Genistein and enterolactone in relation to Ki-67 expression and HER2 status in postmenopausal breast cancer patients
Abstract
Scope
Phytoestrogens (PE) may improve breast cancer prognosis by modifying tumor prognostic markers, such as cell proliferation marker Ki-67 and human epidermal growth factor receptor 2 (HER2). Epidemiological evidence linking lignans and isoflavones to Ki-67 and HER2 is limited. We examined associations between the major metabolites of lignans and isoflavones – enterolactone (ENL) and genistein (GEN) – respectively, and Ki-67 expression and HER2 in tumor tissue of breast cancer patients.
Methods and results
Data from 1060 invasive breast cancer patients from the population-based MARIE study were used. Multivariate-adjusted linear (Ki-67 log-transformed) and quantile regression, and logistic regression analyses (HER2, Ki-67 dichotomized) were performed to calculate β estimates and ORs, respectively. Median post-diagnostic ENL and GEN concentrations were 19.5 and 4.8 nmol/L, respectively. Median Ki-67 was 12.0%, and 21.2% of the tumors were HER2+. After adjustment, there was an inverse association between GEN and Ki-67 at high expression levels (OR for Ki-67 ≥20% versus <20% of 0.93 (95%CI [0.87;0.99]) per 10 nmol/L GEN increment).
Conclusion
Our findings indicate an inverse association between GEN and Ki-67 at high levels of Ki-67 expression. Additional investigations are recommended to confirm our findings and to further elucidate mechanisms linking PE metabolites to breast cancer survival.
Abbreviations
-
- ß
-
- beta estimate
-
- DAI
-
- daidzein
-
- ENL
-
- enterolactone
-
- ER
-
- estrogen receptor
-
- FS
-
- flaxseed
-
- GEN
-
- genistein
-
- HER2
-
- human epidermal growth factor receptor 2
-
- hrs/w
-
- hours per week
-
- HRT
-
- hormone replacement therapy
-
- IQR
-
- interquartile range
-
- LTPA
-
- leisure time physical activity
-
- MARIE
-
- Mammary Carcinoma Risk factor InvEstigation
-
- MET
-
- metabolic equivalent
-
- Mo
-
- months
-
- OR
-
- odds ratio
-
- PE
-
- phytoestrogens
-
- PR
-
- progesterone receptor
-
- SDG
-
- secoisolariciresinol diglycoside
-
- TMA
-
- tissue-microarrays
-
- TR-FIA
-
- time-resolved fluoroimmunoassays
-
- 95%CI
-
- 95% confidence interval
1 Introduction
Phytoestrogens (PE) are postulated to be protective for hormone-dependent diseases such as breast cancer because of the estrogenic and anti-estrogenic effects the metabolites can exhibit 1, 2. Lignans and isoflavones are the two main sub-groups of PEs, and their major sources are oilseeds (i.e. flaxseed (FS)) and grains, and soy products, respectively 2. Lignans are metabolized by the gut microflora to enterolactone (ENL) while isoflavones are converted, for example, to daidzein (DAI) and genistein (GEN) 3, 4. The structural similarities between these metabolites and the mammalian estrogen, 17β-estradiol, allow them to mimic and antagonize endogenous 17β-estradiol. By blocking estrogen-binding to the estrogen receptor of the tumor, PEs can potentially lower the hormonal effects of endogenous estrogens and reduce tumor cell proliferation. Phytoestrogens can also reduce the bioavailability of estrogens by activating or inhibiting key enzymes of hormone metabolism 2, 4, 5. In addition to these hormone-dependent effects, PEs also exhibit several hormone-independent benefits including their ability to lower metastasis, inflammation, and angiogenesis 2-5. On the other hand, PEs have also been suggested to exert possible adverse effects. For instance, there is evidence from experimental studies that they may also have potential tumour-promoting effects 2, 5.
Results from several large meta-analyses of cohort and case-control studies have shown that high lignan 6, 7 and isoflavone 8, 9 exposure is associated with decreased breast cancer risk 6, 8, all-cause mortality 7, breast cancer-specific mortality 7, 9, and recurrence 9. The biological mechanisms underpinning these relationships have not yet been fully elucidated, and in this analysis, we focus on the mechanisms that may link PEs to breast cancer survival. Several studies have examined the potential influence of PEs on expression of the cell proliferation marker Ki-67 and the human epidermal growth factor receptor 2 (HER2), which are established prognostic breast cancer factors in addition to age, tumor size, nodal status, and hormone receptor status (ER, progesterone receptor (PR)) 10-12. These studies exploring the influence of PEs on Ki-67 and HER2 expression in breast cancer have been predominantly human intervention studies using FS 13, 14 or soy supplementation 15-17, and with the exception of two case-control studies 18, 19, the current evidence is mainly limited to experimental studies 20-27.
Results from human interventions studies examining daily supplementation with FS, lignans, and soy on breast tumor Ki-67 expression have been somewhat equivocal, in that FS 14 and soy 15, 17 supplementation did not affect Ki-67 expression compared to those receiving placebo, but lignan supplementation decreased Ki-67 expression, although a control group was not included for comparison 13. Likewise, supplementation with FS compared to those taking placebo 14 and increased soy intake compared to low intake 19 has been associated with decreased breast tumor HER2 expression, although lignan intake was not associated with HER2 in a case-control study 18.
Overall then, there is suggestive epidemiological evidence to support an association between PEs and both Ki-67 and HER2. And given the importance of both lignans and isoflavones in impacting breast cancer survival, exploring potential relationships between lignans and isoflavones and Ki-67 and HER2, two prognostic markers of breast cancer, is crucial to improving our knowledge about possible etiological mechanisms relating lignans and isoflavones to breast cancer prognosis.
Here, we evaluate the influence of two PEs – ENL and GEN – on Ki-67 and HER2 expression. To our knowledge, this is the largest cross-sectional investigation into the associations between serum ENL and GEN, two metabolites of lignans and isoflavones, respectively, and Ki-67 expression and HER2 status in breast tumors of postmenopausal breast cancer patients, while also accounting for hormone-related factors in addition to other relevant covariates.
2 Materials and methods
2.1 Study population
We included postmenopausal breast cancer patients from the MARIE (Mammary Carcinoma Risk Factor Investigation) study, a large population based case-control study that has previously been used to assess associations between lignan exposure and breast cancer risk 6, 28 and survival 7. In brief, 3,813 patients aged 50–74 years at diagnosis of a histologically confirmed primary invasive breast cancer (ICD-10 C50) or in situ carcinoma (D05) were recruited between 2001 and 2005 from two study regions (Hamburg and Rhine-Neckar-Karlsruhe) in Germany through participating clinics and cancer registries. Participants completed a face-to-face interview and were requested to provide a blood sample. The present analysis was restricted to data from the Rhine-Neckar-Karlsruhe region due to the availability of paraffin tissue blocks for standardized determination of Ki-67 expression and HER2 status. Patients with ductal carcinoma in situ, patients who had undergone neo-adjuvant chemotherapy, and who had blood drawn before diagnosis or before the operation were excluded, resulting in a total of 1060 invasive breast cancer patients.
Ethical approval was granted by the ethics committees of the University of Heidelberg, the University of Hamburg, and the Medical Board of the State of Rhineland-Palatine. The study was conducted in accordance with the Declaration of Helsinki. All study participants provided written informed consent.
2.2 Phytoestrogen measurements
Non-fasting serum samples collected after patient recruitment were processed, divided into aliquots, and stored at -80°C. The median time between operation and blood collection was 148 days (SD 184 days, Min 0 days, Max 1,109 days). ENL and GEN concentrations in serum samples were determined using time-resolved fluoroimmunoassays (TR-FIA) (Labmaster Ltd., Turku, Finland) according to validated methods 29. The refinement process consisted mainly of enzymatic hydrolysis and liquid extraction (20 μL) of the samples (for details http://www.labmaster.fi/index.php?cat = 3&lang = en&project = ). To assess intra-assay coefficient of variation (CVs), the first 160 samples were measured in duplicate. The mean intra-assay and inter-assay CVs were 7.2 and 14.6% for ENL and 16.4 and 24.2% for GEN, respectively, which were adequately low. In addition, two quality control samples at varying plate positions were measured (mean ENL concentrations: 33.2 and 33.0 nmol/L; mean GEN concentrations: 3.2 and 3.3 nmol/L). Twenty-four ENL samples and fifty-seven GEN samples below the lower limit of detection (LOD) of 0.5 nmol/L for ENL and 1.0 nmol/L for GEN were reviewed and original values were taken. Four ENL samples and eighteen GEN samples were above the upper LOD of 300 nmol/L for both metabolites. Samples with values beyond the standard curve at ≥500 nmol/L for both metabolites (ENL n = 2; GEN n = 9) were excluded (Supporting Information Fig. 1).
2.3 Tumor markers Ki-67 and HER2
Tissue microarray (TMA) samples were provided by the tissue bank of the National Center for Tumor Diseases (NCT, Heidelberg, Germany) in accordance with the regulations of the tissue bank and the approval of the ethics committee of the University of Heidelberg. Hematoxylin&Eosin tissue sections from paraffin tissue blocks collected from 1,505 MARIE patients were prepared and scanned using the Aperio Scanscope CS system at the Department of Pathology, University of Heidelberg. TMAs using two cores from representative tumor areas per tissue block were created. Antibodies against HER2 (clone SP3, dilution 1:1000 CC1-standard; catalog no. 237R-16, Cell Marque) and Ki-67 (clone 30–9, dilution RTU CC1-standard; catalog no. 790–4286, Ventana) were assessed by immunohistochemistry (IHC) and scored by two pathologists. Ki-67 expression is defined as “the percentage of positively stained malignant cells among the total number of malignant cells” 30. HER2 expression was scored using IHC as 0, 1+, 2+, or 3+, and considered as positive for a score of 3+. For tumors scored as 2+ using IHC, HER2 status was determined by chromogenic in situ hybridization (CISH) with a ratio of >2.2 in a dual color assay. Additionally, pathology reports were available for 1739 breast cancer patients, and when information was available for Ki-67, HER2, ER, and PR from both TMAs and pathology reports, the highest value was used. Otherwise, values for Ki-67, HER2, ER and PR were derived from pathology reports. For Ki-67 and HER2 status, values were adopted from local pathology reports for 82 (10.2%) and 350 patients (36.5%), respectively.
After the exclusion criteria were applied, 1060 patients with available data on Ki-67 expression and/or HER2 status and ENL and/or GEN concentrations were included in our analyses (Supporting Information Fig. 1).
2.4 Statistical analysis
Differences in mean ENL and GEN concentrations according to population characteristics were assessed using ANOVA. Characteristics of the patients according to Ki-67 exposure (<20% (low); ≥20% (high)) 31, 32 and HER2 status (HER2+/-) were evaluated by using the chi-squared test.
The associations between ENL and GEN (continuous per 10 nmol/L increments and in quartiles) with Ki-67 as a log-transformed variable were estimated by calculating beta (β) estimates and corresponding 95% confidence intervals (95%CIs) using linear regression (PROC GLM) to explore potential linear relationships. Interpretation of the β is such that β is the expected percentage change in Ki-67 expression for a 10nmol/L increase in ENL/GEN (http://www.kenbenoit.net/courses/ME104/logmodels2.pdf). Additionally, quantile regression analyses (PROC QUANTREG) were carried out for 0.10 to 0.90 percentiles to model the associations of the predictors ENL and GEN on the conditional percentiles of Ki-67. Considering a more clinical point of view, a binary logistic regression analysis using Ki-67 expression dichotomized (≥20% (high) versus <20% (low) 31, 32 was also performed (PROC LOGISTIC) to test possible relationships with PE metabolites (continuous per 10 nmol/L increment and in quartiles). The associations between ENL and GEN (continuous per 10 nmol/L increment and in quartiles) with HER2 status were estimated using binary logistic regression (HER2+ versus HER2-) to estimate odds ratios (ORs) and 95%CIs.
The following were tested as potential confounders in regression models: age at time of diagnosis (years), education level (low/medium/high), BMI (kg/m²), smoking status (never or former/current), alcohol consumption (g/day), leisure time physical activity (LTPA) aged ≥ 50 years (Metabolic equivalent (MET) hours per week (hrs/w)), and hormone replacement therapy (HRT) use at diagnosis (never/past/current (≤ 6 months)). Two different timing variables for blood draw – time of blood draw with respect to chemotherapy (no chemotherapy or blood draw before chemotherapy/blood draw during chemotherapy or ≤ 3 months after chemotherapy/ > 3 months after chemotherapy) and time between surgery (when tumor tissue was taken) and blood draw (in days) – were included as covariates to account for possible influences of chemotherapy on ENL and GEN concentrations 33 and to account for changes in dietary patterns following recovery after surgery 34. ER status (ER+ versus ER-), and PR status (PR+ versus PR-)) were included as covariates to account for hormone-related effects of PEs 2, 4, 5. In contrast, tumor characteristics such as tumor size, grade, nodal status and metastases were not included in the models, as Ki-67 and HER2 overexpression are associated with increased tumor growth, enhanced invasive aggressiveness, and development of metastases and thus strongly correlated with tumor characteristics 11, 35.
Covariates were selected on the basis of a priori assumptions and considered confounders in the associations between PE metabolites with both Ki-67 expression (linear) and HER2 status (logistic regression), separately, if they changed the ß estimates and ORs, respectively, by at least 10% using manual forward selection. The same covariates used in the linear model were also used in the quantile and logistic models with Ki-67 as the dependent variable.
All statistical analyses were performed using the statistical software package SAS, version 9.3 (SAS Institute, Cary, North Carolina, U.S.). For all analyses, two-sided p-values <0.05 were considered significant.
3 Results
Median age at diagnosis of an incident breast cancer for 1060 patients was 63.8 years (interquartile range (IQR) 7.9). Out of a total of 1060 tumors, 21.2% were HER2+, 83.2% were ER+, and 73.4% were PR+. Median Ki-67 expression in 803 patients was 12.0% (IQR 19.0, range 0.0-98.0). The distribution was slightly right skewed and therefore log-transformed. Patients with HER2- tumors and low Ki-67 expression tumors (<20%) were more likely to have ER+, PR+, and low or moderate grade tumors compared to patients with HER2+ tumors or higher Ki-67 expression (≥20%). As expected, low Ki-67 expression was associated with smaller tumor size, no affected lymph nodes, and HER2- tumors (Supporting Information Table 1).
3.1 Serum PE concentrations
The mean concentrations of serum ENL and GEN were 32.2 nmol/L (median 19.5, IQR 33.4, 0.0-458.7) and 19.2 nmol/L (median 4.8, IQR 10.4, 0.1-527.9), respectively. The distribution of both PE concentrations was slightly right skewed. Higher mean ENL concentrations were significantly associated with older age, higher education, underweight or normal weight (BMI<25kg/m²), never/former smokers, current HRT use (≤6 months), and no chemotherapy or blood drawn before chemotherapy as well as smaller tumor size (≤2cm) (all p<0.05) (Table 1). GEN concentrations were solely significantly related to current HRT use.
| Characteristics | ENL (nmol/L) | GEN (nmol/L) | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| n | Median | Mean | IQR | pa | Median | Mean | IQR | pa | ||
| Total | 1058/1050 | 19.5 | 32.2 | 33.4 | 4.8 | 19.2 | 10.4 | |||
| Age at diagnosis (years) | 50–54 | 74 | 13.3 | 24.2 | 23.7 | <0.01 | 4.1 | 27.3 | 6.6 | 0.12 |
| 55–59 | 204 | 16.1 | 25.1 | 28.0 | 4.4 | 24.2 | 10.6 | |||
| 60–64 | 338 | 22.8 | 34.1 | 35.0 | 5.2 | 16.4 | 10.5 | |||
| 65–69 | 295 | 21.6 | 33.0 | 37.0 | 4.2 | 14.4 | 10.6 | |||
| 70–75 | 149 | 24.3 | 39.8 | 35.0 | 5.7 | 23.7 | 10.9 | |||
| Ki-67 expression (%) | < 20 | 541 | 20.1 | 33.4 | 33.5 | < 0.05 | 5.0 | 20.3 | 10.2 | < 0.01 |
| ≥ 20 | 262 | 16.9 | 27.3 | 32.1 | 4.0 | 10.8 | 8.5 | |||
| HER2 status | HER2- | 820 | 20.1 | 32.6 | 34.4 | 0.55 | 4.8 | 19.9 | 11.1 | 0.50 |
| HER2+ | 220 | 17.3 | 30.7 | 30.0 | 4.2 | 17.0 | 9.0 | |||
| ER status | ER- | 178 | 16.4 | 27.5 | 34.1 | 0.10 | 4.2 | 16.6 | 9.0 | 0.50 |
| ER+ | 882 | 20.1 | 33.1 | 33.1 | 4.8 | 19.7 | 10.6 | |||
| PR status | PR- | 282 | 17.3 | 29.0 | 35.8 | 0.13 | 4.2 | 14.8 | 8.8 | 0.12 |
| PR+ | 778 | 21.1 | 33.3 | 32.4 | 5.0 | 20.8 | 11.1 | |||
| Tumor size (cm) | ≤ 2 | 595 | 22.5 | 34.8 | 36.2 | < 0.05 | 5.0 | 21.2 | 11.9 | 0.23 |
| > 2-5 | 385 | 18.8 | 29.2 | 31.8 | 4.4 | 17.4 | 8.2 | |||
| > 5/chest wall | 77 | 15.9 | 26.2 | 24.7 | 5.1 | 10.6 | 10.3 | |||
| Affected lymph nodes at diagnosis | 0 | 694 | 21.1 | 33.9 | 35.2 | 0.16 | 4.9 | 21.3 | 11.0 | 0.15 |
| 1–3 | 241 | 17.0 | 29.6 | 33.0 | 4.4 | 16.7 | 9.3 | |||
| ≥ 4 | 123 | 17.6 | 27.8 | 28.2 | 4.2 | 11.6 | 10.6 | |||
| Metastases at diagnosis | No | 1017 | 19.4 | 32.2 | 33.4 | 0.99 | 4.8 | 19.6 | 10.5 | 0.20 |
| Yes | 43 | 24.0 | 32.3 | 35.9 | 3.3 | 8.6 | 10.6 | |||
| Grade | Low/moderate | 761 | 21.3 | 33.1 | 34.2 | 0.13 | 4.8 | 20.3 | 10.6 | 0.26 |
| High | 294 | 17.3 | 28.9 | 30.0 | 4.7 | 16.0 | 10.1 | |||
| Educational level | Low | 694 | 17.6 | 29.4 | 31.4 | <0.01 | 4.4 | 18.7 | 9.6 | 0.72 |
| Medium | 217 | 24.3 | 34.4 | 34.0 | 5.2 | 18.4 | 11.1 | |||
| High | 149 | 26.0 | 41.6 | 35.7 | 5.5 | 22.6 | 14.6 | |||
| BMI (kg/m²) | < 25 | 438 | 23.9 | 35.6 | 34.0 | 0.02 | 5.3 | 21.4 | 10.4 | 0.36 |
| 25-<30 | 411 | 21.4 | 31.2 | 32.3 | 4.6 | 19.2 | 11.1 | |||
| ≥ 30 | 208 | 10.9 | 26.2 | 25.4 | 3.8 | 14.7 | 9.7 | |||
| Smoking status | Never/former | 915 | 21.4 | 33.7 | 34.6 | <0.01 | 4.8 | 18.9 | 10.6 | 0.76 |
| Current | 145 | 15.0 | 22.4 | 27.6 | 4.2 | 20.5 | 9.2 | |||
| LTPA (MET hrs/w) | < 28 | 384 | 18.1 | 31.1 | 33.3 | 0.52 | 4.8 | 21.1 | 10.7 | 0.41 |
| ≥ 28 | 672 | 20.1 | 32.8 | 33.4 | 4.7 | 18.1 | 10.4 | |||
| Alcohol (g/day) = | 0-< 0.5 | 305 | 17.6 | 34.1 | 32.5 | 0.50 | 4.9 | 21.2 | 10.9 | 0.53 |
| 0.5-< 6 | 366 | 19.4 | 32.0 | 31.4 | 4.5 | 17.4 | 9.3 | |||
| 6-< 12 | 151 | 21.4 | 33.8 | 39.0 | 8.3 | 23.5 | 18.7 | |||
| ≥ 12 | 238 | 22.2 | 28.9 | 33.6 | 4.0 | 16.5 | 7.6 | |||
| HRT at diagnosis | Never/past | 628 | 17.6 | 29.4 | 31.3 | 0.02 | 4.7 | 15.7 | 9.9 | 0.02 |
| Current (≤ 6 months) | 422 | 23.0 | 35.6 | 33.7 | 4.8 | 24.2 | 11.7 | |||
| Time of blood draw and chemotherapy | No/before CT | 793 | 22.4 | 34.2 | 33.9 | 0.01 | 4.8 | 19.8 | 10.6 | 0.83 |
| During/≤ 3 months after | 144 | 11.2 | 23.4 | 25.1 | 4.3 | 18.7 | 9.6 | |||
| > 3 months after CT | 115 | 18.6 | 30.4 | 36.1 | 4.7 | 16.2 | 9.4 | |||
| Time OP and blood draw | ≤ 3 months after OP | 564 | 19.3 | 30.9 | 32.7 | 0.27 | 4.7 | 18.5 | 10.0 | 0.69 |
| > 3 months after OP | 496 | 20.0 | 33.7 | 34.0 | 4.8 | 19.9 | 11.2 | |||
- a Differences in mean ENL and GEN concentrations by characteristics of the population were tested by ANOVA (p<0.05 for statistical significance).
Mean PE concentrations did not differ between patients with HER2- (ENL 32.6; GEN 19.9) and HER2+ tumors (ENL 30.7; GEN 17.0). On the other hand, mean ENL and GEN concentrations were significantly lower in patients with higher Ki-67 expression tumors (≥20%) (ENL 27.3; GEN 10.8) than in patients with lower Ki-67 expression (<20%) (ENL 33.4; GEN 20.3) (p<0.05 and p<0.01, respectively) (Table 1).
3.2 Associations between serum PE concentrations and Ki-67 expression
Using a log-linear model, after adjustment for age at diagnosis, timing of blood draw with respect to chemotherapy, time between operation and blood draw, HRT use at diagnosis, ER status, and PR status, there were weak inverse associations between ENL and GEN with Ki-67 expression (ß of -0.8% [-2.5%; 0.8%] per 10 nmol/L increments of ENL, p = 0.31; ß of -1.3% [-2.7%; 0.2%] per 10 nmol/L increments of GEN, p = 0.08), although not significant (Table 2). The explained variance of the variables in both models was 0.25 (adjusted R-square).
| Ki-67 (log-transformed)a | Ki-67 (≥20% versus <20%)b | HER2 status (HER2+ versus HER2-)c | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Exposure | n | ßd | [95%CI] | pcate | ptrende | n | OR | [95%CI] | pcate | ptrende | n | OR | [95%CI] | pcate | ptrende |
| ENL per 10 nmol/L increment | 786 | −0.008 | [−0.025; 0.008] | 0.31 | 785 | 0.97 | [0.93; 1.02] | 0.30 | 1030 | 1.00 | [0.96; 1.04] | 0.99 | |||
| ENL in quartiles | 786 | 785 | 1030 | ||||||||||||
| Q1 (0.0-≤7.5) | Ref. | Ref. (1.0) | Ref. (1.0) | ||||||||||||
| Q2 (>7.5-≤19.5) | 0.173 | [−0.001; 0.358] | 0.07 | 0.99 | [0.61; 1.60] | 0.74 | 1.34 | [0.88; 2.04] | 0.10 | ||||||
| Q3 (>19.5-≤41.2) | 0.136 | [−0.053; 0.325] | 0.16 | 0.93 | [0.57; 1.52] | 0.94 | 1.03 | [0.66; 1.62] | 0.78 | ||||||
| Q4 (>41.2-458.7) | −0.050 | [−0.239; 0.139] | 0.61 | 0.85 | [0.51; 1.40] | 0.52 | 0.97 | [0.62; 1.51] | 0.44 | ||||||
| GEN per 10 nmol/L increment | 779 | −0.013 | [−0.027; 0.002] | 0.08 | 778 | 0.93 | [0.87; 0.99] | 0.02 | 1023 | 0.99 | [0.96; 1.02] | 0.62 | |||
| GEN in quartiles | 779 | 778 | 1023 | ||||||||||||
| Q1 (0.1-≤ 2.1) | Ref. | Ref. (1.0) | Ref. (1.0) | ||||||||||||
| Q2 (>2.1-≤4.7) | 0.040 | [−0.144; 0.224] | 0.67 | 1.05 | [0.65; 1.70] | 0.28 | 1.11 | [0.72; 1.70] | 0.34 | ||||||
| Q3 (>4.7-≤12.4) | 0.037 | [−0.146; 0.220] | 0.69 | 0.79 | [0.49; 1.28] | 0.42 | 1.03 | [0.67; 1.59] | 0.68 | ||||||
| Q4 (>12.4-527.9) | −0.018 | [−0.206; 0.170] | 0.85 | 0.77 | [0.47; 1.27] | 0.35 | 0.79 | [0.51; 1.25] | 0.15 | ||||||
- a Linear regression models between ENL and GEN concentrations (continuous or in quartiles) separately with Ki-67 as dependent variable (log-transformed) to calculate ß estimates and 95%CIs; all adjusted for age at diagnosis, blood draw during chemotherapy, time between operation & blood draw, HRT use at diagnosis, ER status and PR status.
- b Logistic regression models between ENL and GEN concentrations (continuous or in quartiles) separately with Ki-67 as 2 categories (≥20% versus <20%) to calculate odds ratios (ORs) and 95%CIs; all adjusted for age at diagnosis, blood draw during chemotherapy, time between operation & blood draw, HRT use at diagnosis, ER status and PR status.
- c Logistic regression models between ENL and GEN concentrations (continuous or in quartiles) separately with HER2 (+ versus -) to calculate odds ratios (ORs) and 95%CIs; adjusted for age at diagnosis, blood draw during chemotherapy, time between operation & blood draw, ER status, PR status.
- d ß 100(eß -1) = change in %.
- e p cat (p categorical), differences between the groups; p trend (p for linear trend).
When ENL and GEN were assessed as quartiles a decrease in Ki-67 was apparent only for the highest quartile (Q4) compared to the lowest quartile (Q1), where Ki-67 was reduced by 4.9% with ENL (p categorical (cat) p = 0.61, p trend = 0.31) and by 1.8% with GEN (p cat = 0.85, p trend = 0.08) (Table 2), although neither of these reached statistical significance.
When the association with PEs was assessed for Ki-67 in quantiles, the reduction in Ki-67 per 10 nmol/L increment of ENL and GEN became increasingly stronger in the upper percentiles of Ki-67. As of the 60th Ki-67 percentile, significant reductions in Ki-67 expression of 14.7% [4.6–24.9%] per 10 nmol/L GEN increment (p < 0.01) increased to 38.2% [4.9–81.3%] for the 90th percentile (p = 0.08) (Table 3).
| Ki-67 Quantile | Ki-67 Mean | ß | [95%CI] | p |
|---|---|---|---|---|
| ENL (per 10 nmol/L increment) | ||||
| 0.1 | 2.0 | −0.005 | [−0.093; 0.084] | 0.92 |
| 0.2 | 2.4 | −0.001 | [−0.119; 0.118] | 0.99 |
| 0.3 | 4.7 | −0.013 | [−0.142; 0.116] | 0.84 |
| 0.4 | 10.3 | −0.059 | [−0.209; 0.090] | 0.44 |
| 0.5 | 11.9 | −0.084 | [−0.301; 0.132] | 0.44 |
| 0.6 | 13.1 | 0.039 | [−0.194; 0.272] | 0.74 |
| 0.7 | 16.4 | 0.007 | [−0.225; 0.238] | 0.96 |
| 0.8 | 24.3 | −0.118 | [−0.426; 0.190] | 0.45 |
| 0.9 | 29.6 | −0.313 | [−0.715; 0.090] | 0.13 |
| GEN (per 10 nmol/L increment) | ||||
| 0.1 | 1.9 | −0.035 | [−0.092; 0.021] | 0.22 |
| 0.2 | 2.3 | −0.058 | [−0.148; 0.031] | 0.20 |
| 0.3 | 4.8 | −0.049 | [−0.147; 0.049] | 0.33 |
| 0.4 | 9.8 | −0.084 | [−0.180; 0.013] | 0.09 |
| 0.5 | 11.0 | −0.062 | [−0.157; 0.033] | 0.20 |
| 0.6 | 13.6 | −0.147 | [−0.249;−0.046] | <0.01 |
| 0.7 | 16.4 | −0.147 | [−0.369; 0.075] | 0.19 |
| 0.8 | 23.6 | −0.280 | [−0.570; 0.010] | 0.06 |
| 0.9 | 30.6 | −0.382 | [−0.813; 0.049] | 0.08 |
- a Quantile regression models between ENL and GEN concentrations (per 10 nmol/L increment) separately with Ki–67 in percentiles to calculate ß estimates and 95%CIs; adjusted for age at diagnosis, blood draw during chemotherapy, time between operation & blood draw, HRT use at diagnosis, ER status and PR status.
When Ki-67 was used as a clinically relevant binary variable, there was a significant inverse association between GEN per 10 nmol/L increment and Ki-67 expression (Table 2). In other words, higher GEN concentrations were related to lower expression Ki-67 in tumors showing ≥20% Ki-67 expression.
3.3 Associations between serum PE concentrations and HER2 status
There were no associations between ENL or GEN (as continuous variables and as quartiles) and HER2 status (Table 2). The highest versus the lowest quartile for GEN was associated with a slightly decreased chance to have a HER2+ classified tumor, although not statistically significant (per 10 nmol/L GEN increment OR 0.79 [0.51; 1.25]). A similar trend was observed for ENL in quartiles and HER2 status following adjustment (Q4 versus Q1: OR 0.97 [0.62; 1.51]).
4 Discussion
In the current analysis of postmenopausal patients with invasive breast cancer, inverse associations between GEN and Ki-67 expression were found starting at the 60th percentile of Ki-67 expression, whereby the significant reduction in Ki-67 expression per 10 nmol/L GEN increment steadily increased to the 90th percentile of Ki-67 expression, although no longer significant in the higher percentiles. Dichotomizing at the clinically relevant cutoff of 20% Ki-67 expression, we found that higher GEN concentrations were significantly associated with lower Ki-67 expression in tumors showing ≥20% Ki-67 expression. We did not find clear evidence of an overall relationship between concentrations of the PE metabolites ENL and GEN with tumor HER2 expression or between ENL with Ki-67 expression after adjustment for relevant covariates.
However, there are limitations to this study – blood samples were collected after tumor development and thus are not prediagnostic concentrations of ENL and GEN, and chemotherapy can influence intestinal metabolism of PEs 33. To account for this, we included the time between operation and blood draw and time between chemotherapy and blood draw in all statistical models. We also performed several sensitivity analyses to ensure that the results obtained were not influenced by timing of blood draw. Blood samples for measuring biomarkers were available from only one single time point. Repeated measurements would give better indications of long-term exposure 36. Furthermore, PE metabolites could have been influenced by disease status and potential changes in diet after diagnosis. Antibiotic intake was not recorded, which could also influence PE bioavailability 37. As with all epidemiological studies, residual confounding due to unmeasured covariates cannot be excluded. Determination of Ki-67 expression was predominantly based on TMAs, which may have led to some misclassification 32. However, this is unlikely to have led to bias in the results since it was independent of PE measurement. Ten percent of Ki-67 values were derived from pathology; therefore we checked to ensure that source of Ki-67 was not a confounder. When we performed sensitivity analyses excluding the patients whose Ki-67 values were derived from local pathology reports, the results for linear, logistic and quantile regression also did not change. Finally, the cross-sectional analysis does not allow for causal inferences.
This study also possesses several strengths – the large sample size, comprehensive clinical data, and the use of serum PE metabolites rather than dietary assessment tools for exposure assessment. Biomarker measurements provide a more precise assessment and reflect individual variation in metabolism, absorption, and bioavailability of PEs in the gut microflora in contrast to estimated dietary intake 38 as well as allowing for the avoidance of recall bias and misclassification. ENL and GEN were measured using TR-FIA, a validated immunoassay which was shown to have good sensitivity and allows for rapid, high-throughput measurement 3, 38. In addition, all tumor blocks for TMAs were evaluated by one pathologist to ensure consistency and IHC markers were scored by two pathologists. Patients with neo-adjuvant chemotherapy were excluded in this analysis, as neo-adjuvant chemotherapy could influence tumor characteristics. To our knowledge, the influence of estimated PE intake or PE metabolites on Ki-67 and HER2 expression in human observational studies have not yet been investigated while concurrently accounting for hormonal factors such as ER status, PR status, and HRT use.
Our finding of an inverse association between GEN at higher Ki-67 expression levels conflict with associations reported in several previous studies. Results from three RCTs have shown no effect of daily isoflavone supplementation on Ki-67 expression in premenopausal breast cancer patients 15, 16, postmenopausal breast cancer patients 15, and healthy women 17. In one RCT with 140 premenopausal breast cancer patients, there was no difference in tumor Ki-67 expression between those supplemented daily with 51.6g soy protein (n = 70) for 7–30 days and those taking a placebo (n = 70) 15. Likewise, there was no difference in Ki-67 expression in histologically normal breast epithelium of 19 premenopausal women supplemented daily with 45mg isoflavones for 14 days and 29 premenopausal women in the placebo arm 16, nor was there a difference in Ki-67 expression in high risk breast tissue of 98 healthy women supplemented with 150 mg of GEN, 74 mg DAI, and 11 mg glycitein daily for 6 months (n = 49) compared with those who received placebo (n = 49) 17.
The differences between our results and those from RCTs could be attributed to several reasons. In these RCTs, participants were supplemented with supraphysiological doses of soy protein, isoflavones, GEN, DAI, and glycitein for a short period of time, which could affect Ki-67 expression differently as compared to physiological concentrations as measured in serum in patients not receiving any form of dietary intervention. We have previously reported median intakes of FS and sesame in breast cancer patients of the MARIE baseline study to be 1.42 g per day, for soy milk 12.3 g, soybeans 0.41 g and tofu 1.64 g 28, which are much lower compared to those in intervention studies 14. Disparities in Ki-67 expression assessment, and menopausal status of participants could further account for the discrepancies in results between the current analysis and those of the RCTs as well as an epidemiological study of 268 pre- and postmenopausal breast cancer patients using data from a cohort and a nested case-control study 19. In the observational study, rather than assessing metabolites of isoflavones in blood, a questionnaire was used to ascertain soy intake (≥1 versus <1 serving/week) consumed both in early life (<20 years) and in adulthood (≥20 years), and soy intake was not found to be associated with Ki-67 expression in benign and malignant breast tissue even after adjustment for ethnicity, age, BMI, and HRT use 19. Last, but highly pertinent given our results, the type of (tumor) tissue assessed (normal epithelium versus high risk breast tissue versus tumor tissue), tumor characteristics, and timing of exposure to PEs would be responsible for differential results between the studies. We observed an inverse association between GEN concentrations and Ki-67 expression but only at higher levels of Ki-67 expression. As such, our results are in line with those from RCTs that investigated Ki-67 expression in normal breast epithelium 16 or breast tissue of healthy women 17, given that Ki-67 expression in normal breast epithelium of healthy women has been demonstrated to be very low (<3%) 39, 40.
In contrast to the aforementioned human studies, animal studies have reported consistent results 22, 23. Daily treatment with GEN (700 μg/g/day) 23 or soy extract (100 mg/kg/day) 22 significantly down regulated Ki-67 expression in juvenile rats 23, and in mammary glands of ovariectomized mice in epithelial compartments 22, compared to controls. Our result of an inverse association between GEN and Ki-67 expression at higher Ki-67 expression levels is in agreement with these animal studies. Thus our study contributes noteworthy information to the current body of evidence by providing a snapshot of serum ENL and GEN concentrations and Ki-67 and HER2 expression in breast cancer tumors. We have found that the impact of GEN on tumors that are not highly proliferative is small but when tumors are highly proliferative as indicated by high Ki-67 expression, GEN is associated with a decrease in tumor proliferation in a non-estrogenic manner. Data from large long-term longitudinal studies would be essential for replication of our findings and to further explore in greater detail the etiological mechanisms linking GEN to Ki-67 in breast cancer patients.
We did not observe an association between serum ENL and Ki-67 expression, and our finding of a null association is in concordance with an intervention study 14 but not with others 13. A RCT conducted in 32 postmenopausal patients with newly diagnosed breast cancer reported no significant change in Ki-67 expression during the pre- and posttreatment period in the FS group, who received a 25g FS muffin daily (n = 19) compared to the placebo group (n = 13) with mean treatment lengths of 32 days and 39 days, respectively 14. In another study, however, among premenopausal women, lignan supplementation with 50 mg secoisolariciresinol diglycoside (SDG) daily for 12 months was associated with significantly decreased Ki-67 expression in hyperplastic benign breast tissue of 80% of the 45 patients who completed the trial from a median of 4% (IQR 2.0–16.8 %) to 2% (0–15.2%), although there was no control group for comparison 13. Our findings are similar to those in the RCT with postmenopausal women, but conflict with findings presented in the study with premenopausal women. These differences could be attributed to most crucially, menopausal status of the study populations, and also the different tissues in which Ki-67 was assessed, and different treatment regimes. Given the divergent findings in the few epidemiological studies that have been conducted 13, 14 and significant effects of FS and SDG on Ki-67 in MCF-7 breast cancer cell studies 20, 21 on Ki-67 expression, further studies would be worthwhile.
We did not find an association between serum ENL with HER2 status, which is in agreement with some epidemiological studies 18 but not with others 14. Total dietary lignan intake and HER2 status was not found to be associated in a case-control-study of 683 postmenopausal breast cancer cases and 611 controls 18. In contrast, HER2 expression was lower following daily supplementation with FS (25g/day) compared to the placebo group in a previously mentioned RCT of 32 postmenopausal breast cancer patients 14. In experimental studies, SDG and lignan sesamin (1g/kg for 8 weeks) were shown to reduce HER2 expression in human MCF-7 breast cancer tumors in athymic mice 21.
We also did not observe an association between serum GEN with HER2 status. A previous study in humans found that women with high soy intake in adulthood (≥ 1 serving/week) were less likely to have HER2 expression in malignant tissue compared to those with low soy intake (OR 0.46 [0.23–0.92], p = 0.03) 19. Experimental studies of human breast cell lines in mice have demonstrated that high doses of GEN (at ≥ 1μM 27, at 25μM 25, ≥ 50 μM 26) can significantly inhibit HER2 expression. Overall, inconclusive results between those in the current analysis, a human intervention study 14 and experimental studies 21, 25-27 may be due to the comparatively low serum ENL and GEN concentrations in our study patients (median 19.5 and 4.8 nmol/L, respectively) as well as differences in exposure and outcome assessment, type of (tumor) tissue assessed, and tumor characteristics. There is a paucity of data, particularly from epidemiological studies, on PEs and HER2 expression. Drawing conclusions from the current insufficient body of evidence would not be possible due to the heterogeneity of studies and a general lack of information 14, 18, 19, 21, 25-27. In light of this and the identification of HER2 as a known prognostic marker of aggressive tumor behavior occurring in approximately 20–30% of breast cancers 11, large long-term longitudinal studies to evaluate the associations of both lignans and isoflavones on HER2 status would be useful in improving our knowledge about these relationships.
In conclusion, the results of the present study indicate an inverse association between serum GEN and Ki-67 expression when Ki-67 is highly expressed in tumors of postmenopausal breast cancer patients even after adjusting for hormone-related and other relevant factors. No associations between serum ENL and GEN concentrations and HER2 expression were observed. Future research should be aimed at implementing large long-term longitudinal studies to confirm our findings and to additionally explore possible biological hormone-independent and hormone-dependent links between ENL/GEN and breast cancer survival.
Acknowledgments
The authors’ responsibilities were as follows: J.C-C. conceived and designed the MARIE study; S.J., A.Y.J. and J.C-C. conceived and designed the manuscript; S.J., A.Y.J., A.R. and J.C-C. contributed to the literature search, analysis and interpretation of data; S.J., A.Y.J. and J.C-C. prepared the manuscript; S.J. drafted the first version of the manuscript; P.S. evaluated all tumor blocks for TMAs and scored the IHC markers; E.H. oversaw TMA construction and provided data from the IHC measurements; K.T. was responsible for the MARIE study center in Hamburg; T.J. supervised the lab work; and all authors approved the final manuscript.
We are grateful to all the MARIE study participants for their contribution and the interviewers who collected the data. We thank U. Eilber and S. Behrens for data management and coordination, D. Sookthai for statistical support, S. Becker, S. Henke, B. Kaspereit, B. Lederer and K. Zaineddin for sample preparation and sample coordination T. Ilola (Labmaster, Finland) for biomarker measurements as well as the NCT tissue bank for providing the TMAs. The MARIE study was funded by the Deutsche Krebshilfe e.V. (grant number 70-2892-BR I) and the German Cancer Research Center.
The authors have declared no conflict of interest.
