HER2 Phenotyping of CTCs By Peptide-Functionalized Nanoparticles As The Diagnostic Biomarkers of Anti-HER2 Treatment in Breast Cancer


 Background: The efficacy of anti-human epidermal growth factor receptor 2 (HER2/neu) treatment is impacted by tissue-based evaluation bias due to tumor heterogeneity and dynamic changes of HER2 in breast cancer. Circulating tumor cell (CTC)-based HER2 phenotyping provides an integral and real-time assessment, benefiting accurate HER2 diagnosis for improved therapeutics.Methods: The study was exploratory. Fifty-four breast cancer patients, including 26 histopathologic HER2/neu-positive (Histo-HER2/neu+) and 28 HER2/neu-negative (Histo-HER2/neu+), were enrolled for blood draws before a new line of treatment. We used peptide-functionalized magnetic nanoparticles to isolate CTCs from 2.0 mL whole blood and determined HER2 phenotypes of the enriched CTCs by immunocytochemistry (ICC). We investigated the correlation of the enumeration and HER2 phenotyping on baseline CTCs with the diagnosis, the prognosis, and the efficacy of the trastuzumab in combination with chemotherapeutic agents of breast cancer. We also explored the dynamic change of HER2 phenotypes on CTCs after the combination therapy in a cohort of Histo-HER2/neu+ individuals (n = 14).Results: We achieved high-efficient detection of the CTC-positive cases (≥ 3 CTCs) (38/54, 70.4%), from whom 71.1% (27/38) had a concordant HER2 status on CTCs and tumor tissues at baseline. Pretheraputic CTC enumeration showed a significant correlation with the histopathologic diagnosis (e.g. estrogen receptor/progesterone receptor (ER/PR) status or proliferation of cancer cells) and the progression-free survival/overall survival (PFS/OS) of breast cancer. However, it was not a practicable biomarker to predict the efficacy of the trastuzumab-based combination therapy. In contrast, we demonstrated a significantly higher possibility of good overall responses in the patients with < 3 CTCs (P = 0.006) or with HER2 overexpression in CTCs (CTC-HER2+) at baseline, as compared to the individuals without HER2 overexpression in CTCs (CTC-HER2–) (P = 0.028).Conclusions: We demonstrate the significance of the peptide-functionalized magnetic nanoparticle in noninvasive detection and HER2 phenotyping of CTCs, and highlight its diagnostic and prognostic values for the patients about to start the anti-HER2 treatment.

Nevertheless, conventional sampling of tumor tissues through surgery or puncture is constrained by complicated manipulation and harmful invasion, therefore not feasible in the repeated, real-time biopsies for a comprehensive tumor portrait (e.g. HER2) [15,16]. Moreover, the HER2/neu analysis by tissue biopsies might produce uninterpretable results due to insufficient tumor cells or strong background [8]. Therefore, it is necessary to achieve more sensitive and real-time biomarkers for the HER2 diagnosis of breast cancer.
A noninvasive blood-based detection using the circulating tumor cells (CTCs) that have similar genetic and epigenetic properties to the primary/metastatic tumor cells is blooming to be a surrogate to the histopathologic assessments [17]. Multiple investigations have been implemented to explore the clinical significances of CTCs in the early diagnosis, prognostication, and mornitoring of solid tumors [18][19][20][21][22]. CTC enumeration has been a valid and independent staging and prognostic biomarker for breast cancer but not for predicting the survival of the HER2-overexpressed patients during anti-HER2 treatment [23][24][25]. Exploring the gene/protein portraits of breast cancer CTCs should becomplementary to the personalized diagnoses and treatments. HER2/neu overexpression or amplification on CTCs has been investigated in breast cancer over the last decade. The discordance of HER2/neu status on tumor tissues and CTCs was present (38% -86%) [26][27][28][29]. Besides, Jordan et al. demonstrated a spontaneous interconversion of HER2 phenotypes on CTCs ex vivo [30]. Hence, HER2 reassessments on CTCs might overcome the spatio-temporal heterogeneity of HER2 expression in breast cancer. Nonetheless, up to date, there is no widely accepted consensus about the criteria of HER2 phenotyping of enriched CTCs, primarily because various tests and definitions of HER2 status were adopted by different trials in a small sample size. On the other hand, as the only U. S. Food and Drug Administration (FDA)-approved detection system of epithelial CTCs in solid tumors, CellSearch was extensively used in various studies, whereas it is inevitably weak in the limited rate of CTC detection from breast cancer patients (22% -70%) [31][32][33][34], leading to a portion of patients that are unqualified in the HER2 analysis by CellSearch. Therefore, it remains necessary to explore the clinical potentials of HER2 phenotyping on CTCs in breast cancer on the basis of a more sensitive detection of CTCs and HER2 proteins.
Previously, our group has developed an innovative EpCAM-targeted peptide (Pep10)-functionalized magnetic nanoparticle (Pep10@MNPs) that can be used to efficiently capture the rare epithelial CTCs in 2.0 mL whole blood (WB) from patients with various solid tumors [35][36][37]. In the present study, the Pep10@MNPs and classical four-color immunocytochemistry (ICC) were applied to detect EpCAM-overexpressed CTCs and to assess HER2 status of the enriched CTCs. We achieved the analysis of the blood samples from 54 breast cancer patients at baseline and from 15 cases of the Histo-HER2/neu + individuals with the second blood draws after anti-HER2 therapy in combination with chemotherapy. Thereafter, we consolidated the diagnostic and prognostic values of CTC enumeration with the determined cutoff that was published in our previous report [36]. Furthermore, the evaluation of HER2 overexpression on CTCs was achieved based on the fluorescent quantification of HER2 immunostaining.
We emphatically analyzed the relationship of the HER2 phenotypes of baseline CTCs or the evolutionary HER2 status in CTCs with the efficacy of the HER2-based combination treatment, which may further help understand the prognostication values of HER2 status of the enriched CTCs in guiding anti-HER2 treatment.

Study design
We were approved by the institutional review board at the Medical Ethical Committee

Isolation of breast cancer cells by Pep10@MNPs in simple cell models
The streptavidin-modified magnetic nanoparticles (SA-MNPs, maghemite γ-Fe2O3) (Ademtech, Gironde, PESSAC, France) and the biotin-labeled EpCAM-targeted peptides (Pep10) (Guoping Pharmaceutical Co., Anhui, China) were purchased. The Pep10@MNPs were constructed as previously described [38]. Briefly, with the volume ratio (v/v) of 1:2, the SA-MNPs (5 mg/mL) were incubated with Pep10 (1 mg/mL) for 60 min at 37 ℃, followed by three-time washing. As a control, the antibody@MNPs were prepared with anti-EpCAM (Abcam, Cambridge, MA, UK) using a similar protocol. The constructed nanoparticles were confirmed by dynamic light scattering (DLS) (Zetasizer Nano ZS90, UK) and transmission electron microscope (TEM) (Tecnai G2 20 S-TWIN, USA). In simple cell models, 1000 cells that were pre-stained with 4',6-diamidino-2-phenylindole dihydrochloride (DAPI) suspended in 1.0 mL PBS and incubated with the Pep10/anti-EpCAM@MNPs for 60 min at 37 ℃. The Olympus IX73 fluorescent microscope (Olympus, Tokyo, Japan) was used to enumerate the added or captured cells. The capture efficiency (%) was determined as the number of the captured cancer cells/the number of the added cancer cells.

HER2 immunostaining on the breast cancer cells in cancer cell models
Cancer cell models were constructed with 1000 breast cancer cells in 1.0 mL PBS (simple, System 1) or in 1.0 mL PBS containing 10 6 monocyte-like U937 cells (complex, System 2). The breast cancer cells were pre-stained with DAPI (nucleic acid dye)/DIO (plasma membrane dye). The U937 cells were only pre-stained with DAPI.
Pep10@MNPs were incubated with System 1/2 for 60 min at 37 ℃, followed by magnetic enrichment. After the fixation with 2% paraformaldehyde (PFA) for 30 min, the captured cells were blocked with 5% BSA solution. HER2 immunostaining of the captured cells was then achieved with Alexa Fluor 647-labeled anti-HER2 for 60 min (BioLegend, San Diego, CA, USA). The Olympus IX73 fluorescent microscope (Olympus, Tokyo, Japan) was used to quantify the mean fluorescent intensities of HER2 immunostaining from 50 captured cells.

Patients and clinical blood samples
The eligible breast cancer patients (N = 54) for baseline blood draws were newly diagnosed locally advanced or recurrent/metastatic individuals and were about to accept new-line treatments (chemotherapy alone or in combination with anti-HER2 or other targeted therapies). The patients with secondary malignancies within 5 years, or already in systematic therapies within 7 days, or painful to blood draw were all excluded. As a negative control, 10 healthy donors were enrolled who were evaluated as no medical history of any serious disease and any surgery within 6 months. The corresponding clinicopathological and radiological evidence of the 54 enrolled patients was documented in Table 1 Totally, we collected 54 baseline blood samples and 14 follow-up blood samples from breast cancer patients, and 10 blood samples from healthy donors. For each blood sample, whole blood was drawn into an anticoagulation-contained plastic vacuum blood collection tube (NANOPEP BIOTECH CO., Beijing, China), and delivered to the lab for Pep10@MNPs-based detection of CTCs within 24 h.

Capture and HER2 immunostaining of CTCs in clinical blood samples
Briefly, 10 μL Pep10@MNPs were incubated with 2.0 mL whole blood for 60 min at or Chi-square test was useful to figure out the relationships among various diagnostic paramters (P < 0.05 as a significant correlation). Chi-square test was used when the sample size was large (n ≥ 40) and the minimum of the observed frequencies (Tmin) was larger than five; otherwise, the Fisher's exact test was applied [32].

Isolation and identification of breast cancer cells
We first functionalized the streptavidin-modified magnetic nanoparticles (SA- Considering the compelling advantages of peptides (e.g. high yield, good stability, low cost, etc.) [35,44], Pep10@MNPs serve as a competitive strategy for cancer cell isolation.
We therefore applied Pep10@MNPs to capture CTCs in 2.0 mL whole blood (WB) from breast cancer patients. On the basis of the four-color immunocytochemical (ICC) staining of nuclei/CKs/CD45/HER2, the captured cells that exhibited Nuclei + /CK + /CD45were identified as CTCs, distinguished from the contaminative WBCs that exhibited Nuclei + /CK -/CD45 + or the other blood components (Fig. 1d) [22].

HER2 phenotyping of breast cancer cells by immunocytochemistry (ICC)
We evaluated the expression level of HER2 in the four breast cancer cell lines. Both , correlated with the reported expression level of HER2 in these cell lines [42]. We therefore chose these cells with high expression of EpCAM and different expression levels of HER2 to establish a model system for cell isolation and HER2 phenotyping.
We achieved the HER2 testing of the captured cells by ICC in model systems. The fluorescent intensities from HER2 immunostaining were 10-fold higher (****P < 0.0001) in the HER2-overexpressed SK-BR-3 cells compared to the other cell lines.

Diagnostic and prognostic significance of baseline CTC enumeration in breast cancer patients
Fifty-four histopathologically confirmed breast cancer patients were enrolled for analysis of CTCs with baseline blood draws. CTCs could be detected in 81.5% (44/54) of the enrolled patients at baseline. We found that the patients had a broad range of CTC enumeration (0 -683 CTCs per 2.0 mL WB; mean ± standard deviation (S. D.), 47.2 ± 111.5) (Fig. 3a), whereas the epithelial cells were rare from healthy donors (mean ± S. D., 0.3 ± 0.7), consistent with the findings from Cristofanilli et al. [18,22]. The receiver operator characteristics (ROC) curve in Supplementary Fig. S5 [33]. Taken together, the detection of the CTC-positive cases (≥ 3 CTCs in 2.0 mL WB) showed promising diagnostic and prognostic values in breast cancer.
Herein, the CTC-HER2 (2+/3+) patients were described as the cases with HER2 overexpression in CTCs (CTC-HER2 + ). As shown in Table 2 Fig. S8; Table S3). Thereafter, we investigated the relationship of the HER2 phenotypes on baseline CTCs (CTC-HER2 + or CTC-HER2 (3+) ) with the histopathologic parameters of breast cancer, whereas no significant associations were achieved (Supplementary Table S4), in concordance with the conclusions by others [16,37]. Nonetheless, we found several intriguing relationships of the CTC-based HER2 phenotyping with the grading, bone metastasis and lung metastasis, which are close to the significant level (Supplementary Table S4).
We also assessed the prognostic significance of the CTC-based HER2 phenotyping at enrollment in predicting the PFS/OS. The CTC-HER2 + patients showed a significantly worse PFS than the CTC-negative cases (log-rank P = 0.024; hazard ratio (HR) of progression = 3.19, 95% CI of HR (1.01 to 10.11)) ( Supplementary Fig. S9).
No significant difference of the PFS was achieved between the patients with various HER2 phenotypes of CTCs, but the CTC-HER2individuals had a shorter median PFS than the one of CTC-HER2 + individuals (7.5 vs 9.2 months) ( Supplementary Fig. S9).
In OS analysis, we found the patients with CTC-HER2phenotype exhibited a worst OS (log-rank test, overall P = 0.015; P of CTC-HER2vs CTC-HER2 + = 0.099; P of CTC-HER2vs CTC-negative = 0.014) (Supplementary Fig. S9). The 20 Histo-HER2/neu + patients were determined into three groups on the basis of CTC enumeration and HER2 phenotyping at baseline, including the CTC-negative (n = 5), the CTC-HER2 -(n = 6) and the CTC-HER2 + (n = 9). We first showed the CTC enumeration at baseline was not a valid biomarker to predict the responses of the combination therapy (Fisher's exact P = 0.051, two tailed) (Fig. 5a). With the contribution of CTC-based HER2 phenotyping at baseline, the CTC-negative patients (n = 5) were detected with 4 cases with PR and one case with Non-CR/non-PD, and meanwhile the CTC-HER2 + patients (n = 9) showed 6 cases with PR, , whereas all of the CTC-HER2individuals (n = 6) reached the SD/PD (Fig. 5b; Supplementary Table   S5), indicating an increased ORR (71.4%) if the cases with negative expression of HER2 on CTCs were excluded. We statistically demonstrated a significant difference of the best overall responses in the three cohorts of patients (Fisher's exact test, overall P = 0.004, two tailed) (Fig. 5b). Furthermore, the CTC-HER2 + phenotype at baseline was assessed with a most potential effect size in predicting the responses of PR/Non-CR/non-PD (AUC = 0.833, P = 0.034; sensitivity, 66.7%; 95% CI of sensitivity, 35.4% to 87.9%; specificity, 100%; 95% CI of specificity, 61.0% to 100%), superior to other CTC-derived paramters (e.g. the count/ratio of the HER2 (2+/3+) CTCs, the count/ratio of the HER2 (3+) CTCs, the count of total CTCs) ( Fig. 5c; Supplementary Table S6).
Thereafter, based on the analysis of the second blood draws of the 14 individuals of the patients with trastuzumab-based combination therapies, we investigated the significance of the dynamic change of CTC count and HER2 phenotype from baseline to the follow-up visit (Supplementary Table S5). At second blood draws, 8 (14) of the patients had the PR or Non-CR/non-PD, 5 (14) of the cases showed the status of SD and one patient already progressed. As shown in Fig.6a, b, we vividly demonstrated the dynamic change of total CTC count from baseline to the follow-up visit, whereas no significant correlation with the real-time responses was found. By contrast, we statistically showed a significant association of the evolutionary HER2 phenotypes on CTCs with the follow-up efficacy of combination therapies (Fisher's exact test, overall P = 0.007) (Fig. 6c, d). The patients with the depletion of CTC-HER2phenotype were all assessed with the responses of SD/PD at second blood draws, contributing to the significant differences from the patients with depletion of CTC-HER2 + phenotype (Fisher's exact test, P = 0.015) or with consistently CTC-negative (Fisher's exact test, P = 0.048). Taken together, the HER2 phenotyping of the baseline CTCs or its dynamic change were valuable for predicting the outcomes of the trastuzumab-based combination therapies in breast cancer.

Discussion
The FDA-approved CellSearch system is extensively utilized to detect CTCs in 7.5 mL whole blood (WB). However, a limited portion of patients with primary breast tumors (22% -24%) [31,32] or metastatic breast tumors (45% -70%) [18,33,34] could be detected with CTCs by CellSearch, which mightbe because of the limited anti-EpCAM molecules on the surface of 10 nm ferrofluid. With the advantages of peptides and larger magnetic nanoparticles, our team developed an innovative EpCAM-targeted technique (i.e. Pep10@MNPs) to enrich CTCs in 2.0 mL WB, which was validated with a high rate of CTC detection (70% -97%) in solid tumors (breast cancer, lung cancer, etc.) [20,36,[45][46][47]. In the present study, 81.5% of the enrolled breast cancer patients were detected with CTCs by Pep10@MNPs (mainly 1 -300 CTCs/2.0 mL WB), competitive to the one of CellSearch (mainly 1 -500 CTCs/7.5 mL blood) [17,48]. Moreover, our data showed that the rate of the cases with more than 3 CTCs that were determined with a worse prognosis by Liu et al. could reach 70.4% [46]. By contrast, by CellSearch, Cristofanilli and Pestrin et al. showed a lower recovery rate of the metastatic breast cancer (MBC) patients with ≥ 2 CTCs (34.5% or 60.6%) [22,49]. Therefore, the efficient capture of CTCs by Pep10@MNPs in 2.0 mL WB might allow us to achieve the analysis of CTCs for the undetectable cases by CellSearch.

The cancer staging guidelines (version 8) of breast cancer by the American Joint
Committee on Cancer (AJCC) recommended CTC enumeration as a valid staging indicator of breast cancer [23]. In this trial, our data demonstrated the high level of CTCs (≥ 3 CTCs) by Pep10@MNPs was significantly correlated with the negative expressionof ER/PR on tumor tissues, different from the stuidies by CellSearch to the best of our knowledge [18,22,33]. Nonetheless, in breast cancer, the basal-like subtype (ER/PR -) are more aggressive and showed a worse prognosis as compared to the luminal A/B subtypes (ER/PR + ), suggesting the rationality of the finding [50].
Furthermore, the expression level of Ki67 antigen (Ki67 LI) represents the cell proliferation, correlated with the malignancy and therapeutic outcomes of various cancers [51][52][53][54][55]. The histopathologic Ki67 LI has been valuable biomarker used to determine the luminal A/B subtypes of breast cancer [39,40]. Two teams (Lucci's and Sandri's) previously investigated the relationship between CTC count and Ki67 LI of breast tumors, but both found no significant correlation [27,56]. Differently, our data demonstrated a significant association (P = 0.003) with the cutoff of ≥ 3 CTCsin 2.0 mL WB. The various dichotomous cutoff of the Ki67 LI and CTC enrichment by CellSearch from Sandri's group might be the reasons for the different result [57]. As demonstrated in some CellSearch-based reports, we did not found significant correlations of CTC count with grading, sites of metastasis and HER2 expression on tumor tissues as well [18,22,31,33,57]. Intringuingly, enumerating CTCs had a strong correlation with the neu amplification in our trial, suggesting the potential to aid the decision making of the IHC-based HER2 (2+) patients. This finding is consistent with the one by Lucci's group, whereas both are yet limited by the sample size [56]. Moreover, as expected, a strong correlation of the CTC enumeration (≥ 3 CTCs) with the PFS/OS of breast cancer patients was achieved, in concordance with the other CellSearch/TumorFisher-based publications [18,22,25,36]. Taken together, the prevalence of epithelial CTC by Pep10@MNPs could serve as a valid diagnostic and prognostic marker of breast cancer. Nevertheless, these findings should be supported by a larger cohort of breast cancer patients in the future.
HER2/neu is predominant in the classification and treatment of breast cancer.
Niikura et al. demonstrated that a cohort of breast cancer patients (6% -64%) showed a loss of HER2/neu + status at primary/metastatic tumors after preoperative systematic therapy and some patients (2% -15%) exihibited an acquisition of HER2/neu + status at metastases, which resulted in a shorter overall survival than the one of concordant HER2/neu status [12]. However, based on the invasive and biased tissue biopsies, it is difficult to overcome the spatio-temporal molecular heterogeneity of breast cancer. The at least one HER2-strongly/moderately stained (HER2 (2+/3+) ) CTCs in the primary/metatstatic breast cancer patients [31,58]. In other previous studies, the metastatic breast cancer patients with at least 5 CTCs and at least one HER2 (3+) CTCs were determined as the CTC-HER2 + cases [16,28]. Besides, some trials explored the contribution of the ratio of HER2-positive CTCs in a reasonable definition of the the CTC-HER2 + , such as 30% or 50% as the cutoff [29,33,49]. By contrast, in the CTCpostive cases (≥ 3 CTCs/2.0 mL WB), we determined the individuals with > 12% HER2 (2+) CTCs or with at least one HER2 (3+) CTCs, whose rationality was supported by the HER2 testing by IHC. The HER2 phenotyping of the enriched CTCs by Pep10@MNPs, could achieve a competitive overall accuracy (~ 71%) in identifying the patients with various histopathologic HER2/neu status, as compared to the reported concordance rates (38% -86%) (Supplementary Table S7). Half of the patients with HER2/neu-overexpressed/amplified tumor tissues were determined with the negative expression of HER2 on CTCs. Various sensitivities from ICC and IHC might be one of the intrinsic reasons. The heterogeneity of HER2 profile in primary/metastatic breast tumors and dynamic change of HER2 on CTCs should also be included in the reasons for this discordance [11,12,30]. With respect to the underlying diagnostic values of the HER2 reassessments on CTCs, our findings demonstrated the potential of the CTCbased HER2 phenotyping in detecting the poordifferentiation or bone/lung metastasisof breast cancer was close to the significant levels, superior to the CTC enumeration.
Similarly, Brandt's group found the prevalence of HER2-positive CTCs in the poorlydifferentiated patients (P = 0.033) [59]. Cristofanilli's team demonstrated an increased likelihood of HER2-positive CTCs in the patients with bone metastasis, whereas they found no correlation with the other metastatic sites [60]. Hence, the significances described in the trials above were controversial at some extent. As compared to the CTC enumeration at baseline, HER2 pheotyping on CTCs achieved a more sensitive  [27,29,37]. Moreover, no obvious difference of OS in the patients with various CTC-based HER2 phenotypes was observed by Munzone et al. [27].
Clinically, breast cancer patients should be recommended to accept anti-HER2 monoclonal antibodies alone or in combination after the HER2/neu-positive primary/metastatic breast tumors are determined. As demonstrated by Jordan et al, the HER2-negative CTCs were insensitive to anti-HER2 monoclonal antibodies and easily resistant to cytotoxic chemotherapy [30]. Hence, the CTC-HER2cases at baseline might suggest an undesirable outcome of trastuzumab-based combination therapies.
Our investigation preliminarily provided proofs to the assumption in a small smaple size (n = 20), where the objective response rate reached 71.4% when the exclusion of the baseline CTC-HER2individuals were done in the analysis. As a comparison, the potential of CTC enumeration in predicting the outcomes of the combination therapies was also investigated in our trial and showed no association, consistent with the implications from a larger retrospective trial by Cristofanilli's group [25]. Therefore, beyond enumerating CTC, the HER2 phenotyping on CTCs before therapies may serve as a more useful predictor for the personalized administration of anti-HER2 therapy in breast cancer. Second, in a small sample size with second blood draws (n = 14), we showed the dynamic change of CTC enumeration and CTC-HER2 phenotypes after the combination therapy. As expected, the depletion of epithelial CTCs was far from a practical predictive biomarker of the treatment outcomes [61]. A few groups

Supplementary data
Supplementary data to this article can be found online at

Availability of data and material
All data generated or analysed during this study are included in this published article and its supplementary information files.

Consent for publication
Not applicable.       from CTC-positive to CTC-negative (i.e. from ≥ 3 CTCs to < 3 CTCs) (n = 11), or from CTC-negative to CTC-positive (n = 1), or with consistently CTC-negative (n = 2). P values were analyzed by the Fisher's exact test (two-tailed). Not significant (ns), P > 0.05. c Heat map of the CTC-HER2 phenotype before and after the combination therapy for each monitored patient. Color blocks represented various CTC-HER2 phenotypes (Cambridge blue, the CTC-negative; gray, the CTC-HER2 -; pink, the CTC-HER2 + ). d Comparison of patient response to the combination therapy in the four groups of individuals with dynamic change of CTC-HER2 phenotypes. The individuals were determined as the cases with depletion of CTC-HER2 + (n = 6), or with depletion of CTC-HER2 -(n = 5), or with acquisition of CTC-HER2 + (n = 1), or consistently with CTC-negative (n = 2). P values were analyzed by the Fisher's exact test (two-tailed). Significant level, *P < 0.05, **P < 0.01, including P = 0.015 (depletion of CTC-HER2 + vs depletion of CTC-HER2 -), P = 0.048 (depletion of CTC-HER2 -vs consistently with CTC-negative) and overall P = 0.007.