SEARCH

SEARCH BY CITATION

Keywords:

  • breast cancer;
  • KIF14;
  • kinesin;
  • prognosis;
  • real-time RT-PCR

Abstract

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Gain of chromosome 1q is a hallmark of breast cancer, and likely reflects oncogene amplification. We previously identified mitotic kinesin KIF14 (kinesin family member 14) as an overexpressed candidate oncogene in the 1q31.3–1q32.1 minimal region of genomic gain in breast cancer cell lines. KIF14 also showed high expression in other cancers, notably an association with survival in lung tumors. We now report KIF14 expression in 99 primary breast tumors and 10 normal breast controls. Measured by real-time RT-PCR, KIF14 was overexpressed 10-fold on average in tumors relative to normals (t test p = 0.000054); expression increased with grade (ANOVA p = 0.000006). Infiltrating ductal carcinomas had higher KIF14 levels than lobular (p = 0.017), and estrogen receptor (ER) negative tumors had higher KIF14 levels than ER positive tumors (t test p = 0.030). KIF14 expression correlated positively with Ki-67 mRNA level (Spearman r = 0.692, p = 0.000001), fraction of positive nodes (r = 0.227, p = 0.024) and percent invasive cells (r = 0.360, p = 0.0002), and negatively with percent fatty stroma (r = −0.258, p = 0.010) and percent normal epithelium (r = −0.291, p = 0.003). KIF14 expression is thus tumor-specific and increased in more aggressive tumors. Indeed, KIF14 expression predicted overall survival (univariate Cox p = 0.010), with an odds ratio of 3.60 (1.37–9.48), in 50 tumors with available outcome data. KIF14 overexpression also predicted decreased disease-free survival (log-rank p = 0.049). These findings are the first evidence of association between expression of a mitotic kinesin and prognostic variables in breast cancer. © 2006 Wiley-Liss, Inc.

Recurrent chromosomal gains and losses are a feature of most cancers, and are presumed to reflect underlying oncogene amplification and tumor suppressor gene loss, respectively. In breast cancer, gain of 1q is the second-most common genomic gain (after 8q). Based on cumulative data in the Progenetix database of cytogenetic changes in cancer,1 the minimal common region of genomic gain encompasses bands 1q31–1q32, gained or amplified in 44% of the 631 breast tumors in the database.1 Gain of the long arm of chromosome 1 is one of the most frequent genomic aberrations in many tumor types, including retinoblastoma, strongly suggesting the presence of oncogene(s) in this region.

Using quantitative multiplex PCR of genetic markers across the 1q31–1q32 region of chromosomal gain, we previously scanned DNA from 55 retinoblastoma and 12 breast cancer cell lines, and identified a 3.06 Mbp minimal region of gain.2 Reverse transcription (RT)-PCR expression analyses of all genes in this region showed that only KIF14 (kinesin family member 14), a poorly characterized kinesin, was overexpressed 31- to 92-fold in 4 breast cancer cell lines compared with healthy breast tissue. Robust, cell line-specific protein expression, paralleling the mRNA level, was confirmed by immunoblot. KIF14 mRNA was also highly overexpressed in retinoblastoma tumors and cell lines; samples with 1q genomic gain had higher KIF14 levels than those without. Medulloblastoma cell lines and primary lung tumors (which display 1q gain in some cases1) also displayed KIF14 overexpression. Among the lung tumors, KIF14 mRNA overexpression showed a strong trend toward association with decreased survival.2

Although cloned several years ago,3 KIF14's cellular function is not yet clear. A recent report indicates that KIF14 inhibition causes a delay in the metaphase-to-anaphase transition, characterized by misaligned chromosomes that oscillate abnormally between the spindle pole body and the metaphase plate.4 This finding is consistent with the known functions of the Drosophila homolog of KIF14, Klp38B, which binds chromatin, acts as a microtubule-dependent molecular motor (like all kinesins) and is involved in formation of the mitotic spindle.5, 6, 7, 8 In a microarray study, KIF14 expression was found to be cell cycle regulated with peak expression in S phase9; the KIF14 promoter is bound by E2F4 and p130.10 In summary, these data suggest that KIF14 plays an important role in mitosis.

Despite the limited existing functional data, the putative cell cycle role and specific overexpression of KIF14 in multiple cancers make it an attractive candidate oncogene. Our analyses in lung tumors also suggest that KIF14 expression may be a predictor of patient outcome.2 Moreover, as a kinesin, and thus an ATPase enzyme, KIF14 is an attractive therapeutic target. In the present study, we demonstrate for the first time KIF14 overexpression in primary breast tumors, and show that KIF14 expression is prognostic for patient outcome in breast cancer.

Material and methods

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Clinical samples

We obtained frozen sections from surgical samples of 99 breast tumors and 10 reduction mammoplasties from the Manitoba Breast Tissue Bank (MBTB).11 All tumors were from females who had not received any therapy at the time of resection. Clinical outcome data were available for 50 of these banked tumor samples. The grade of all samples was uniformly scored by a single pathologist, using the 7-point Nottingham score, and subsequently collapsed into Elston grade I (score 3–5), II (score 6–7) or III (score 8–9). Estrogen receptor (ER) and progesterone receptor (PR) status were determined by the Manitoba provincial laboratory, using ligand binding assays; a result of <3 was scored as an ER negative tumor and a result of <15 was scored as a PR negative tumor. Histological analyses of percentage cell type and inflammation (on a 5-point scale) were performed on hematoxylin and eosin-stained sections obtained from paraffin tissue blocks immediately adjacent and mirror image to the frozen tissue block sections used for RNA extraction and RT-PCR. The University Health Network Research Ethics Board and the MBTB Access Review Committee approved this study, and all subjects provided informed consent to the MBTB.

RNA extraction and reverse transcription

Molecular assays were performed blinded to the clinical details of the tumor samples. Total RNA was extracted from two 5 μm frozen sections by homogenizing in TRIzol reagent (Invitrogen, Burlington, ON, Canada), followed by chloroform extraction and isopropanol/ethanol precipitations. RNA was dissolved in RNase-free water and stored at −70°C. RNA yield and purity were evaluated by UV spectroscopy. To minimize RNA loss, no DNase treatment was performed. In pilot real-time reverse transcription (RT)-PCR experiments, negligible differences in KIF14 relative quantity were observed between identical samples with and without DNase treatment (data not shown).

For reverse transcription, 500 ng total RNA plus RNase-free water to a volume of 10 μl was combined with 3 ng random decamers (Invitrogen) and 500 μM RNase-free dNTPs (Invitrogen), heated at 65°C for 5 min, then snap cooled on ice. To each sample was added 1× SuperScript™ II reaction buffer (Invitrogen), 10 mM dithiothreitol and 40 U RiboLock™ RNase inhibitor (Fermentas, Burlington, ON, Canada). This reaction mixture was incubated at room temperature for 10 min, then warmed to 42°C for 2 min prior to the addition of 200 U SuperScript™ II RT (Invitrogen) to each reaction, except “no RT” control tubes. Samples were incubated at 42°C for 50 min, followed by enzyme inactivation at 70°C for 15 min. To ensure consistency, all elevated temperature incubations were performed in a thermal cycler (RoboCycler 96, Stratagene, La Jolla, CA). Samples were processed in 6 separate RT reactions, each including “no RT” and “no RNA template” negative controls.

To confirm RT efficacy, 1 μl of each reaction was tested in an end-point PCR for KIF14 and HPRT (hypoxanthine phosphoribosyl transferase, a housekeeping gene), using primers as follows: KIF14 forward 5′-AGCAGTTCTGAAAGGGAGCA-3′, reverse 5′-ATCACTGGCCAAGTTGCGAA-3′; HPRT forward 5′- TGCTCGAGATGTGATGAAGG-3′, reverse 5′-TCCCCTGTTGACTGGTCATT-3′. The 20 μl reaction mixture contained 1× KOD PCR buffer (Novagen, Madison, WI), 1 mM MgSO4, 0.2 mM each dNTPs, 1.3 M betaine, approximately 100 pmol each primer and 0.2 U KOD polymerase (Novagen). The reaction consisted of 2 min at 94°C, 30 cycles of 15 sec at 94°C, 30 sec at 55°C and 1 min at 68°C, with a final extension at 68°C for 10 min. The presence of an appropriately sized (192 bp) band for (at least) HPRT was confirmed by agarose gel electrophoresis.

Real-time RT-PCR

The products of RT reactions were diluted 2-fold, and 1 μl was added to 12.5 μl reactions with 1× TaqMan® PCR master mix including UNG (Applied Biosystems Incorporated [ABI], Foster City, CA or Eurogentec, San Diego, CA) and 1× TaqMan® Gene Expression Assay primer-probe mix (ABI) for KIF14 (Hs00978216_m1) or TBP (TATA-box binding protein), a house-keeping gene used as an endogenous control (Hs99999910_m1). Diagnostic markers ERBB2 (Hs00170433_m1) and MKI67 (Hs00606991_m1) were similarly assayed with TBP in a separate set of reactions. All primer sets used span exons. Triplicate reactions for each gene for each sample were prepared in 384-well plates, and PCR was performed using an SDS 7900HT system (ABI). Cycling conditions were 2 min at 50°C to activate UNG, 10 min at 95°C to activate the polymerase and 40 cycles of 15 sec at 95°C plus 1 min at 60°C. SDS 2.1 software (ABI) was used to calculate ΔΔCt relative expression values, normalized to TBP and calibrated to 1 sample in the assay (normal breast N19H). Relative quantity (RQ) was calculated according to the formula RQ ± SD = equation image. Calculations were performed using Microsoft Excel. Confirmation of equal efficiency of PCR amplification between each gene of interest and TBP was confirmed in pilot experiments by amplification of serial dilutions of a pool of RT products.

Statistical analyses

The right-skewed RQ distributions for KIF14, ERBB2 and MKI67 were normalized by log10 transformation. Associations between KIF14 expression (logRQ) and binary clinical variables were analyzed by unequal variance t test, among categorical variables by one-way ANOVA followed by Tukey's posthoc comparisons, and with continuous variables by Spearman's correlation coefficient. Survival analyses on dichotomized expression levels were performed using Kaplan-Meier log-rank statistics, while univariate and multivariate Cox regressions were used to analyze the effects of KIF14 level and covariates on survival. All statistical analyses were performed with SPSS v.11 for Mac OS X, and p-values ≤ 0.05 were considered significant in all tests.

Results

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

KIF14 expression, tumor grade and histology

We examined KIF14 mRNA expression by real-time RT-PCR in 99 female breast tumors and 10 normal breast tissue samples from reduction mammoplasties. Of the former samples, 85 were infiltrating ductal carcinoma, 9 were infiltrating lobular carcinoma and 5 were classified as showing mixed ductal and lobular features. Age at diagnosis, where known (n = 50), was 61.2 ± 16.2 years.

With normal samples excluded, relative KIF14, ERBB2 and MKI67 expression did not vary significantly between RT samples prepared on different dates or between different 384-well assay plates, plate rows or columns, and did not correlate with age at diagnosis (data not shown).

We initially asked whether KIF14 expression differed between normal breast and the entire population of tumors. The 10 normal samples had a mean KIF14 RQ, relative to a single normal sample, of 1.72 ± 1.99, while the 99 tumors had a mean RQ of 17.54 ± 19.9 (t test p = 0.000054) (Fig. 1a). Importantly, KIF14 expression increased with tumor grade (Fig. 1a) (Grade I mean RQ = 7.95 ± 10.0; Grade II = 18.2 ± 20.9; Grade III = 25.8 ± 21.7; ANOVA p = 0.000006); each grade differed significantly from the other two by Tukey posthoc comparisons. KIF14 expression also differed between the three histologic subtypes present in this cohort (Ductal mean RQ = 18.5 ± 19.2; Mixed = 23.1 ± 39.3; lobular = 5.4 ± 3.1; ANOVA p = 0.020), with ductal carcinoma having significantly higher KIF14 levels than lobular (Tukey posthoc p = 0.017) (Fig. 1b).

thumbnail image

Figure 1. KIF14 is overexpressed in breast tumors; expression increases with grade and is highest in ductal carcinoma. (a) Boxplot of KIF14 mRNA levels (relative to a single normal sample) determined by RT-PCR in 10 normal breast samples and 99 breast tumors stratified by grade. KIF14 levels were higher in the entire tumor set versus normals (t test p = 0.000054), and each grade had higher levels than the previous (ANOVA p = 0.000006), (b) KIF14 levels were higher in ductal carcinoma than lobular (Tukey posthoc p = 0.017); mixed ductal and lobular histology tumors had mid-range KIF14 levels.

Download figure to PowerPoint

KIF14 expression and other clinical variables

KIF14 expression was higher in ER negative than ER positive tumors (Fig. 2a) (mean RQ = 29.7 ± 26.6 versus 15.2 ± 17.6; t test p = 0.030). However, the trend to higher KIF14 expression in PR negative versus PR positive tumors was not significant (Fig. 2b) (mean RQ = 21.1 ± 21.1 versus 15.6 ± 19.1; t test p = 0.111). Among the 80 tumors for which reliable tumor size data were available, there was a trend toward a correlation between KIF14 expression and tumor size (Spearman r = 0.201, p = 0.073). KIF14 expression did not associate with inflammation score on a 5-point scale (I = 13.4 ± 12.8; II = 18.2 ± 22.3; III = 17.6 ± 25.1; IV = 18.9 ± 14.5; V = 24.7 ± 25.1; ANOVA p = 0.648), nor did KIF14 expression differ between node-negative and node-positive tumors (mean RQ = 16.3 ± 20.2 versus 19.1 ± 19.7; t test p = 0.247). However, among patients who had nodes removed, KIF14 expression correlated positively with the fraction of nodes that were tumor positive (r = 0.227, p = 0.024; Fig. 3a).

thumbnail image

Figure 2. KIF14 expression is higher in estrogen receptor negative tumors than positive, but does not differ between progesterone receptor positive and negative tumors. Boxplots of relative KIF14 mRNA levels stratified by (a) estrogen receptor status or (b) progesterone receptor status. * = significant difference, t test p = 0.030; n.s. = nonsignificant, p = 0.111.

Download figure to PowerPoint

thumbnail image

Figure 3. KIF14 expression correlates positively with fraction of nodes removed that were tumor-positive and percent invasive cells; and negatively with percent fatty stroma and percent normal epithelial cells. Scatterplots of relative log-transformed KIF14 mRNA expression (y-axes) with (a) fraction of positive nodes, (b) percent invasive cells, (c) percent fatty stromal cells and (d) percent normal epithelial cells. Spearman correlation coefficients and p-values are listed, and a line of best-fit ± 95% CI (dashed line) is indicated. Note different x-axis scales between graphs.

Download figure to PowerPoint

Extensive histological characterization of the tissue section adjacent to the section used for KIF14 analysis was available for these tumors. KIF14 expression correlated positively with percent invasive cells (r = 0.360, p = 0.0002; Fig. 3b), and negatively with percent fatty stroma (r = −0.258, p = 0.010; Fig. 3c) and percent normal epithelial tissue (r = −0.291, p = 0.003; Fig. 3d). There was no correlation between KIF14 expression and percent in situ tumor or percent collagenous stroma (data not shown).

We determined the mRNA levels by real-time RT-PCR of independent prognostic factors ERBB2 (HER-2/neu) and MKI67 (encoding the proliferation marker recognized by the Ki-67 antibody) in this cohort. As expected, both these markers showed higher expression in tumors than normal samples (ERBB2: 7.01 ± 16.0 versus 0.75 ± 0.51, t test p = 0.000002; MKI67: 9.71 ± 7.2 versus 1.37 ± 1.16, p = 0.000005), while MKI67 mRNA level increased with grade (ANOVA p = 0.000001), as has been observed previously.12 Importantly, amongst the tumor samples, KIF14 mRNA level correlated strongly with MKI67 level (Spearman r = 0.692, p = 0.000001; Fig. 4a). However, KIF14 mRNA level and ERBB2 mRNA level taken as a continuous variable did not correlate. The majority of tumor samples (87 of 99) had ERBB2 levels less than 7-fold higher than the normal calibrator sample. However, 12 samples were clear outliers, with ERBB2 levels 11-fold to 88-fold higher than normal. These 12 samples, which likely represent a subpopulation with ERBB2 gene amplification, had KIF14 levels higher than the remaining 87 samples (28.6 ± 28.0 versus 16.0 ± 18.2, t test p= 0.041; Fig. 4b).

thumbnail image

Figure 4. KIF14 expression correlates positively with mRNA levels of the proliferation marker MKI67 and is higher in ERBB2 overexpressing tumors. (a) Scatterplot of relative log-transformed KIF14 mRNA expression versus relative log-transformed MKI67 mRNA expression. Spearman correlation coefficient and p-value are listed, and a line of best-fit ± 95% CI (dashed line) is indicated, (b) Twelve of 99 tumors had outlying high ERBB2 mRNA levels (>11-fold of normals), suggestive of gene amplification. Boxplot stratifies KIF14 expression by ERBB2 mRNA level. * = significant difference, p = 0.041.

Download figure to PowerPoint

KIF14 expression and overall survival

Detailed treatment and outcome data were available from the MBTB for 50 of the 99 tumors analyzed, with a median follow-up time of 111 months. KIF14 expression was split around the mean RQ of the 10 normal samples plus 3 SD (1.72 + 3 × 1.99 = 7.69), offering a conservative upper limit to “normal” expression. This division yielded a group of 15 patients with “normal” KIF14 levels (mean RQ = 5.2 ± 1.7), and a group of 35 patients with “high” KIF14 levels (mean RQ = 22.0 ± 18.8). There was a significant overall survival advantage to the patients whose tumors were in the “normal KIF14 level” group (mean survival±SEM = 147.5 ± 17.8 versus 91.6 ± 11.0 months; Kaplan-Meier log rank p = 0.026; Fig. 5a). Results were similar when only death due to breast cancer was scored as an event (mean survival±SEM = 161.4 ± 17.1 versus 105.9 ± 11.8 months; p = 0.041; Fig. 5b).

thumbnail image

Figure 5. KIF14 expression predicts overall survival and disease-free survival in breast cancer. Survival plots of (a) overall survival, (b) death due to primary breast tumor alone and (c) disease-free survival. Relative KIF14 mRNA expression in 50 tumors was dichotomized based on the mean relative expression level in 10 normal samples plus 3 SD, offering a conservative estimate of the “normal” KIF14 expression range. KIF14 overexpression (solid lines) was associated with reduced survival in all 3 analyses when compared with normal KIF14 expression (dashed lines). Kaplan-Meier log-rank p-values are indicated. Black diamonds indicate censored cases.

Download figure to PowerPoint

When KIF14 level was used in a univariate Cox regression analysis, it significantly correlated with survival (p = 0.010), with an odds ratio (OR) of 3.60 (95% CI: 1.37–9.48). Separate univariate Cox survival analyses indicated that overall survival time was decreased for patients with PR negative tumors (p = 0.017), larger tumors (p = 0.049), older age at diagnosis (p = 0.008) and who were treated with surgery only (p = 0.026). Increasing grade showed a trend toward decreased survival (p = 0.063). However, in this dataset, ER status, tumor histology, inflammation score, ERBB2 and MKI67 mRNA levels, other treatment modalities (radiotherapy, chemotherapy, or hormone therapy, treated as independent variables) and nodal status did not significantly influence survival. Based on these data, we performed a multivariate Cox regression with KIF14 level, PR status, tumor size, age at diagnosis, surgery status, and grade as covariants. Once these covariants were included in the model, KIF14 expression was no longer a significant predictor of overall survival.

Similarly, in a univariate Cox regression analysis, KIF14 level significantly correlated with death due only to breast cancer: p = 0.018, OR: 3.86 (1.26–11.79). Univariate analyses of all other variables indicated decreased survival due to breast cancer for patients with PR negative tumors (p = 0.004), ER negative tumors (p = 0.029), increasing grade (p = 0.024), increasing tumor size (p = 0.003), increasing MKI67 mRNA level (p = 0.027) and percent invasive cells (p = 0.031). However, tumor histology, inflammation score, ERBB2 mRNA level, treatment modality, age at diagnosis and nodal status did not influence time to death from breast cancer alone. In a multivariate Cox regression with PR status, ER status, grade, tumor size, MKI67 level, percent invasive and KIF14 level as covariants, KIF14 logRQ was, again, not a significant predictor of survival.

Since ductal carcinoma showed higher KIF14 levels than lobular samples (Fig. 1b), we performed similar survival analyses on the ductal carcinoma samples alone (n = 46). Results were comparable to those for the entire group. Although Kaplan-Meier log-rank statistics for dichotomized samples were not significant (data not shown), in univariate Cox regressions, KIF14 expression significantly predicted overall survival: p = 0.031, OR 3.02 (1.11–8.25); and death by breast cancer only: p = 0.043, OR 3.32 (1.04–10.60). However, KIF14 was not a significant predictor of survival in multivariate analyses.

KIF14 expression and disease-free survival

When patients were divided into “high” and “normal” KIF14 expression groups, as described above, disease-free survival time was longer in the “normal” group (mean survival±SEM = 135.3 ± 17.4 versus 74.2 ± 12.5 months; Kaplan-Meier log-rank p = 0.049; Fig. 5c). However, in a univariate Cox regression, KIF14 expression was not a significant predictor of disease-free survival (p = 0.11), with an OR of 2.24 (0.83–6.07). Disease-free survival in this cohort did not differ significantly by any other clinical variable except tumor size (p = 0.014), MKI67 mRNA level (p = 0.013) and fraction of positive nodes (p = 0.034).

Discussion

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

We investigated the association of KIF14 expression with pathological and prognostic variables in breast cancer, and found that not only is KIF14 overexpressed in tumors in a grade-dependent fashion, it is also a predictor of overall and disease-free survival. Although we saw considerable variability in KIF14 expression, the overexpression phenotype is specific to breast tumor cells; ninety-one of the 99 breast tumors had KIF14 levels higher than all but one of the 10 control tissue samples (Fig. 1a). The negative correlation between KIF14 expression and the percentage of fatty stromal cells (Fig. 3c) and normal epithelial cells (Fig. 3d) in each sample corroborates this tumor cell specificity, and suggests that KIF14 expression is higher in malignant than in normal epithelial tissue. Our previous finding of very low KIF14 expression in other normal, highly epithelial tissues (lung, pancreas, kidney) supports this assertion.2 Nor is KIF14 overexpression likely to be part of an inflammatory response. Rather, increased KIF14 expression seems to be associated with more advanced, aggressive tumors; its expression increases pronouncedly with grade (Fig. 1a), fraction of tumor-positive nodes removed (Fig. 3a) and percent invasive cells (Fig. 3b). It is also expressed at higher levels in tumors with ductal (Fig. 1b), ER negative (Fig. 2a) and ERBB2 overexpression (Fig. 4b) phenotypes. In sum, these data suggest that KIF14 expression may be associated with high mitotic rate; it is a potential novel proliferation marker.

This putative role of KIF14 as a proliferation marker is underscored by the strong correlation we observed between KIF14 and MKI67 mRNA levels (Fig. 4a). Proliferation index (which may be measured by MKI67 mRNA levels12) was one of the 8 independent diagnostic parameters for guiding adjuvant treatment recommended by the consensus report of the St. Gallen 2005 conference.13 However, KIF14 level is unlikely to be a substitute for Ki-67 staining in this context, since in the present cohort its expression is not independent of 6 of the other St. Gallen parameters: size, histology, grade, hormone receptor status, lymph node status, ERBB2 status (which may be ascertained by mRNA levels).14 The final St. Gallen parameter, peritumor vascular invasion, could not be assessed in the present study.

KIF14 may not show independent diagnostic utility in the context of the St. Gallen criteria, but our demonstration that KIF14 is markedly and specifically overexpressed in almost all breast tumors supports consideration of KIF14 as a drug target with a high degree of tumor cell specificity. As an enzyme, KIF14 could be a more tractable antiproliferative drug target than other proliferation markers. It remains possible that KIF14 activation is a late event in tumorigenesis, suggested by the increase in KIF14 expression with grade (Fig. 1a). However, a cytogenetic study of a large cohort of breast tumors has suggested that gain of 1q is an early event in oncogenesis15; if KIF14 is indeed the gene “driving” 1q gain in breast tumors, it may play a role in early tumor development.

Although KIF14 expression is not a predictor of nodal metastasis, KIF14 expression does correlate with the severity of nodal invasion among node positive tumors (Fig. 3a) and with grade (Fig. 1a), indicating potential utility in molecular staging, to supplement histopathological evaluation. It will be interesting to ascertain whether KIF14 expression remains high in metastatic nodes and other metastases.

Increased KIF14 expression is associated with decreased survival time, when considering either death by all causes (Fig. 5a), or death by breast cancer alone (Fig. 5b). As mentioned, this effect is not independent of other prognostic variables, but could contribute to a novel molecular tumor signature, providing prognostic information from a very small biopsy (ideally suited to real-time RT-PCR analysis), or even circulating tumor cells. Even if KIF14 mRNA level is simply a novel proliferation marker, it offers an appealing alternative or supplement to other proliferation-based prognostic indicators such as mitotic index, Ki-67 or cyclin A expression (reviewed in Ref.16). However, we recognize the limited power of our relatively small cohort of patients with available survival data (n = 50) in this initial study. There may perhaps be a random bias in this population, too, since increasing age at diagnosis decreased survival; the opposite is usually the case.17 Analysis of KIF14 expression in a larger cohort of tumor samples might yet show a significant independent effect of KIF14 expression on survival.

Although KIF14 overexpression predicted disease-free survival in a Kaplan-Meier analysis (Fig. 5c), this effect was not evident when KIF14 expression was examined as a continuous, independent predictor of disease-free survival in a Cox regression. Again, limited power or a random bias in the small study population might explain this result, suggested by the fact that few other of the “expected” clinical variables significantly predicted disease-free survival in this cohort; confirmation in a larger cohort is warranted. Moreover, disease-free survival is not as valuable an epidemiological variable as overall survival in retrospective studies such as this, since time of death is definitive, but time of recurrence is subjective.

To our knowledge, this is the first report of an association between expression of a kinesin family member and grade or survival in breast cancer. In fact, there is a paucity of data linking expression of kinesins with outcome in any cancer, despite the proven preclinical therapeutic efficacy of a number of kinesin inhibitors considered as antimitotic drugs.18, 19, 20 However, KIF2 and a gene later identified as KIF15 were identified as breast tumor antigens21; the latter gene was overexpressed in breast tumors. The aurora kinases, which interact with kinesins to regulate the mitotic spindle, are oncogenes. They are overexpressed in numerous cancers and correlate with aggressiveness in breast cancer (reviewed in Ref.22). Dynein light chain, a component of another microtubule-dependent cellular motor, the dynein complex, has also recently been shown to be overexpressed in breast tumors and displays increasing expression in the MCF10AT cell culture model of breast tumor progression.23 Beyond these studies, we are not aware of any other epidemiological analyses of individual kinesin or related protein expression levels in primary tumors. Examination of other mitotic kinesins as potential prognostic factors is clearly warranted, as is investigation of KIF14's effects on survival in other tumor types.

Because of the lack of an effective, specific anti-KIF14 antibody for immunohistochemistry, we elected to perform real-time RT-PCR in the current study. Immunohistochemistry remains the gold standard for pathological evaluations, since formalin fixing and paraffin embedding is the preferred method for tumor archiving, and protein levels equate more closely with function than mRNA levels (although KIF14 protein levels do correlate with its mRNA levels in breast cancer cell lines,2 and ERBB2 mRNA levels are reflective of both protein level and genomic copy number14). Thus, application of the methodology described here for routine laboratory prognostic analyses of breast tumors remains challenging. However, real-time RT-PCR, particularly for determination of overexpression as a prognostic variable, holds 3 key advantages over immunohistochemistry. First, real-time RT-PCR makes very efficient use of tumor material; enough RNA for expression analysis of many genes of interest, in triplicate, can be obtained from 1 or 2 frozen sections. For example, we were able to rapidly assess MKI67 and ERBB2 mRNA levels along with KIF14 in this study. Second, TaqMan-based RT-PCR chemistry is exquisitely specific and sensitive, whereas nonspecific binding can confound antibody-based techniques. Third, and most importantly, real-time RT-PCR gives a sensitive, robust, objective and above all quantitative indication of transcript abundance. This allows for the application of powerful, parametric statistical analyses, and raises the possibility of defining a threshold RQ for pathological overexpression.

Although the exact function of KIF14 remains obscure, its role in aligning chromosomes at metaphase4 suggests that it could play a role in the development of aneuploidy, and its overexpression may function to accelerate the cell cycle in cancer cells. Whatever its mode of action, its specific overexpression in breast tumors and association with grade and survival demonstrate its importance as a putative oncogene, therapeutic target and prognostic indicator.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

This study would not have been possible without tissue samples and clinical data supplied by Peter Watson, Michelle Parisien, Linda Curtis and Yulian Niu of the Manitoba Breast Tumor Bank, Winnipeg, Manitoba, which is funded in part by Cancer Care Manitoba and the Canadian Institutes of Health Research. We also thank Lisa Wang for statistical advice, and Peter Watson, Michael Jewett and members of the Gallie Laboratory for critical reading of the manuscript. This work was supported by grants to B.L.G. from the National Cancer Institute of Canada with funds from the Terry Fox Run and the Canadian Cancer Society, the Canadian Genetic Diseases Network, the Grand Chapter of Royal Arch Masons of Canada, and the Keene Annual Perennial Plant Sale. T.W.C. was supported by a Canadian Institutes of Health Research Canada Graduate Scholarship.

References

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  • 1
    Baudis M, Cleary ML. Progenetix.net: an online repository for molecular cytogenetic aberration data. Bioinformatics 2001; 17: 12289.
  • 2
    Corson TW, Huang A, Tsao M-S, Gallie BL. KIF14 is a candidate oncogene in the 1q minimal region of genomic gain in multiple cancers. Oncogene 2005; 24: 47414.
  • 3
    Nomura N, Nagase T, Miyajima N, Sazuka T, Tanaka A, Sato S, Seki N, Kawarabayasi Y, Ishikawa K, Tabata S. Prediction of the coding sequences of unidentified human genes. II. The coding sequences of 40 new genes (KIAA0041–KIAA0080) deduced by analysis of cDNA clones from human cell line KG-1. DNA Res 1994; 1: 25162.
  • 4
    Zhu C, Zhao J, Bibikova M, Leverson JD, Bossy-Wetzel E, Fan JB, Abraham RT, Jiang W. Functional analysis of human microtubule-based motor proteins, the kinesins and dyneins, in mitosis/cytokinesis using RNA interference (RNAi). Mol Biol Cell 2005; 16: 318799.
  • 5
    Alphey L, Parker L, Hawcroft G, Guo Y, Kaiser K, Morgan G. KLP38B: a mitotic kinesin-related protein that binds PP1. J Cell Biol 1997; 138: 395409.
  • 6
    Molina I, Baars S, Brill JA, Hales KG, Fuller MT, Ripoll P. A chromatin-associated kinesin-related protein required for normal mitotic chromosome segregation in Drosophila. J Cell Biol 1997; 139: 136171.
  • 7
    Ohkura H, Torok T, Tick G, Hoheisel J, Kiss I, Glover DM. Mutation of a gene for a Drosophila kinesin-like protein, Klp38B, leads to failure of cytokinesis. J Cell Sci 1997; 110: 94554.
  • 8
    Ruden DM, Cui W, Sollars V, Alterman M. A Drosophila kinesin-like protein, Klp38B, functions during meiosis, mitosis, and segmentation. Dev Biol 1997; 191: 28496.
  • 9
    Whitfield ML, Sherlock G, Saldanha AJ, Murray JI, Ball CA, Alexander KE, Matese JC, Perou CM, Hurt MM, Brown PO, Botstein D. Identification of genes periodically expressed in the human cell cycle and their expression in tumors. Mol Biol Cell 2002; 13: 19772000.
  • 10
    Cam H, Balciunaite E, Blais A, Spektor A, Scarpulla RC, Young R, Kluger Y, Dynlacht BD. A common set of gene regulatory networks links metabolism and growth inhibition. Mol Cell 2004; 16: 399411.
  • 11
    Watson PH, Snell L, Parisien M. The NCIC-Manitoba Breast Tumor Bank: a resource for applied cancer research. CMAJ 1996; 155: 2813.
  • 12
    Esteva FJ, Sahin AA, Cristofanilli M, Coombes K, Lee SJ, Baker J, Cronin M, Walker M, Watson D, Shak S, Hortobagyi GN. Prognostic role of a multigene reverse transcriptase-PCR assay in patients with node-negative breast cancer not receiving adjuvant systemic therapy. Clin Cancer Res 2005; 11: 33159.
  • 13
    Goldhirsch A, Glick JH, Gelber RD, Coates AS, Thurlimann B, Senn HJ. Meeting highlights: international expert consensus on the primary therapy of early breast cancer 2005. Ann Oncol 2005; 16: 156983.
  • 14
    Vanden Bempt I, Vanhentenrijk V, Drijkoningen M, Wlodarska I, Vandenberghe P, De Wolf-Peeters C. Real-time reverse transcription-PCR and fluorescence in situ hybridization are complementary to understand the mechanisms involved in HER-2/neu overexpression in human breast carcinomas. Histopathology 2005; 46: 43141.
  • 15
    Höglund M, Gisselsson D, Hansen GB, Sall T, Mitelman F. Multivariate analysis of chromosomal imbalances in breast cancer delineates cytogenetic pathways and reveals complex relationships among imbalances. Cancer Res 2002; 62: 267580.
  • 16
    van Diest PJ, van der Wall E, Baak JP. Prognostic value of proliferation in invasive breast cancer: a review. J Clin Pathol 2004; 57: 67581.
  • 17
    Kroman N, Jensen MB, Wohlfahrt J, Mouridsen HT, Andersen PK, Melbye M. Factors influencing the effect of age on prognosis in breast cancer: population based study. BMJ 2000; 320: 4748.
  • 18
    Leizerman I, Avunie-Masala R, Elkabets M, Fich A, Gheber L. Differential effects of monastrol in two human cell lines. Cell Mol Life Sci 2004; 61: 206070.
  • 19
    Mayer TU, Kapoor TM, Haggarty SJ, King RW, Schreiber SL, Mitchison TJ. Small molecule inhibitor of mitotic spindle bipolarity identified in a phenotype-based screen. Science 1999; 286: 9714.
  • 20
    Sakowicz R, Finer JT, Beraud C, Crompton A, Lewis E, Fritsch A, Lee Y, Mak J, Moody R, Turincio R, Chabala JC, Gonzales P, et al. Antitumor activity of a kinesin inhibitor. Cancer Res 2004; 64: 327680.
  • 21
    Scanlan MJ, Gout I, Gordon CM, Williamson B, Stockert E, Gure AO, Jager D, Chen YT, Mackay A, O'Hare MJ, Old LJ. Humoral immunity to human breast cancer: antigen definition and quantitative analysis of mRNA expression. Cancer Immun 2001; 1: 4.
  • 22
    Katayama H, Brinkley WR, Sen S. The Aurora kinases: role in cell transformation and tumorigenesis. Cancer Metastasis Rev 2003; 22: 45164.
  • 23
    Vadlamudi RK, Bagheri-Yarmand R, Yang Z, Balasenthil S, Nguyen D, Sahin AA, den Hollander P, Kumar R. Dynein light chain 1, a p21-activated kinase 1-interacting substrate, promotes cancerous phenotypes. Cancer Cell 2004; 5: 57585.