For instance, many of the observations, although previously undescribed, remain primarily correlational. How is metastasis increased under dual activation? What are the most important effector genes induced by the combination, what is their relative biological contribution to metastasis and are any of them druggable targets? It also remains to be seen whether approved receptor tyrosine kinase inhibitors, including the epidermal growth factor–targeted antibodies cetuximab and panitumumab13, have similar prometastatic effects, as receptor tyrosine kinase inhibition can, in some contexts, inhibit PI3K-AKT signaling.

Interesting clinical challenges lie ahead, as well. One is that colorectal cancer tumors are heterogeneous regarding amounts of elevated nuclear β-catenin protein14, which is a continuous variable that may be challenging to dichotomize. Delineating cutoffs and the most suitable predictive diagnostic assay to define 'high' nuclear β-catenin protein are major challenges. How such an assay will be deployed and prospectively validated in the clinical setting, given that clinical trials are already underway for several inhibitors, is an equally difficult question. One might also wonder how elevated β-catenin may affect efficacy in the background of other candidate predictive biomarkers, such as PIK3CA mutations. Careful clinical trial design and patient stratification with mandatory biomarker assessments will be crucial in taking this story to the next level. The work by Tenbaum et al.5 is a step in the right direction in the quest for more personalized therapies for patients with colorectal cancer.


  1. 1

    Howlader, N. et al. SEER Cancer Statistics Review. National Cancer Institute, Bethesda, MD (2008).

  2. 2

    Samuels, Y. & Velculescu, V.E. Cell Cycle 3, 1221–1224 (2004).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  3. 3

    Zhang, Y., Gan, B., Liu, D. & Paik, J.H. Cancer Biol. Ther. 12, 253–259 (2011).

    CAS  Article  PubMed  Google Scholar 

  4. 4

    Yang, J.Y. & Hung, M.C. Curr. Drug Targets 12, 1284–1290 (2011).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  5. 5

    Tenbaum, S.P. et al. Nat. Med. 18, 892–901 (2012).

    CAS  Article  PubMed  Google Scholar 

  6. 6

    Aberle, H., Bauer, A., Stappert, J., Kispert, A. & Kemler, R. EMBO J. 16, 3797–3804 (1997).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  7. 7

    Dehner, M., Hadjihannas, M., Weiske, J., Huber, O. & Behrens, J. J. Biol. Chem. 283, 19201–19210 (2008).

    CAS  Article  PubMed  Google Scholar 

  8. 8

    Biechele, T.L. et al. Sci. Signal. 5, ra3 (2012).

  9. 9

    Yang, J.Y. et al. Nat. Cell Biol. 10, 138–148 (2008).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  10. 10

    Wallin, J.J. et al. Mol. Cancer Ther. 10, 2426–2436 (2011).

    CAS  Article  PubMed  Google Scholar 

  11. 11

    Huang, S.M. et al. Nature 461, 614–620 (2009).

    CAS  Article  Google Scholar 

  12. 12

    Rojo, F., et al. J. Clin. Oncol. 24, 4309–4316 (2006).

    CAS  Article  PubMed  Google Scholar 

  13. 13

    Ciardiello, F. & Tortora, G. N. Engl. J. Med. 358, 1160–1174 (2008).

    CAS  Article  PubMed  Google Scholar 

  14. 14

    Brabletz, T. et al. Proc. Natl. Acad. Sci. USA 98, 10356–10361 (2001).

    CAS  Article  PubMed  Google Scholar 

Download references

Author information



Corresponding authors

Correspondence to Yibing Yan or Mark R Lackner.

Ethics declarations

Competing interests

The authors declare no competing financial interests.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Yan, Y., Lackner, M. FOXO3a and β-catenin co-localization: double trouble in colon cancer?. Nat Med 18, 854–856 (2012).

Download citation

Further reading


Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing