The names of cell lines used in our investigations are listed in Table 1. The derivation, sources, and maintenance of most of the breast cancer cell lines used in this study have been reported previously 13 or were provided in Table 2. These cell lines have been previously analyzed for genomic aberrations by comparative genomic hybridization (CGH) and for gene-expression profiles using Affymetrix microarrays (Santa Clara, CA, USA) 13. The information on growth conditions of additional cell lines was listed in Table 2. Cells at 50% to 75% confluence were washed in ice-cold phosphate buffered saline (PBS). Then cells were extracted with a lysis buffer (containing 50 mM HEPES (pH 7.5), 150 mM NaCl, 25 mM β-glycerophosphate, 25 mM NaF, 5 mM EGTA, 1 mM EDTA, 15 mM pyrophosphate, 2 mM sodium orthovanadate, 10 mM sodium molybdate, 1% Nonidet-P40, 10 mg/ml leupeptin, 10 mg/ml aprotinin, and 1 mM PMSF). Cell lysates were then clarified by centrifugation and frozen at -80°C. Protein concentrations were determined using the Bio-Rad BCA protein assay kit (Cat# 23227, Pierce Biotechnology, Rockford, IL, USA).

For Western blots, 10 μg of protein extracts per lane were electrophoresed with denaturing sodium doedecyl sulfate (SDS)-polyacrylamide gels (4% to 12%), transferred to PVDF membranes (Millipore, Temecula, CA, USA), and incubated with HJURP antibody 1:500 (Rabbit, HPA008436, Sigma-Aldrich, St. Louis, MO, USA) and actin (goat, sc-1616, Santa Cruz Biotechnology, Santa Cruz, CA, USA) diluted with blocking buffer (927-40000, LI-COR Biosciences, Lincoln, NE, USA) The membranes were washed four times with TBST and treated with 1:10,000 dilution of Alex Fluor 680 donkey anti-rabbit (A10043, Invitrogen, Carlsbad, CA, USA) and IRDye 800CW conjugated donkey anti-goat (611-731-127, Rockland, Gilbertsville, PA, USA) to detect HJURP and actin respectively. The signals were detected by infrared imaging (LI-COR Biosciences, Lincoln, NE, USA). Images were recorded as TIFF files for quantification.

Protein levels were measured by quantifying infrared imaging recorded from labeled antibodies using Scion Image 14. For each protein, the blots were made for 7 sets of 11 cell lines, each set including the same pair (SKBR3 and MCF12A) to permit intensity normalization across sets. A basic multiplicative normalization was carried out by fitting a linear mixed effects model to log intensity values, and adjusting within each set to equalize the log intensities of the pair of reference cell lines across the sets.

Detailed patient information has been described in our previous studies 15. This analysis is based on previously reported comparative genomic hybridization (CGH) and a gene expression profile of 130 tumors from UC San Francisco and the California Pacific Medical Center collected between 1989 and 1997.

The association of HJURP expression levels and survival among patients with breast tumors was examined in existing microarray data sets of primary tumor samples that had been profiled with an Affymetrix microarray assay (either HG-U133A or HG U133 Plus 2.0) ((GEO:GSE1456), (GEO:GSE7390), (GEO:GSE2034), (GEO:GSE4922)) or Agilent oligo microarray (Santa Clara, CA, USA)(Table 3). Probe 218726_at and 20366 (GenBank: NM_018410) were used to measure HJURP expression in Affymetrix and Agilent GeneChip, respectively. The process data from GEO website were downloaded for analysis.

The shRNA sequences were (forward) 5'-GATCCCC GAGCGATTCATCTTCATCA TTCAAGAGA TGATGAAGATGAATCGCTC TTTTTGGAAA-3' and (reverse) 5'-AGCT TTTCCAAAAA GAGCGATTCATCTTCATCA TCTCTTGAA TGATGAAGATGAATCGCTC GGG-3' synthesized from IDT (Integrated DNA Technologies, Inc., San Diego, CA, USA). HJURP shRNA was cloned into BglII and HindIII cleavage sites of pSUPER.retro.puro vector based on manufactory's instruction (OligoEngine, Seattle, WA, USA). HJURP shRNA expression vector were confirmed by direct DNA sequencing.

HJURP shRNA (or empty) retroviral vectors along with packaging system pHit60 and pVSVG vectors were then co-transfected into the HEK 293 Phoenix ampho packaging cells (ATCC, Manassas, VA, USA) by using FuGENE6 transfection reagent (Roche, Lewes, UK) according to the instruction to produce retroviral supernatants. Forty-eight hours after transfection, the virus-containing supernatant was filtered through a 0.45 μm syringe filter. Retroviral infection was performed by adding filtered supernatant to a MDAMB231 cell line cultured on 10 cm dishes with 50% confluent in the presence 4 ug/ml of polybrene (Sigma, St. Louis, MO, USA). Six hours after infection, the medium was changed with fresh medium. After 48 hours, infected cells were selected by adding 5 μg/ml puromycin (Sigma) to the culture medium for 72 hours and then maintained in complete medium with 2 μg/ml puromycin. Down-regulation of HJURP expression was confirmed by Western blot analysis.

The effects on cell growth and apoptosis were assessed by a Cellomics high-content image screening system (Cellomics, Thermo Fisher Scientfic Inc., Pittsburgh, PA, USA) after breast cancer cells exposed to a single dose of 0 (sham), 1, 2, 4, 6, 8 or 10 Gy X-ray radiation emitted from an irradiator (model 43855F, Faxitron X-ray Corporation, Lincolnshire, IL, USA). Live cells in 96 well plates with six replicates from each treatment were stained with 1 μmol/L YO-PRO-1 positive cells.

Spearman's correlation coefficient and test were used to examine the relationship between HJURP mRNA level and its protein level in the cell line studies, and the relationship with age, tumor size in the tumor studies, and CENPA mRNA level. The association between HJURP mRNA level and clinical factors, such as estrogen receptor (ER), progesterone receptor (PR), ERBB2 and lymph node status, pathological stage, Scarff-Bloom-Richardson (SBR) grade, was analyzed by Mann-Whitney U (for two groups) or Kruskal-Wallis H (for more than two groups) test. Kaplan-Meier plots were constructed and a long-rank test was used to determine differences among disease free and overall survival curves according to HJURP expression level or radiotherapy. Multivariate analyses were carried out to examine whether HJURP expression is an independent prognostic factor for survival when adjusting for other covariates (age, ER, PR, lymph node, pathologic stage, SBR grade, tumor size) or the molecular subtypes (normal like, luminal, Erbb2 and Basal) using Cox proportional-hazard regression. In addition, the relation between HJURP expression and survival was explored in microarray data sets by dividing the cases from each cohort into a group with high (top one-third), moderate (middle one-third), and low (bottom one-third) level of expression. All analyses were performed by SPSS 11.5.0 for Windows. A two-tailed P-value of less than 0.05 was considered to indicate statistical significance.

We examined the protein levels of HJURP in a large panel of human breast cancer cell lines and immortalized non-malignant mammary epithelial cells, which have been analyzed for genomic aberrations by comparative genomic hybridization (CGH) and for gene-expression profiles using Affymetrix microarrays 13. Although we found few genetic alterations in the HJURP locus by inspection of these CGH microarray data, the protein levels of HJURP were elevated in about 50% of these breast cancer cell lines when compared to immortalized but non-malignant mammary epithelial cells 184A1N4, 184B5, and S1 (Figure 1a, b). In order to determine whether mRNA expression reflected protein levels, we quantified and normalized HJURP protein expression in each cell line and demonstrated a significant correlation between mRNA expression and protein levels (the Affymetrix probe for HJURP is 218726_at: Spearman's correlation coefficient R = 0.55, P < 0.001; Figure 1c). Next we examined whether HJURP protein level is associated with cell proliferation. In order to do so, we measured the doubling time for each cell line and found that the doubling time of cell lines was negatively correlated with HJURP protein levels (Spearman's correlation coefficient R = -0.395, P = 0.005; Figure 1d). Furthermore, HJURP mRNA levels in invasive ductal carcinomas (IDC) were statistically significantly higher than its levels in the normal breast ducts (P < 0.0001) (Figure 1e) 16.

We assessed the association between HJURP mRNA levels and clinical factors and outcomes using a cohort of breast cancer patients in our previous studies 15. HJURP expression level is measured as log₂ (probe intensities) by Affymetrix microarray. In univariate analysis, HJURP mRNA levels were not associated with pathological stage, tumor size, ERBB2 positive, or lymph node positive status (Figure 2a, b, c, d). However, high HJURP mRNA levels were significantly associated with estrogen-receptor negative (ER-) (P < 0.0001), progesterone-receptor negative (PR-) P < 0.0001), advanced SBR grade (P < 0.0001), young age (P < 0.001) and Ki67 proliferation indices (P < 0.001) (Figure 2e, f, g, h, i). When we divided HJURP expression levels into three groups (low = bottom third, moderate = middle third, and high = top third), patients whose tumor with high HJURP expression levels had significantly shorter disease free survival (P = 0.0009) and overall survival (P = 0.0017) period using a Kaplan-Meier log rank analysis (Figure 3a). Interestingly, although HJURP expression significantly correlated with Ki67 proliferation indices, Ki67 proliferation indices are not significantly associated with both disease-free and overall survival (Figure 3b).

In multivariate analyses (including age, pathological stage, SBR grade, ER status, PR status, lymph node status, tumor size, HJURP mRNA levels), lymph node positive and high pathological stage were associated with poor disease free survival, whereas lymph node positive, big tumor size, and age were associated with poor overall survival (Table 4). HJURP expression level is an indicator of a poor prognosis for disease-free survival (hazard ratio, 2.05; 95% CI, 1.18 to 3.58; P = 0.011), and for overall survival (hazard ratio, 1.83; 95% CI, 1.11 to 3.01; P = 0.018) (Table 4).

To validate our findings, we used several independent breast cancer cohorts with previously reported microarray data deposited in the Gene Expression Omnibus (GEO) database 17, to compare mRNA level of HJURP in tumor tissue with patient survival (Table 3). In agreement with our initial findings, decreased disease-free and overall survival rate was associated with high mRNA level of HJURP in all of the datasets (Figures 4 and 5).

Finally, we investigated whether HJURP mRNA levels were an independent prognostic factor over molecular subtypes (normal like, luminal, Erbb2 and basal) using Cox regression. In order to do so, three data sets (reference 14, Dataset 1 and 3), in which the information of the molecular subtypes was available, were combined because there were few patients in each subtype using each data set. As showed in Table 5, both HJURP mRNA levels and molecular subtypes were independently significantly associated with survival.

It has been reported that HJURP is involved in the DNA repair pathway, thus next we investigated whether the HJURP mRNA level is a predictive marker for radiotherapy in our cohort of breast cancer patients. As shown in Figure 6a, the radiotherapy significantly increased disease-free survival of patients within the high HJURP mRNA level group (P = 0.022) whereas radiotherapy did not within the low HJURP mRNA level group. The data showed a trend toward increased overall survival within the high and moderate HJURP mRNA level group (Figure 6b).

In order to confirm the relationship between HJURP mRNA levels and radiation sensitivity, we selected two cell lines, one had high levels of HJURP (MDAMB231), the other had a low level of HJURP (T47D), and treated them with different doses of x-ray irradiation. Seventy-two hours after radiation, we measured cell growth and apoptosis using high-content image analysis. Our data showed that the response to radiation in breast cancer cell line MDAMB231 (IC50 = 3.5 Gy) was more sensitive than T47D (IC50 = 8.6 Gy) (Figure 7a). Consistent with radiation sensitivity, MDAMB231 cells had ahigher rate of apoptosis than T47D cells (Figure 7b). Similar results were found in additional cell lines BT20 with high levels of HJURP and MCF10A with low levels of HJURP (Figure 7c, d). Finally we designed small interfering RNA (shRNA) against HJURP and generated stable transfectants in a human breast cancer cell line (MDAMB231). The shRNA down-regulated HJURP protein levels by 75%, as assessed by Western blotting assays (Figure 7e). Knockdown of the HJURP gene reduced the sensitivity to radiation (Figure 7f).

Recently it has been shown that HJURP interacts with CENPA for localization to centromeres and for accurate chromosome segregation. Thus we examined the expression pattern between HJURP and CENPA at the mRNA level. Surprisingly, HJURP levels were significantly and positively correlated with CENPA levels in human breast cancer cell lines (Figure 8a) and primary breast tumors (Figure 8b). Such highly significant correlation was confirmed in four independent cohorts with breast tumors (Figure 8c, d, e, f).