Fig 1.
Prolonged passage of HPV16 E6/7 immortalised keratinocytes generates an invasive epithelium.
A) H+E stained sections of organotypic cultures containing early and late passage E6/7-HFKs cultured with pro-invasive pRb-depleted fibroblasts. Late-passage but not early passage E6/7-HFKs invade into the collagen matrix. B) Medium transfer from late-passage E6/7-HFKs (Late-CM) can promote invasion in early pass E6/7-HFKs unlike medium from normal HFKs (HFK-CM) and early-passage E6/7-HFKs (Early-CM). The frequency of invasion is shown in C), data represents the average of three independent experiments, error bars represent standard error of the means (SEM). CM = conditioned medium. D) Growth factor array analysis of conditioned medium, which is quantified in E) demonstrates late-passage E6/7-HFKs secrete less IGFBP2 and GM-CSF. F and G) GM-CSF and IGFBP2 mRNA transcript levels are reduced in late-passage E6/7-HFKs compared to non-invasive early pass E6/7-HFKs. n = 3, error bars represent SEM. H) Medium transfer from non-invasive early E6/7-HFKs blocks invasion of late-passage E6/7-HFK cultures as quantified in I). n = 3, error bars represent SEM. Scale bars represent 100 μM.
Fig 2.
IGFBP2 is transcriptionally repressed in late-passage E6/7-HFKs and HPV positive cervical cancer cell lines.
A) Continued expression of E6/7 in HFKs leads to dramatic down-regulation of IGFBP2 mRNA levels, n = 4, error bars represent SEM, this is also observed at a protein level (B). C) Real-time PCR analysis of the IGFBP family of proteins demonstrated a specific down regulation of IGFBP2 expression. n = 3, error bars represent SEM. Note IGFBP1 is not expressed by HFKs but was readily detected in human fibroblasts (S4A Fig). D) Gene rank analysis of IGFBP2 expression in cervical cells lines identified that IGFBP2 expression is low in HPV positive cell lines in comparison to cervical HPV negative cell lines (C33A and HT-3). E) The results in D were confirmed by western blot (E) and real-time PCR analysis (F) for the C33a, Caski and Hela cervical cell lines. n = 3, error bars represent SEM.
Fig 3.
IGFBP2 is a regulator of the invasion process.
A) Recombinant IGFBP2 was added to organotypic cultures at the indicated concentrations over a 14 day period. Addition of IGFBP2 inhibited the invasive behaviour of late-passage E6/7-HFKs and the invasive frequencies are quantified in B). n = 3, error bars represent SEM. C) IGFBP2 levels were depleted from early-passage E6/7-HFKs as demonstrated by western blot (C) and real-time analysis (D). n = 3, error bars represent SEM. E) Depletion of IGFBP2 from early-passage E6/7-HFKs resulted in enhanced invasive potential as quantified in F). However this was not observed when IGFBP2 was depleted in primary HFKs (G, also E and F). n = 3, error bars represent SEM. Scale bars represent 100 μM.
Fig 4.
IGFBP2 expression is repressed through enhanced HDAC function.
A) Addition of 1 μM Trichostatin A (TSA) or 5 mM sodium butyrate (SB) restores expression of IGFBP2 in late passage E6/7-HFKs to levels similar to primary keratinocytes, as demonstrated by Western blot and real-time PCR (B). n = 3, error bars represent SEM. C) The expression levels of histone deacetylases were assessed by real-time PCR in control primary keratinocytes (pBabe) early and late passage E6/7-HFKs. n = 3, representative experiment shown, error bars represent standard deviation (SD). D) Addition of HDAC inhibitors to organotypic cultures significantly reduced the invasive frequency of the epithelium, as quantified in (E). n = 3, error bars represent SEM.
Fig 5.
HDAC3 is a critical regulator of IGFBP2 expression.
A) Using selective (HDAC6i, HDAC1/2i and HDAC3i) HDAC inhibitors and a class I inhibitor (entinostat – entino), we were able to restore IGFBP2 expression in late passage E6/7-HFKs both at the protein and transcriptional level (B). n = 3, representative experiment shown, error bars represent SD. C) Following 48 hours treatment with siRNA targeting HDAC1, 2 or 3, western blot analysis confirmed the specific depletion of individual HDACs siRNA and restoration of IGFBP2 following HDAC3 depletion. D) This was also confirmed at the transcriptional level n = 3, error bars represent SEM. E) Encode histone 3 lysine 9 acetylation (H3K9Ac) CHIP-seq from normal human epithelial keratinocytes (NHEK) at the IGFBP2 locus. Blocks represent primer locations to assess the levels of H3K9Ac binding at these sites. F) H3K9Ac CHIP-qPCR at the three sites identified in E, demonstrates the loss of the modification in late passage E6/7-HFKs which lack IGFBP2 expression. n = 3, error bars represent SEM G) HDAC3 CHIP-qPCR was conducted at locations throughout the IGFBP2 locus, predicted from previous HDAC3 CHIP-seq experiments in human and mouse derived cells. HDAC3 binding was enriched at the transcriptional start site (TSS) and was enhanced in late passage E6/7-HFKs. n = 3, error bars represent SEM. H) Western blot of NcoR1 and NcoR2 in HFKs, early and late passage E6/7-HFKs. I) The co-repressors NcoR1 and NcoR2 are elevated in late passage E6/7 cells at the protein and transcriptional level (I). n = 3, error bars represent SEM J) CHIP-qPCR analysis also demonstrated enhanced binding of both NcoR1 and NcoR2 at the TSS. Average of two experiments shown, error bars represent SD.
Fig 6.
IGFBP2 blocks pro-invasive IGF1R signalling.
A) Late-passage E6/7-HFKs were pre-treated with 10ng/mL IGFBP2 for 1 hour prior to 10 minutes of 10ng/ml of IGF1/2 treatment, as described in the Methods section. IGF1/2 induced activation of AKT and ERK pathways, which was inhibited by IGFBP2. B) Similar treatment of primary HFKs and early-passage E6/7-HFKs showed that these cells do not activate AKT in response to IGF1/2 treatment. C) The expression of the IGF and insulin receptors (IGF1R/2R and INSR-A/B, respectively) were assessed by real-time PCR and demonstrated significant increases in IGF1R and IGF2R mRNA. n = 3, error bars represent SEM. This was also detected at a protein level for IGF1R in cycling cultures of late passage E6/7-HFKs D). E) IGFBP2 blocks signalling pathways mediated by KGF, including those that result in activation of Ets1 and MMP1. F) The role of IGF1R signalling in regulating invasion was assessed by depletion of IGF1R using siRNA (siIGF1R) and knockdown confirmed by western blotting (F) and real-time PCR (G). n = 3, error bars represent SEM. H) siIGF1R treatment inhibited the invasion process, as quantified in I). Average of three experiments, error bars represent SEM. J) IGF1R knockdown inhibited the KGF-dependent activation of Ets2 and MMP1. Scale bars represent 100 μM.
Fig 7.
Signalling mediated by KGF requires IGF1R.
A) KGF treatment of primary HFKs leads to activation of the AKT pathway in late-passage E6/7-HFKs only. B) siRNA depletion of IGF1R (siIGF1R) prevents the activation of AKT by KGF suggesting that IGF1R forms part of the signalling pathway. C) ADAM17 expression is elevated in late-passage, invasive, E6/7-HFKs, compare to early-passage E6/7-HFKs. n = 3, error bars represent SEM. D) Knockdown of ADAM17. n = 4, representative experiment shown, error bars represent SD, confirmed by western blot and real-time PCR (E) resulted in inhibition of AKT activation by KGF (F), as quantified in (G). H) Following 30 mins treatment of KGF, IGF1 was shed from late-passage E6/7-HFKs in an ADAM17 dependent manner. n = 4, error bars represent SEM.
Fig 8.
IGFBP2 is frequently down regulated in HPV16 infected CIN3 lesions.
A) Immunohistochemical (IHC) and immunofluorescence staining of IGFBP2 localises the protein in the cytoplasm of epithelial cells of the cervix with regions of low IGFBP2 observed. In order to assess whether these were in HPV infected regions the same sample was co-stained with IGFBP2 and p16, the latter a surrogate marker of HPV infection. IGFBP2 was found to be commonly down-regulated in HPV infected regions (CIN3). HPV infected epithelium is indicated with a red arrow, and neighbouring normal tissue with a white arrow. Epi = epithelium; str = stroma. B) To further address the role of IGFBP2 in the progression of cervical cancer, 40 CIN1, 39 CIN3 lesions and 50 invasive carcinomas were assessed for IGFBP2 staining intensity and scored as-, +, ++, +++, for normal and p16 positive regions. If IGFBP2 scores decreased by a factor of 2 or more between p16 positive compared to normal regions then samples were identified as ‘reduced IGFBP2’. Shown are representative images of CIN1 where IGFBP2 expression was un-altered and invasive disease where IGFBP2 is reduced in p16 positive regions compared to neighbouring normal epithelium *indicates non-specific staining of red blood cells. Arrows are as stated above. C) Quantification across all samples. D) We had data for 13 patients who had low levels of IGFBP2 in the original CIN1 biopsy. 10 patients progressed, whilst 1 patient regressed and 2 had persistent CIN1 lesions. E) Analysis of publically available microarray datasets for IGFBP2 in oro-pharyngeal cancers demonstrated that in tumours where p16 is readily detected by IHC, IGFBP2 is reduced compared to tumours with no p16 staining. p16 expression in these dataset was also assessed by mRNA expression and was used as a surrogate of HPV infection.
Fig 9.
Proposed mechanism of IGFBP2 function.
In the presence of IGFBP2, IGF1/2 cannot be released from the cell surface and therefore cannot activate the IGF1 receptor. However, when IGFBP2 is lost, KGF activation of ADAM17 leads to cleavage of the unprotected IGF1 leading to IGF1R activation and subsequent AKT activation, which we have previously demonstrated leads to expression of MMP1 [22].