Article Text

Download PDFPDF

Cyclin E low molecular weight isoforms occur commonly in early-onset gastric cancer and independently predict survival
  1. A N A Milne1,2,
  2. R Carvalho3,
  3. M Jansen1,
  4. E K Kranenbarg4,
  5. C J H van de Velde4,
  6. F M Morsink1,
  7. A R Musler2,
  8. M A J Weterman5,
  9. G J A Offerhaus1,2
  1. 1
    Department of Pathology, University Medical Centre, Utrecht, The Netherlands
  2. 2
    Department of Pathology, Academic Medical Center, Amsterdam, The Netherlands
  3. 3
    MRC Holland, Amsterdam, The Netherlands
  4. 4
    Department of Surgery, Leiden University Medical Centre, Leiden, The Netherlands
  5. 5
    Department of Neurogenetics, Academic Medical Centre, Amsterdam, The Netherlands
  1. Dr Anya N A Milne, Pathology Dept H04.2.25, University Medical Centre Utrecht, Postbus 85500, 3508GA, Utrecht, The Netherlands; a.n.a.milne{at}


Background: Post-translational cleavage of full-length cyclin E from the N-terminus can produce low molecular weight (LMW) isoforms of cyclin E containing the C-terminus only.

Aim: To assess their presence in early-onset gastric cancer (EOGC), stump cancers and conventional gastric cancers and ascertain how they influence survival in EOGC.

Methods: The expression of full-length and LMW isoforms of cyclin E in 330 gastric cancers, including early-onset gastric cancer (EOGC), stump cancer and conventional gastric cancer (>45 years old) was compared using antibodies targeted to the N- and C-terminals.

Results: LMW isoforms were found in 35% of EOGCs, compared to 8% of conventional gastric cancers and 4% of stump cancers; their presence was visualised in cell lines using western blot analysis. In addition, C-terminal staining was a positive predictor of survival in EOGC. In contrast, no correlation with survival was found with the N-terminal antibody which detects only full-length cyclin E.

Conclusion: EOGCs have a unique molecular phenotype and LMW isoforms of cyclin E may independently influence survival in EOGC.

View Full Text

Statistics from

Gastric cancer is the second most common cause of cancer-related death in the world.1 Approximately 5–16% of gastric cancer occurs in patients ⩽45 years old.2 3 The clinicopathological and molecular features are believed to be different between gastric cancer occurring in young and old (>45 years of age) patients.46 There are two main histological types of gastric cancer, diffuse and intestinal, as described by Laurén,7 who also noted that diffuse cancers were more likely to occur in younger patients as confirmed in further reports.810 Diffuse histology and young age have previously been considered to have a poor prognosis, but this subject has been widely debated.2 3 11 12 It is known that regardless of histology or age, patients with small and early cancer lesions who undergo surgical resection have a better chance of survival. However, most patients are diagnosed with advanced stage disease and the 5-year survival rate is generally less than 10%.13

Gastric carcinogenesis occurs through a combination of environmental and genetic factors although the latter may be more important in young patients.14 Chronic inflammation caused by Helicobacter pylori infection is believed to be a critical environmental factor in gastric carcinogenesis.1517 However, as the pathogenesis related to H pylori exposure is suspected to take decades, it is more likely to account for cases of gastric cancer in older patients.16

Deregulation of the cell cycle is known to be a critical event in the onset of tumourigenesis. Cyclin E is a G1 type cyclin, which forms complexes with cyclin dependent kinase 2 (CDK2) and subsequently phosphorylates the retinoblastoma protein (Rb), an important tumour suppressor, thereby facilitating S phase entry through the release of E2F transcription factors from Rb. In normal cells, cyclin E accumulates at the G1/S boundary and is degraded as the cell passes through the S phase. This periodicity is controlled by cell-cycle dependent transcription and post-translational control by ubiquitin-dependent proteolysis (i.e. proteasome degradation).18 Cyclin E has also been described as having many different functions such as centrosome duplication,19 20 histone biosynthesis, transcriptional regulation and pre-mRNA splicing.21 However, as illustrated by several mouse models, where the function of cyclin E was not found to be critical,2224 the importance of each of these is not yet fully clear. Cyclin E is amplified and overexpressed in many different tumour types, including gastric cancer.25 26 In breast cancer low molecular weight (LMW) isoforms of cyclin E as well as the full length form have been reported to act as predictors of poor survival. These isoforms, described to be tumour specific, are produced at the protein level through unique N-terminal post-translational proteolytic mechanisms,27 28 and it has been found that in breast cancer, they are not subject to cell cycle regulation but instead are constitutively active.2931

Accurate predictors of outcome in cancer enable the administration of appropriate treatment for various patient groups and may ultimately help prolong survival or increase quality of life for these patients. Thus whether cyclin E can help predict survival in gastric cancer is of great interest. LMW isoforms of cyclin E are not detected by all available cyclin E antibodies as an N-terminal directed antibody only detects the full length form whereas a C-terminal antibody will detect both full length and LMW isoforms of cyclin E.32 33 The presence of LMW isoforms of cyclin E has not yet been investigated in gastric cancer. In this study we used N- and C-terminal directed antibodies to assess the presence of full length and LMW isoforms of cyclin E in 330 different gastric cancers, including 184 early-onset gastric cancers (EOGCs), and we assessed the impact on survival for a subset of these EOGCs where a 12 year follow-up was available. Finally, we visualised these LMW isoforms of cyclin E in gastric cancer, by western blot of gastric cell lines and confirmed that they are not recognised by the N-terminal antibody. Our findings highlight molecular differences between EOGC and conventional gastric cancer with respect to LMW isoforms of cyclin E and our survival statistics present a role for cyclin E as a positive prognostic indicator in EOGC.


Patients/study groups

This research was carried out in accordance with the ethical guidelines of the research review committee of the Academic Medical Centre, Amsterdam. Twenty-eight stump cancers from the Amsterdam post-gastrectomy cohort34 and 119 conventional gastric cancers from the Academic Medical Centre, Amsterdam (>45 years old) were used. A total of 184 cases of gastric carcinoma in patients under 45 years of age were obtained from institutions throughout The Netherlands through the nationwide search system. Follow-up survival data were available for 44 of these young cases. These 44 cases are recruited from a well characterised large scale study in The Netherlands with long and complete follow-up.35 36 The original study group consisted of 1078 patients, of whom only 44 were below the age of 45. Only archival paraffin embedded tissue and no frozen material was available from these 44 patients. For the normal control in western blot analysis, fresh frozen histologically normal non-cancerous gastric mucosa, obtained from a gastrectomy specimen (distant to the site of a perforation) was used.

Construction of tissue microarray

A total of 113 of the EOGCs and 91 of the conventional gastric cancers were assessed immunohistochemically using tissue microarrays (TMAs) constructed from formalin-fixed and paraffin-embedded archival specimens as described previously.37 Three core biopsy specimens (0.6 mm cylinders) were taken from histologically representative regions (including heterogeneous areas) of paraffin-embedded gastric carcinomas and arranged in a new recipient paraffin block (tissue array block) using a trephine apparatus (Beecher Instruments, Silver Spring, Maryland, USA). Normal gastric mucosa from each case was also included where available. Cores were placed in three separate subdivisions together with insertion of liver, lymph node and kidney cores to assist analysis. Unassessable cores were excluded from the analysis. The highest score of the three cores was used for the overall score for a case.


Paraffin-embedded material: immunohistochemistry was performed using 4 µm sections which were deparaffinised, blocked for endogenous peroxidase activity by immersion in 0.3% H2O2 for 20 minutes and heat treated at 100°C in Tris/EDTA (pH 9) for 10 minutes. Non-specific binding was blocked using 5% normal goat serum for 10 minutes followed by incubation for one hour with the primary antibody. The following antibodies were used: HE12, sc-247, targeting C-terminal38 (Santa Cruz Biotechnology, Heidelberg, Germany) 1:3200, polyclonal N-terminal antibody H-145, sc-20684 (Santa-Cruz) 1:500. The Powervision Plus poly-HRP detection system (ImmunoVision Technologies, Daly City, California, USA) in combination with 3,3-diaminobenzadine was used to visualise the antibody binding sites. Sections were counterstained with haematoxylin.

Frozen material: 4 µm sections were cut at −20°C and fixed with acetone. Non-specific binding was blocked using 5% normal goat serum for 10 minutes followed by incubation for one hour with the primary antibody, as above. Endogenous peroxidase activity was blocked using 0.1% natriumazide and 0.3% H2O2 in phosphate buffered saline (PBS) for 9 minutes. Peroxidase activity was then detected using 3,3 amino-ethyl carbazole (Sigma) in dimethylformamide and sections were counterstained with haematoxylin.

The negative controls for the immunohistochemistry were carried out when optimising the antibody. Here we performed immunohistochemistry using a variety of antibody dilutions and buffers on both normal gastric mucosa and gastric mucosa with cancer to optimise the antibody and attain minimal background. During this procedure it could be clearly seen that the normal gastric mucosa was negative with occasional nuclear staining (which also confirms that the staining has been successful), whereas the gastric cancers often showed stronger and more widespread nuclear staining. In this way the “cleanness” and accuracy of the staining could be assessed.

All tumours were initially scored using four immunohistochemical categories: 0–4%, 5–10%, 11–50% and 51–100% positivity in the nucleus.39 Thereafter, cases in the 0–10% categories were deemed negative/normal and cases in the 11–100% category were deemed positive/overexpressing. Of note, as 204 cases were examined using TMA and the remaining 126 on full sections, 63 cases on the TMA were also examined using conventional full section immunohistochemistry which showed a 98.4% concurrence of scoring between these techniques. All cases used for survival analysis were scored on full tumour sections.

Cell culture and lysates

Cell lines MB-MDA-157 (breast carcinoma known to produce LMW isoforms of cyclin E30), Hs746T and AGS (gastric carcinomas), were grown using RPMI medium supplemented with 10% fetal calf serum, glutamine and antibiotics. To obtain lysates, cells were washed twice with ice-cold PBS and ice-cold modified RIPA buffer (Tris–HCl 50 mM, pH 7.4, 1% NP-40, 0.25% Na-deoxycholate, 150 mM NaCl, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM Na3VO4, 1 mM NaF, 0.1% SDS and 1 μg/ml each of aprotinin, leupeptin and pepstatin) was added. Adherent cells were scraped with a chilled rubber policeman and the lysate gently rocked for 15 minutes at 4°C to lyse the cells. Lysates from fresh frozen tissue were made by cutting 4 μm frozen sections which were immediately incubated in ice-cold RIPA buffer. The lysates were subsequently centrifuged and the pellet discarded. Protein concentrations were measured using Bicinchoninic Acid Solution according to Sigma procedure TPRO-562 (Sigma-Aldrich, Zwijndrecht, The Netherlands).

Western blot analysis

A 60 μg aliquot of protein heated for 10 minutes in sample buffer was used per lane. Proteins were resolved by SDS–PAGE, and electrotransferred to nitrocellulose membranes. The nitrocellulose membranes were blocked overnight in 5% milk in Tris buffered saline with 1% Tween (TBST) and incubated with the primary antibody in 5% milk/TBST (HE-12 at 1:10, H-145 at 1:100) at 4°C for 3 hours. After washing 4×5 minutes at room temperature a secondary antibody coupled to HRP (Dako) was added and incubated for 1 hour. Detection of proteins was performed using ECL Western Blotting Detection Reagents (Amersham Biosciences, Buckinghamshire, UK).40


The SPSS V.11.5 software package was used for statistical analysis including χ2 tests, cumulative survival curves with the univariate Kaplan–Meier method and the stepwise Cox regression approach (proportional hazard model) to identify prognostic significance.41 All clinical variables were included in this multivariate analysis. A χ2 test was applied to the groups of gastric cancer to determine whether the differences found between antibodies were statistically significant and a binary logistic regression model was used to adjust for diffuse histology.


Overexpression of cyclin E

The prevalence of cyclin E overexpression was assessed in 184 early-onset gastric cancers, 118 conventional gastric cancers and 28 stump cancers using immunohistochemistry. This was carried out using an antibody directed towards the N-terminus (recognising full length cyclin E only), and an antibody directed towards the C-terminus (recognising both full length and LMW isoforms of cyclin E). Thus cases with C-terminal staining but no N-terminal staining contain only the LMW isoforms of cyclin E.

Table 1 summarises immunohistochemistry results. Overexpression of full-length cyclin E was seen in 26% (48/184) of EOGC cases. This contrasted with conventional gastric carcinomas which showed overexpression of full-length cyclin E in 51% (61/118). Strikingly, when detection of LMW isoforms was included by staining with a C-terminal directed antibody, the overexpression of cyclin E in EOGC was 2.3-fold higher (113/184); this difference was statistically significant (p<0.001) using a χ2 test. This increase in staining due to presence of LMW isoforms was not seen in the conventional gastric cancers, where only an additional 9 positive cases were seen with C-terminal staining, i.e. isoforms were present in 35% of EOGCs compared to only 8% of conventional gastric cancers. Figure 1 shows a typical immunohistochemical staining with N- and C-terminal antibodies in EOGC.

Figure 1 The panel on the left shows immunohistochemical staining (×100) with the N-terminal antibody, negative (A) and positive (C) in tumour tissue and normal gastric mucosa (E). The right panel shows staining with the C-terminal antibody, negative (B) and positive (D) in tumour tissue and in normal gastric mucosa (F). Of note (A) and (D) show staining of the same tumour with different antibodies.
Table 1 Results of immunohistochemistry with N- and C-terminal cyclin E antibodies

A larger proportion of the early-onset gastric cancers had a diffuse histology than in the conventional group, which consisted mainly of intestinal cancers, as can be seen in table 2. Despite this epidemiological overrepresentation of diffuse cancer in young patients, the presence of isoforms of cyclin E (as detected by C-terminal staining only) correlated significantly with age, even when adjusted for diffuse histology using a binary logistic regression model (p<0.001, 95% CI 0.095 to 0.382).

Table 2 Classification of tumours according to Laurén7

Low molecular weight isoforms of cyclin E

Western blot analysis was carried out in order to visualise the LMW isoforms of cyclin E in gastric cancer and to confirm that they were not recognised by the N-terminal antibody (fig 2). Of note, the C-terminal antibody has been widely described in the literature as recognising LMW isoforms of cyclin E.2831 33 The breast cell line MD-MBA-157, known to contain isoforms of cyclin E,30 was used as a positive control; normal gastric mucosa from a patient without cancer which was confirmed as negative for cyclin E overexpression on immunohistochemistry, was used as a negative control. As can be seen in fig 2, the N-terminal antibody (upper panel) recognised the full length form of cyclin E only (50 kDa), with the exception of one possible isoform in the AGS cell line at 47 kDa. On the other hand, the C-terminal antibody (lower panel) clearly recognised low molecular weight isoforms of cyclin E in both gastric and breast cell lines, with the predominant protein at 43 kDa and other isoforms seen at 47, 40 and 33 kDa. The normal gastric mucosa showed the presence of full length cyclin E only, using both the N- and C-terminal antibodies.

Figure 2 Western blot for cyclin E using both N-terminal and C-terminal antibodies. 1, normal gastric mucosa; 2, MB-MDA-157 breast cell line; 3, Hs764T gastric cell line; 4, AGS gastric cell line.

Association of cyclin E expression with survival

In order to assess the possible clinical relevance of cyclin E overexpression in gastric cancer, survival analysis was carried out on 44 EOGCs for which follow-up data were available. In this group survival was correlated with eligibility status (curative/non-curative resection) (p<0.001), the presence of residual disease (no residual tumour, positive margins microscopically, or macroscopic/non-curative resection) (p<0.001), UICC 1997 lymph node status (p<0.001), positive versus negative lymph node status (p = 0.006), and UICC 1997 TNM stage (p<0.001), but not with histological type. Using a univariate analysis, cyclin E staining did not correlate with lymph node status, residual disease status, T stage, TNM stage, or histological type. Notably, positive C-terminal staining (therefore including both full length and LMW isoforms of cyclin E) was found to be an independent predictor of survival using a Cox regression stepwise analysis, whereby negative staining had a hazard ratio of 5.63 (p<0.001, 95% CI 2.05 to 15.46) (table 3). This model included the presence of residual disease (no residual tumour, positive margins microscopically, or macroscopic/non-curative resection), positive versus negative lymph node status, T stage and cyclin E staining. It can be seen from table 3 that residual disease (which affected 7 of the 44 patients) was the most important prognostic factor in this group of patients, and interestingly the T-stage did not prove to be an independent prognostic indicator of survival in this group.

Table 3 Results of the Cox regression (proportional hazard model) analysis with and without inclusion of cyclin E

This association with C-terminal cyclin E staining and survival remained statistically significant with a hazard ratio of 5.166 (p = 0.002) when only the cases positive for isoforms were included (as assessed by negative N-terminal staining and positive C-terminal staining), thus demonstrating that the presence of the LMW isoforms alone is an independent predictor of survival in early-onset gastric cancer. Positive cyclin E staining using the N-terminal antibody staining on the other hand was not a predictor of survival (p = 0.71).

Kaplan–Meier analysis showed that the median survival for the cyclin E negative group was 2.14 years, whereas the median survival for the cyclin E positive group was greater than 12 years (still alive at the time of follow-up) as can be seen in fig 3. It is interesting to note that this group of young patients had a particularly good survival rate, even with the inclusion of those patients with residual disease (7/44).

Figure 3 Cox regression analysis results for cyclin E C-terminal staining. The median survival for the negative group was 2.14 years, whereas the median survival for the positive group was greater than 12 years (alive at the time of follow-up). EOGC, early-onset gastric cancer.


The function and mechanisms of the tumor specific LMW isoforms of cyclin E, which are produced by post-translational cleavage of full length cyclin E by calpain28 and elastases,29 are of great interest and it is known that they are not subject to cell cycle regulation.29 30 In this study we find that, LMW isoforms of cyclin E occur more commonly in EOGC than conventional gastric cancers or stump cancers, thus adding additional evidence that EOGC have a different molecular phenotype than conventional gastric cancer.6 42 Furthermore, contrary to findings in breast cancer, these isoforms act as a positive predictor of survival in EOGC.

The clinicopathological implications of cyclin E overexpression have been reported to vary with the type of tumour.33 4345 Our findings suggest that function of LMW isoforms of cyclin E is different from that of the full length form in EOGC as demonstrated by the increase in survival in patients expressing LMW isoforms, as based on preferential N- and C-terminal staining. This unique role of LMW isoforms in gastric cancer is highlighted by the fact that our findings differ from the classical role of cyclin E as an oncogene, as initially established in carcinoma of the colon46 and breast38 47 where it was found to be a negative prognostic factor. In breast cancer deregulation of cyclin E/Cdk2 kinase activity was also shown to induce chromosomal instability and aneuploidy.48 However, it has also been reported that cyclin E is not significantly correlated with polyploidy or aneuploidy, when tumours of similar grade are evaluated, unless aberrations in other genes such as hCDC4 are present.49 Furthermore, novel functions of cyclin E which conflict with previous dogma have recently been reported,50 whereby cyclin E overexpression in medullary breast cancer was shown to lead to decreased invasive potential. Here, cyclin E overexpression was shown to induce differences in gene expression patterns associated with cell adhesion, as well as rearrangements of the actin cytoskeleton, increased adhesive properties, decreased motility, and decreased invasive potential, indicating an overall abrogated mobility, as a result of cyclin E overexpression and providing a possible molecular basis for our survival findings in gastric cancer.

Our findings are in keeping with previous literature that cyclin E positivity correlates with good prognosis in gastric cancer,51 although in some reports, no correlation was found.52 53 To our knowledge this is the first study where LMW isoforms of cyclin E have been described in gastric cancer; the association of these isoforms with patient age may provide an explanation for the discrepancy between previous reports.

Our results lend further support to the growing evidence that the role of cyclin E in carcinogenesis is far more complex than first thought; its role appears to vary depending on the cancer type,51 as shown by its role as a positive prognostic indicator in bladder tumours,45 54 certain lung tumours,55 skin tumours56 as well as gastric cancers.51 This complexity of molecular wiring in carcinogenesis has also been emphasised in recent literature,57 58 with the conclusion that cancer can no longer be viewed purely in terms of a network of oncogenes and tumour suppressor genes. Our findings also add to the increasing literature that EOGCs have a molecular phenotype distinct from that of conventional gastric cancers.6 42 59

Gastric carcinoma continues to be a cause of premature death, despite progress in detection and treatment, and despite advances in our understanding of the molecular basis of cancer. The need to develop efficient and effective cancer-specific drugs is coupled with the importance of accurate prediction of disease outcome for various patient groups some of whom, due to the biology of their disease, will do better than others and may warrant a different treatment protocol. In this study we found a difference in the survival of EOGC patients between those expressing full-length and LMW isoforms of cyclin E, as based on preferential N- and C-terminal staining. In addition, our findings highlight that early-onset gastric cancers have a unique molecular phenotype when compared to conventional gastric cancers.

Take-home messages

  • Low molecular weight isoforms of cyclin E occur commonly in early-onset gastric cancer.

  • C-terminal staining of cyclin E is a positive predictor of survival in early-onset gastric cancer.

  • Early-onset gastric cancers have a unique molecular phenotype, distinct from conventional gastric cancers.


The authors thank Wilfred Meun for assistance with the figures.


View Abstract


  • Funding: The study was funded by the Vanderes Foundation.

  • Competing interests:None declared.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.