Collagen Alignment as a Predictor of Recurrence after Ductal Carcinoma In Situ

Abstract

Background: Collagen fibers surrounding breast ducts may influence breast cancer progression. Syndecan-1 interacts with constituents in the extracellular matrix, including collagen fibers, and may contribute to cancer cell migration. Thus, the orientation of collagen fibers surrounding ductal carcinoma in situ (DCIS) lesions and stromal syndecan-1 expression may predict recurrence.Methods: We evaluated collagen fiber alignment and syndecan-1 expression in 227 women diagnosed with DCIS in 1995 to 2006 followed through 2014 (median, 14.5 years; range, 0.7-17.6). Stromal collagen alignment was evaluated from diagnostic tissue slides using second harmonic generation microscopy and fiber analysis software. Univariate analysis was conducted using χ² tests and ANOVA. The association between collagen alignment z-scores, syndecan-1 staining intensity, and time to recurrence was evaluated using HRs and 95% confidence intervals (CIs).Results: Greater fiber angles surrounding DCIS lesions, but not syndecan-1 staining intensity, were related to positive HER2 (P = 0.002) status, comedo necrosis (P = 0.03), and negative estrogen receptor (P = 0.002) and progesterone receptor (P = 0.02) status. Fiber angle distributions surrounding lesions included more angles closer to 90 degrees than normal ducts (P = 0.06). Collagen alignment z-scores for DCIS lesions were positively related to recurrence (HR = 1.25; 95% CI, 0.84-1.87 for an interquartile range increase in average fiber angles).Conclusions: Although collagen alignment and stromal syndecan-1 expression did not predict recurrence, collagen fibers perpendicular to the duct perimeter were more frequent in DCIS lesions with features typical of poor prognosis.Impact: Follow-up studies are warranted to examine whether additional features of the collagen matrix may more strongly predict patient outcomes. Cancer Epidemiol Biomarkers Prev; 27(2); 138-45. ©2017 AACR.

Figure 1

Process for evaluating collagen fiber alignment around DCIS lesions. A) Routine H&E slides from the tissue blocks at the baseline diagnosis. The scale bar indicates 100 μm. Abbreviations: D, DCIS lesion; S, stroma. B) Second-harmonic generation (SHG) microscopy generated a high-contrast image of the lesion comprised solely of collagen. C) The SHG image was transformed using the curvelet algorithm and the angle of curvelets with respect to the DCIS foci boundary within a 100 μm radius (green) was calculated using customized software. Images are 750 μm².

Figure 2

DCIS lesion with staining for syndecan-1 in the stromal fibroblasts. Rectangular boxes placed around the lesion boundary were evaluated for staining intensity. The scale bar indicates 100 μm.

Figure 3

Proportion of collagen fiber angles in 5 degree bins relative to the ductal boundary for median collagen alignment score in normal ducts (white) and DCIS lesions (black).

Figure 4

Kaplan-Meier estimates for time to recurrence as a function of collagen alignment score (above and below the median) from the first imputed dataset for (A) normal ducts and (B) DCIS lesions. Univariate hazard ratio for interquartile range (IQR) increase in collagen alignment score from Cox proportional hazards regression with collagen alignment score as a continuous (linear) predictor. Shaded regions reflect 95% confidence bands for the survival curves.

Figure 5

Kaplan-Meier estimates for time to recurrence as a function of average syndecan-1 staining intensity (above and below the median). Univariate hazard ratio for interquartile range (IQR) increase in staining intensity from Cox proportional hazards regression with average syndecan-1 staining intensity as a continuous (linear) predictor. Shaded regions reflect 95% confidence bands for the survival curves.