Expression analysis of candidate breast tumour suppressor genes on chromosome 16q

Material and RNA isolation

From a series of fresh frozen breast tumours, tested for LOH on the long arm of chromosome 16 as described previously [3], we selected a representative panel of tumours with different LOH status at chromosome 16q and with at least 50% tumour cells on examination of a hematoxylin and eosin-stained section by a pathologist. The series consists of 189 patients operated on between 1986 and 1993 in three Dutch hospitals [3]. Patient material was obtained on approval of local medical ethics committees. RNA from cell lines and snap-frozen tumours was isolated using TRIZOL (Invitrogen, Breda, The Netherlands) and subsequently purified with Qiagen RNeasy columns combined with the RNase-free DNase kit (Qiagen Sciences, Germantown, MD, USA) according to the manufacturer's instructions. cDNA was made using AMV reverse transcriptase (Roche Diagnostics, Basel, Switzerland).

Quantitative reverse transcriptase PCR

qPCR primers were designed in Primer Express (Applied Biosystems Applied Biosystems, Foster City, CA, USA) and primers for fragment analysis were designed using the primer3 program [28]. qPCR reactions were performed on an iCycler (Biorad, Hercules, CA, USA) using the SybrGreen qPCR core-kit (Eurogentec, Seraing, Belgium). Cycle conditions were: 10 minutes at 94°C followed by 40 cycles of 10 s at 94°C and 1 minute at 60°C. Cycle threshold extraction was performed using the iCycler IQ software (version 3, Biorad).

The primers used were: CPSF6, 5'-AAGATTGCCTTCATGGAATTGAG-3', 5'-TCGTGATCTACTATGGTCCCTCTCT-3'; CYPA, 5'-TCATCTGCACTGCCAAGACTG-3', 5'-CATGCCTTCTTTCACTTTGCC-3'; FANCA, 5'-TTAATACCTCGGTGCCCGAA-3', 5'-AGTCCCCACGATCAGCCA-3'; LRRC29, 5'-CCTGCACGCCTGCCC-3', 5'-TGCAGTCAGCTCATAGAGCAGACACTGGA-3'; HNRPM, 5'-GAGGCCATGCTCCTGGG-3', 5'-TTTAGCATCTTCCATGTGAAATCG-3'; CBFA2T3, 5'-ACATCTGGAGGAAGGCTGAAGAG-3', 5'-GCTCCATCTTGGCACGCT-3'; PFKP, 5'-ACCCCTTCGGCATTCGAC-3', 5'-AGCAAGGCGATGACTGCC-3'; TERF2IP, 5'-AAGCTCAAGCGGAAGGCG-3', 5'-TCTGGAGTTCTCTTATTCTGTGGTTC-3'; TAF1C, 5'-GACCGCACCGGAGTGAAG-3', 5'-AACGAAAAAGCAACAGACCACA-3'; and TERF2, 5'-GGTACGGGGACTTCAGACAG-3', 5'-CGCGACAGACACTGCATAAC-3'. For all PCRs, a standard curve was generated using five 1:5 dilutions of pooled cDNA from normal breast epithelial cell lines (MCF10A, MCF10F, MCF12). Relative concentrations of mRNA for each gene were calculated from the standard curve. After qPCR, dissociation curves were made to check the quality of the reaction. Reactions with more than one peak in the dissociation curve were discarded. Using the GeNorm applet, stably expressed control genes for normalization were selected; the three most stable expressed genes were used to calculate normalization factors for each cell line or tumour cDNA [29]. For normalization, the highest expression values for each gene were set to 1 and subsequently divided by the normalization factor generated by the GeNorm applet.

Alternatively, we calculated the rlative fold variability index (RFVI) for each gene, as described [30]. The baseline RFVIs were calculated for the control genes CPSF6, HNRPM, TBP and CYPA. These ranged between 11 and 42, reflecting experimental or population variations.

Fragment analysis

Standard fragment analysis on genomic DNA was performed using fluorescent-labelled primers (Isogen Life Science, IJsselstein, The Netherlands) on an ABI377 and analysed using GeneScan and Genotyper software (Applied Biosystems). Sequencing was performed at the sequence core of the Leiden Genome Technology Center. Four genes were screened to detect small insertions and deletions. Products ranging from 200 to 500 base pairs were generated to screen the exons of TERF2IP, TERF2, LRRC29 and FBXL8 (4, 10, 4 and 6 products, respectively). Fluorescent fragment analysis detects most if not all insertions and deletions due to the size differences and detection of mutations varies from 60% to 88% [31, 32].

Quantitative reverse transcriptase PCR expression analysis

RNA expression of the candidate TSGs at the frequently deleted long arm of chromosome 16 (FANCA, TERF2IP, TERF2, FBXL8, LRRC29 and CBFA2T3 [9]) was studied using qPCR. Expression of the genes in normal breast tissue is a prerequisite for a function as a TSG. We therefore first tested the expression by qPCR in three normal breast cell lines and two normal breast tissues. FANCA, TERF2IP, TERF2, LRRC29 and CBFA2T3 were expressed in breast tissue. For FBXL8, however, we failed to show any expression in normal breast, breast cell lines or breast tumours using different combinations of RT-PCR primers or northern blot analysis (not shown). We therefore excluded FBXL8 from any further analysis and as a candidate TSG. Expression analysis of the genes was studied using real-time qPCR. Relative expression levels of the genes for each sample were calculated from a standard curve from the pooled normal breast cell lines, MCF10A, MCF10F and MCF12A.

Control genes for normalization

For accurate normalization of qPCR data, multiple stably expressed control genes are required [29] because expression variations in a single control gene could have significant impact on the relative expression levels of genes under study. Several control genes are widely used for normalization, such as HPRT (encoding hypoxanthine phosphoribosyl transferase 1), GAPDH (encoding glyceraldehyde-3-phosphate dehydrogenase), TBP (encoding TATA box binding protein) and PBGD (encoding porphobilinogen deaminase or hydroxymethylbilane synthase) [29]. The use of CYPA (encoding cyclophilin A or peptidylprolyl isomerase A) has also been reported for normalization of breast cancer cell lines using qPCR [30].

For each qPCR experiment, the optimal set of controls needs to be tested and the most used controls are not necessarily the best controls. To identify additional control genes specifically suitable for breast cancer cell lines and tumours, we selected possible control genes from gene-expression microarrays hybridised with breast cancer cell lines (Lombaerts,M., van Wezel,T., Philippo,K., Dierssen,J.W.F., Zimmerman,R.M.E., Oosting,J., van Eijk,R., Eilers,P.H., Van De Water,B., Cornelisse,C.J., and Cleton-Jansen,A.M.manuscript submitted). From these cDNA arrays, the most stably expressed possible control genes with the least expression variation in multiple breast cancer cell lines were selected. These were CPSF6 (encoding cleavage and polyadenylation specific factor 6), HNRPM (encoding heterogeneous nuclear ribonucleoprotein M), PFKP (encoding the phosphofructokinase PFKP) and TAF1C (encoding TBP-associated factor, RNA polymerase IC). CPSF6, HNRPM, PFKP and TAF1C, together with HPRT, GAPDH, TBP, PBGD and CYPA, were compared for expression stability in the panel of breast cancer cell lines. Four colon tumour cell lines LS411, LS180, SW480 and SW837 were also included. The most stable control genes for normalization of the relative concentrations of mRNA were identified using the GeNorm software [29]. In the panel of breast cancer cell lines, the most stably expressed genes were TBP, CYPA and CPSF6 (Fig. 1), and in the colon cancer cell lines, CPSF6, HNRPM and TBP appeared the most stably expressed genes (data not shown). For the breast tumours, we tested only the genes HPRT, CYPA, TBP, CPSF6 and HNRPM as the availability of tumour RNA is limited. GeNorm calculates the stability factor 'M' of all control genes by comparing the variation in expression for all genes. A low M-value represents low variation in expression. Table 2 lists the stability factors for all control genes tested. CPSF6, CYPA and TBP were the most stable and were used in subsequent experiments. The commonly used control genes HPRT and GAPDH are much more variable and less suitable as control genes in breast tumours.

The identification of control genes for the normalization of qPCR experiments proved useful as both CPSF6 and HNPRM were very stably expressed in both breast cancer cell lines and tumours. In general, the control genes were more stable in the tumours than in the cell lines (for CPSF6, M = 0.58 for the cell lines and M = 0.38 in tumours). This is possibly due to multiple cell types that are present in a tumour (tumour and stromal cells and infiltrating lymphocytes) whereas cell lines usually are monocultures.

Expression study in cell lines and tumours

We used a panel of three normal breast cell lines (MCF10A, MCF10F and MCF12A), and 27 breast cancer cell lines (Table 1), 1 with LOH q22, 3 with LOH q24, 9 with retention of 16q and 14 with loss of 16q [25, 27]. Furthermore, we tested a panel of two normal breast samples and 40 breast tumours with loss of (part of) the q-arm or with retention of 16q (Table 1). Expression values were normalized by a factor calculated from the three most stable control genes for the cell lines and tumours, (Table 2). Additionally, we calculated the RFVI for each gene between the samples with the highest and lowest expression, using the base-line expression value of each gene in normal breast tissue or cell line, as described previously [30]. The baseline RFVI for the control genes CPSF6, HNRPM, TBP and CYPA ranged between 11 and 42, reflecting experimental or population variations.

As FBXL8 expression could not be detected, this gene was not included in further analysis. LRRC29 was tested first in the cell lines (Fig. 1a). This showed no difference between cell lines with and without LOH, and also the RFVI index for LRRC29 was below the baseline (Fig. 2). These data exclude LRRC29 as a candidate gene and LRRC29 expression was, therefore, not studied in the tumours.

Both TERF2IP and TERF2 show somewhat reduced levels of expression in cell lines with LOH of chromosome arm 16q when compared with those with 16q retention (Fig. 1a); however, these differences are not significant. Both genes were subsequently tested in the tumour panel; again, slightly reduced expression was detected in samples with LOH (Fig 1b). However, the expression was also reduced in tumours with loss of 16q24 alone whereas TERF2IP and TERF2 are located on 16q22. The expression appears to be lower in all tumours compared to normal tissue, regardless of LOH or even grade (data not shown). Also, RFVI levels for TERF2IP and TERF2 are 38 and 41, respectively, which are in the same range as the housekeeping control genes (i.e. below the baseline level). Based on their gene expression variation, therefore, neither genes are likely to be candidate TSGs.

CBFA2T3 encodes a translocation partner of AML1 in myeloid leukaemia and a transcription repressor. Based on its location on chromosomal band 16q24.3, high variation in expression in breast cancer cell lines, loss of protein expression in breast tumours and in vitro growth inhibition of breast cancer cell lines, CBFA2T3 has been proposed as a candidate TSG [9, 30]. We therefore tested this gene in our panel of cell lines and tumours to confirm this finding. We found a large expression variation for CBFA2T3 in cell lines, resulting in a high RFVI value of 122 (Fig. 2), which, however, was not associated with LOH at 16q (Fig. 1a). In primary tumours, there was a tendency for higher CBFA2T3 expression in tumours with LOH of 16q24 (Fig. 1b), where this gene is located, and the RFVI value was below the baseline. These data do not support the candidacy of CBFA2T3 as a TSG on 16q.

FANCA is involved in DNA repair and located at 16q24.3. Its expression in tumours is higher than in normal tissue and, remarkably, expression increases with tumour grade (Fig. 3). There is no association between LOH at 16q and tumour grade, but there is a difference in the mechanism of LOH when comparing low- and high-grade breast cancers. Whereas low-grade tumours show preferential physical loss of 16q, high-grade tumours show mitotic recombination [33]. This could indicate that FANCA is involved only in well differentiated breast cancer. In cell lines, FANCA levels are slightly reduced in tumours with LOH. Interestingly, in tumours with loss of the complete q-arm, FANCA expression is lower than in those with retention of 16q. In those with loss of q21-ter, however, expression is higher than in tumours with retention, suggesting again that the mechanism of 16q LOH may be associated with the targeted TSG.

Fragment analysis

TERF2IP, TERF2, FBXL8 and LRRC29 were screened using fragment analysis for (small) genomic deletions or insertions in their exons in 21 breast cancer cell lines and 32 breast tumours. CBFA2T3 and FANCA were previously screened for mutations, without identifying any inactivating or somatic mutations [5, 9]. Although TSGs are, in many cases, inactivated through deletions and insertions, no mutations were found.