1.

Introduction

Collagen is an important protein within the extracellular matrix (ECM) of normal and malignant tissues. In normal breast tissue, increased collagen is associated with increased risk of developing breast cancer and increased disease aggressiveness.¹ Recent studies of intratumoral collagen demonstrate that collagen fiber alignment and density associate with aggressive disease and increased invasion.^1–3 Furthermore, a high density of interconnected collagen fibers within the ECM of tumors can block the diffusion of drugs into the tumor.⁴ These studies highlight the importance of collagen within the tumor microenvironment (TME), particularly with respect to tumor invasion and drug response. They also demonstrate a need for studies that monitor dynamic changes in collagen during breast tumor progression and elucidate how intratumoral collagen may change in response to anticancer treatments.

Breast cancers are divided into three clinical subtypes: estrogen receptor (ER) positive, human epidermal growth factor receptor 2 (HER2) positive, and triple negative breast cancer (lacking expression of ER, HER2, and progesterone receptor). HER2 is overexpressed in $\sim 20\%$ of breast cancers, correlating with poor clinical outcome and increased tumor malignancy.⁵ Fortunately, several drugs are approved for direct targeting of HER2, including trastuzumab (Tz, a monoclonal antibody) and lapatinib (a tyrosine kinase inhibitor). HER2 inhibitors increase the survival of patients when used alone and in combination with standard chemotherapies.^6–9 Unfortunately, many patients do not initially respond to HER2 inhibitors or develop acquired resistance to these targeted agents.⁹

Tz specifically binds to the extracellular domain of HER2 and prevents activation of the intracellular HER2 tyrosine kinase. Tz has other mechanisms of action, including blockade of HER2 dimerization, increased HER2 endocytosis and degradation, and induction of antibody-dependent cellular cytotoxicity.^10,11 Tz induces cytostatic effects by arresting the cell cycle in the G1 phase and inhibits signaling involved in cell survival.^12,13 Additionally, Tz has been shown to inhibit angiogenesis and reduce the expression of growth factors, such as VEGF and transforming growth factor (TGF)-$\alpha$. However, additional changes that occur in the TME, particularly changes in collagen content and organization, following Tz treatment are poorly understood.

Images of collagen within the TME can be obtained from second harmonic generation (SHG) microscopy. SHG is a nonlinear optical phenomenon in which two photons of the same wavelength are combined through a material to yield one photon with twice the energy (one-half the wavelength) of the incident photon. Collagen SHG imaging probes the fibrillar collagen content of tissues and generates photon intensity images of collagen fibers within the tissue, which can be analyzed to assess collagen density and alignment.^14–18 Collagen fibers, such as type I, II, III, V, and XI, exhibit a strong SHG signal. In particular, increased amounts of collagen types I, III, and V have been found within breast cancers.^19,20 SHG imaging of breast cancer samples has revealed distinct collagen alignment configurations that correlate with increased tumor aggressiveness and invasion.^1–3,21 SHG images can be assessed for collagen density and relative alignment of collagen fibers. Additionally, tumor collagen can be assessed by trichrome staining of histological slides. Trichrome stains all types of collagen nondiscriminately and provides information complementary to that quantified from SHG images of fibrillar collagen. In contrast to trichrome stains, SHG microscopy can be performed in vivo.

Few studies have characterized changes in the TME following chemotherapy or targeted therapy. Therefore, this study investigates changes in the collagen content of HER2+ tumors following treatment with Tz. Two isogenic HER2+ breast cancer cell lines, one Tz-responsive (BT474) and the other selected for Tz resistance (HR6), were established as xenografts in athymic mice. Tumor-bearing mice treated with Tz or with a control immunoglobulin (IgG) were used to assess collagen content from SHG images in vivo at three time points (2, 5, and 14 days of treatment) and in histological trichrome-stained tumor sections harvested at days 2, 5, and 14. These studies tested the hypothesis that Tz treatment affects collagen in the TME, independent of the effect of Tz on the tumor cells themselves.

2.

Materials and Methods

3.

Results

The collagen content of Tz-responsive (BT474) and Tz-resistant (HR6) tumors was examined by in vivo collagen SHG imaging and trichrome staining over a 14-day time course of control-IgG or Tz treatment. BT474 tumors respond to Tz with increased cell death and decreased tumor volume, while Tz does not induce cell death or decrease growth in HR6 tumors.^22,24,25 Collagen SHG images were analyzed using CurveAlign to extract the relative fiber alignment and density of collagen positive pixels (Fig. 1). Figure 1(c) shows the thresholded intensity mask image of collagen density, where all collagen positive pixels have a value of 1 and all negative pixels have a value of 0. Trichrome stained tumor sections were scored from 0 to 8 based on the presence and intensity of the intra- and extratumor collagen. Representative images of trichrome stained tumor sections demonstrate trichrome scoring heterogeneity [Figs. 1(d)–1(f)].

Fig. 1

(a) Original collagen second harmonic generation (SHG) image. (b) Overlay image of identified collagen fiber angles (green arrows). (c) Collagen density obtained from a thresholded image. (d) to (f) Representative trichrome images obtained at $400\times$ to demonstrate a low [(d), score of 2], medium [(e), score of 5], and high [(f), score of 8] score. Scale bars are $100\mu \mathrm{m}$.

Collagen analysis revealed significant differences as a result of tumor growth and treatment with Tz in BT474 xenografts. Representative collagen SHG images demonstrate the typical size, density, and alignment of collagen fibers within control and Tz-treated BT474 tumors over the time course [Fig. 2(a)]. Collagen density was obtained from collagen SHG images as the portion of collagen positive pixels. Control BT474 tumors initially had a collagen density of 0.38 on day 2, which significantly ($p<0.05$) increased to 0.47 on day 5 and was maintained at 0.49 on day 14 [Fig. 2(b)]. No significant differences in collagen density were detected between control and Tz-treated BT474 tumors. Likewise, no differences in collagen alignment were detected in BT474 tumors treated with Tz for two or five days. On day 14, however, there was a significant reduction in the collagen alignment of Tz-treated BT474 tumors compared with control BT474 tumors [$p<0.05$; Fig. 2(c)]. The intratumor variance for the density of the BT474 tumors was 0.002, while the intergroup variance was 0.004, suggesting greater differences between tumors within a time point and treatment group than within individual tumors. The intratumor variance for the coefficient of alignment of collagen fibers within BT474 tumors was 0.008, while the intergroup variance was 0.02 ($p<0.05$), again indicating greater heterogeneity between individual tumors rather than within a tumor.

Fig. 2

(a) Representative collagen SHG images from BT474 tumors. Scale bar is $100\mu \mathrm{m}$. (b) Collagen density of SHG images for control and trastuzumab (Tz)-treated BT474 tumors. (c) Coefficient of alignment of SHG images for control and Tz-treated BT474 tumors. (d) Trichrome staining score for control and Tz-treated BT474 tumors. † $p<0.05$ versus day 2. * $p<0.05$ versus control.

Trichrome staining of collagen revealed significant differences in collagen between control and Tz-treated BT474 tumors. A trichrome staining score was assigned to each histology section based on visual analysis of the intra- and extratumoral collagen, both dense and diffuse. An increased score indicates increased collagen content. Significant reductions in collagen trichrome scores were detected on 2, 5, and 14 days ($p<0.05$) of treatment in the BT474 tumors [Fig. 2(d)].

Representative collagen SHG images from control and Tz-treated HR6 tumors are shown in Fig. 3(a). Control HR6 tumors showed an increase in collagen density over time, from 0.38 on day 2 to 0.55 on day 14 [$p<0.05$; Fig. 3(b)], similar to the increase in collagen density observed in the control BT474 tumors over time [Fig. 2(b)]. Tz resulted in stable collagen SHG density in both the BT474 and HR6 tumors across all time points [Figs. 2(b) and 3(b)]. No significant differences in collagen density between the Tz-responsive and Tz-resistant tumors (within control and Tz-treated groups) were observed at any time point ($p>0.05$). The intratumor variance for the collagen density of the HR6 tumors was less than the variance between HR6 tumor groups (0.003 and 0.007, respectively; $p<0.05$). Likewise, the intratumor variance for the collagen coefficient of alignment of the HR6 tumors was less than the variance between HR6 tumor groups (0.01 and 0.02, respectively; $p<0.05$). Tz failed to induce a decreased collagen alignment in HR6 tumors over time or between control and Tz-treated tumors [Fig. 3(c)]. A significant reduction in collagen alignment ($p<0.05$) was detected in control HR6 tumors compared with control BT474 tumors on days 5 and 14. A significant reduction in the trichrome score was detected in Tz-treated HR6 tumors compared to control tumors at every time point [Fig. 3(d)].

Fig. 3

(a) Representative collagen SHG images from HR6 tumors, scale bar is $100\mu \mathrm{m}$. (b) Collagen density of SHG images for control and Tz-treated HR6 tumors. (c) Coefficient of alignment of SHG images for control and Tz-treated HR6 tumors. (d) Trichrome staining score for control and Tz-treated HR6 tumors. † $p<0.05$ versus day 2. * $p<0.05$ versus control.

Correlation analysis was performed between collagen endpoints from control BT474 and HR6 tumors using matched tumor data to investigate relationships between SHG imaging endpoints and trichrome scores. Neither the collagen density nor the coefficient of alignment determined from SHG images [Fig. 4(a)] correlated with the trichrome score. Additionally, none of the endpoints, trichrome score, SHG density, or SHG coefficient of alignment, correlated with tumor volume [Figs. 4(b)–4(d)]. The SHG density did not correlate with the coefficient of alignment [Fig. 4(e)]. However, a significant correlation was detected between SHG density and the age of the tumor [Fig. 4(f)].

Fig. 4

(a) Correlation between trichrome score and SHG density. (b) Correlation between trichrome score and tumor volume. (c) Correlation between SHG density and tumor volume. (d) Correlation between SHG coefficient of alignment and tumor volume. (e) Correlation between SHG density and SHG coefficient of alignment. (f) Correlation between SHG density and age of tumor. $r$ is Spearman’s correlation coefficient. ns, not significant. Tumor volume measurements were obtained from paraffin embedded tumors, which result in tissue shrinkage.

4.

Discussion

Tz is an HER2-targeted antibody prescribed clinically for treatment of breast cancers that overexpress HER2. While many studies have characterized the effects of Tz on the tumor epithelium, few studies have investigated the effects of Tz on the tumor microenvironment. This study investigated the effects of Tz on breast tumor collagen in both Tz-responsive and Tz-resistant tumors.

Representative collagen SHG images demonstrate the typical size, alignment, and density of collagen fibers within the tumors over time and with treatment [Figs. 1(a), 2(a), and 3(a)]. These images were acquired from a microscope system designed for multiphoton fluorescence imaging and fluorescence lifetime imaging. The only modification necessary to image collagen SHG was the addition of a narrow emission filter at one-half the wavelength of the incident near-infrared light. Collagen SHG imaging is sensitive to the polarization of the incident light. The incident light in our experimental setup is attenuated by a Pockel cell to vary the power of the illumination light, which results in an elliptically polarized beam. Therefore, we should capture collagen fibers at all angles, but the intensity of the fibers may be direction-dependent due to the polarization of the incident beam. In order to account for this bias in intensity with fiber orientation, we applied an intensity threshold to our images for further analysis so that our endpoints are independent of absolute pixel intensity. Collagen SHG images were analyzed to extract collagen density and relative fiber alignment [Figs. 2(b), 2(c), 3(b), and 3(c)].

Analysis of the collagen SHG images revealed a statistically significant increase in the density of collagen in control BT474 and HR6 tumors over time [Figs. 2(b) and 3(b)]. Correlation analysis (Fig. 4) does not show a correlation between larger tumors and collagen density. However, there is a significant correlation between SHG density and tumor age. These results suggest increased fibrillar collagen within aging tumors, which is consistent with previous reports of increased collagen content in malignant and progressing tumors.¹ Upon treatment with Tz, the BT474 tumors shrink over 14 days,²⁵ and the collagen density within these tumors showed no significant change over the time course [Fig. 2(b)]. This suggests that Tz decreases the amount of fibrillar collagen in parallel with decreases in tumor volume in responsive tumors, thus maintaining constant collagen density over time.

The relative alignment of collagen fibers is associated with tumor invasion and progression, with radially aligned fibers characteristic of aggressive disease.^1–3 To assess collagen fiber alignment, a coefficient of alignment was determined for each image using CurveAlign. Images with perfectly parallel fibers would have an alignment coefficient of 1, while images with perpendicular fibers would have a value of 0. The Tz-treated BT474 tumors showed a significant decrease in coefficient of collagen alignment compared to control BT474 tumors on day 14 [Fig. 2(c)], suggesting less aggressive characteristics upon treatment with the therapeutic antibody. No change in collagen alignment was detected in the HR6 tumors treated with Tz, suggesting no significant effect of Tz on the collagen fiber alignment of these nonresponding tumors. A significant reduction in collagen alignment was detected in control HR6 tumors compared with control BT474 tumors on days 5 and 14. This may be due to differences in the characteristics and behavior of these two tumors. The HR6 tumors grow at a faster rate than the BT474 tumors, which may influence microenvironment composition.²⁵ HR6 cells also express higher levels of epidermal growth factor receptor and phosphorylated epidermal growth factor receptor compared to the BT474 cells,²² which may influence tumor behavior and microenvironment composition.^27,28

Histological analysis of collagen stained with trichrome dye revealed significant reductions in collagen content in BT474 tumors treated with Tz at all time points [Fig. 2(d)]. The trichrome staining score is a sum of four measures, diffuse and dense collagen both intra- and extratumoral, and all four individually show similar trends. Trichrome stains all forms of collagen, including fibrillar collagen. Tumors contain multiple types of collagen, including types I, II, III, and IV. Collagen type IV is particularly important within tumors because it composes the basal membrane and is involved in tumor invasion.^29,30

The HR6 SHG and trichrome results suggest that Tz may have direct effects on the tumor microenvironment. The Tz-treated HR6 tumors did not show an increase in SHG density over the 14-day study, consistent with the Tz-treated BT474 tumors and contrary to the control HR6 and BT474 tumors [Figs. 2(b) and 3(b)]. Furthermore, the trichrome scores revealed significantly reduced scores at all time points for the Tz-treated HR6 tumors compared to the control HR6 tumors [Fig. 3(d)]. Few studies have investigated the effects of Tz treatment on nonresponsive tumor cells and the tumor microenvironment. In a study of a novel glaucoma surgical intervention, Tz was tested as an antiscar agent,³¹ revealing that Tz suppressed fibroblast proliferation and reduced tissue levels of TGF-$\beta$, fibroblast growth factor-$\beta$, and platelet-derived growth factor.³¹ This and the HR6 results suggest that Tz may affect the collagen content of tumors by directly modifying fibroblast behavior or by modulating paracrine growth factors secreted by tumor cells. Furthermore, our previous studies of Tz effects on the tumor vasculature of BT474 and HR6 tumors revealed significant decreases in vessel oxygen saturation with five days of Tz treatment in both tumor models.²⁴ Our studies and these previous results suggest that Tz treatment may affect the tumor microenvironment independent of its effects on the tumor cells.

The correlation analysis (Fig. 4) indicates that the SHG endpoints, collagen density, and coefficient of alignment do not correlate with the trichrome staining score or with each other. These results agree with findings of Strupler et al. who directly compared sequential slides of collagen SHG and trichrome strain and reported that trichrome is less specific at identifying fibrillar collagen and is difficult to score and compare with SHG.³² This suggests that SHG density, SHG coefficient of alignment, and trichrome score are different measurements. The inconsistencies between trichrome staining and SHG may be due to the source of contrast. Trichrome stains all collagen, including fibrillar and nonfibrillar collagen as well as other components of the ECM, while only certain fibrillar collagen strands will exhibit SHG.³² Tumors can have increased content of nonfibrillar collagens, in particular, collagen IV,^29,30 which does not exhibit SHG.³² Additionally, discrepancies between SHG density and trichrome staining may be due to sampling error in either the SHG measurements or trichrome imaging, both of which are limited to small fields of view and are, thus, probing only a small portion of the tumor. Imaging additional trichrome stained histological slices of the tumor or obtaining three-dimensional stacks of SHG images may provide increased representation of the tumor and more accurate measurements.

The results of this study highlight several changes in the collagen content and organization of the tumor microenvironment both in growing tumors and with Tz treatment. However, there may be confounding factors that influence the composition of the tumor microenvironment and the effectiveness of Tz. The tumor ECM can affect drug diffusion^18,33,34 and extracellular proteins, such as laminin, can co-localize with HER2 and limit the efficacy of Tz.³⁵ Therefore, the initial composition of the tumor ECM may influence Tz treatment. Furthermore, variability in tumor growth rates may cause variable vascular instability, which may affect the tumor microenvironment, collagen measurements, and Tz delivery and effectiveness.³⁶ In this study, variability across tumors was controlled by the use of a single batch of cells to initiate tumors, the use of age-matched mice, injection of the same number of cells per tumor, and the use of mouse weight to determine the Tz dose. However, some variability occurred between tumors within groups due to experimental error and biological variance. This inherent variability may mask some effects of Tz on the tumor microenvironment and may limit the effectiveness of Tz inhibition of HER2.

In conclusion, these studies suggest that Tz treatment may affect the tumor microenvironment independent of its effects on the tumor cells. We have also demonstrated that collagen SHG and histological trichrome staining analysis of collagen provide complementary endpoints of tumor collagen content and morphology. Collagen SHG imaging can be performed in vivo through window chambers for longitudinal studies,^24,37 while trichrome staining is inherently postmortem, potentially limiting the number of time points and increasing animal burden and expense. Trichrome staining revealed decreased collagen content in both Tz-responsive and nonresponsive tumors treated with Tz. Furthermore, collagen SHG imaging revealed similar trends in collagen density in both Tz-responsive and nonresponsive tumors treated with Tz, suggesting consistent effects of Tz on the tumor microenvironment irrespective of Tz-induced tumor size reduction. However, there was a decrease in the alignment of collagen fibers of Tz-treated responsive tumors only after 14 days of treatment. These results highlight the changes occurring within the collagen compartment of the tumor microenvironment following systemic Tz treatment.