Open Access
Add to BibSonomy
Abstract
Overexpression of heat shock protein 90 (Hsp90) on the surface of breast cancer cells makes it an attractive molecular biomarker for breast cancer diagnosis. Before a ubiquitous diagnostic method can be established, an understanding of the systematic errors in Hsp90-based imaging is essential. In this study, we investigated three factors that may influence the sensitivity of ex vivo Hsp90 molecular imaging: time-dependent tissue viability, nonspecific diffusion of an Hsp90 specific probe (HS-27), and contact-based imaging. These three factors will be important considerations when designing any diagnostic imaging strategy based on fluorescence imaging of a molecular target on tissue samples.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
The vast majority of patients with breast cancer are diagnosed by core needle biopsy (CNB) and undergo treatment with breast conserving surgery (BCS) followed by radiation therapy [1,2]. The current gold standard for diagnostic CNB and post-operative margin assessment of BCS requires histological analysis of excised tissues [3,4]. The major limitations for traditional histology are multi-step tissue preparation and reliance on specialized personnel, resulting in inefficiencies between diagnosis and treatment. There is a wide-range of re-excision rates reported, but most groups report 20-40% of patients undergo at least one re-excision [3]. Secondary surgeries both increase cost of care dramatically and result in surgeons performing fewer primary surgeries, leading to delays in treatment for women. Ex vivo tissue imaging can be used during CNB to reduce time to diagnosis and intraoperatively to perform margin assessment in real time to reduce the number of secondary surgeries by providing an on-site evaluation of excisional specimens prior to full pathological review.
Advances in new microscopy technologies as well as molecular imaging have the potential to identify malignant tissue ex vivo, providing opportunities to augment traditional post-procedure pathology with real-time feedback either during CNB or in the operating room. For example, advances in microscopy with ultraviolet surface excitation (MUSE) [5] and light-sheet fluorescence microscopy [6] demonstrate the capability to provide 3D high-resolution pathology-like images of biopsies in lieu of histological sections, and machine learning algorithms can automate this process. Other groups have demonstrated that molecular imaging can be used to perform cancer diagnosis, margin assessment, and fluorescence-guided surgery [7–12]. Studies have shown that cancer-specific biomarker can be used enhance the specificity of molecular-imaging based margin assessment [13–15].
Point of care molecular imaging has the potential to enhance current histology practice and can be tailored to visualize different proteins that are overexpressed in breast cancer. We have previously leveraged the unique over expression of Heat shock protein 90 (Hsp90) on the surface of breast cancer cells for diagnostic applications in breast cancer [16–19]. Compared to other molecular markers for breast cancer, Hsp90 has the advantage of expression across all different sub-types of breast cancer, including ductal carcinoma in situ (DCIS). Hsp90 plays an important role in breast cancer formation, orchestrating stress responses and protein folding as a chaperone protein [20,21]. Overexpression of Hsp90 across all breast cancer receptor subtypes and DCIS [22] provides the rationale for pursuing Hsp90 as a molecular target for point-of-care breast cancer diagnosis.
We have previously established a non-invasive point of care imaging approach to quantify Hsp90 expression using a FITC-tethered Hsp90 inhibitor (HS-27) that specifically binds to Hsp90 expressed on the surface of breast cancer cells [17,23,24]. In our in vivo pre-clinical studies, we found that highly glycolytic tumors had the highest surface Hsp90 expression, while non-tumorigenic cells showed little to no surface Hsp90. We also demonstrated that HS-27 binds to surface Hsp90 on excised tissue biopsies, which can be rapidly imaged at the point of care to differentiate breast cancer from benign tissues [23,25]. This provides a practical approach to provide automated diagnostics of image guided CNBs or fine needle aspirates (FNA) and tumor margins.
Our goal is to develop a low-cost, non-contact molecular diagnostic platform to discriminate between malignant and benign tumor margins and breast core needle biopsies (CNBs) as an alternative to labor-intensive pathology, with the ultimate goal of applying it to tumor margin assessment with a wide field, high resolution probe. However, it is important to characterize the sources of systematic error that can lead to diminished contrast and account for that prior to designing such a system. Here, we investigate how three sources of systematic error affect ex vivo Hsp90 imaging. Specifically, we explored the effect of tissue viability, molecular probe diffusion kinetics, and contact vs. non-contact Hsp90 imaging methods as potential error sources. We used 4T1 mammary tumors grown in a nude mouse model, from which CNBs were obtained to emulate the clinical scenario yet provide a controlled environment in which to test these variables. Our results were critical in identifying the window in which variables that confound Hsp90 contrast are minimized. These studies also provide a methodology for optimizing point of care molecular imaging of tissue biopsies with other agents.
2. Methods
2.1 Cell culture
The murine mammary carcinoma cell line (4T1) and normal murine mammary epithelial cell line (NMuMG) were used in the in vitro and ex vivo studies. All cell lines were obtained from the American Type Culture Collection (ATCC) and cultured under sterile conditions at 37°C and 5% CO2 level. NMuMG cells were cultured in 90% Dulbecco's Modified Eagle's Medium with 4.5 g/L glucose and 10 mcg/ml insulin, 10% fetal bovine serum (FBS), and 1% penicillin-streptomycin. 4T1 cells were cultured in 90% RPMI-1640 (L-glutamine), 10% FBS, and 1% penicillin-streptomycin. All cell lines were used within 10 cell passages.
2.2 In vitro effect of cell viability on Hsp90 fluorescence imaging
NMuMG cells were treated with 35 mM hydrogen peroxide (H2O2) for 15 minutes to compromise cell viability [26]. Negative control cells received no treatment. Both H2O2-treated and control NMuMG cells were stained with ReadyProbes Cell Viability Imaging Kit, Blue/Red (Invitrogen), containing Hoechst 33342 for cell nuclei staining and propidium iodide (PI) for cell nuclei with compromised plasma membrane integrity. All cells were stained with Hoechst 33342 (blue fluorescence) and 100 µM HS-27 (green fluorescence) for 15 minutes or Hoechst 33342 and PI (red fluorescence) for 15 minutes. All cells were imaged with a Zeiss Upright 780 Confocal Microscope using a 405 nm laser and 458 ± 25 nm emission filter for Hoechst, a 488 nm laser and 525 ± 25 nm emission filter for HS-27, and a 561 nm laser and 615 ± 25 nm emission filter for PI.
2.3 Animal studies
All animal studies were approved by the Duke University Institution for Animal Care and Use Committee under protocols A216-15-08 and A138-18-05.
2.4 Ex vivo effect of cell viability on hsp90 fluorescence imaging
Murine mammary tumors (4T1) were grown in the flank of female athymic nude mice (6–8 weeks old) by injecting a suspension of 106 cells suspended in serum free media into the right flank. After tumors grew to a volume of 1 cm3, a 12-gauge automated biopsy gun was used to collect up to three randomly located biopsies from each tumor. Mice were anesthetized with vaporized isoflurane (1–1.5% v/v) in air during biopsy. All mice were immediately euthanized after the biopsy procedure while under anesthesia. Sixty biopsies from 20 mice were randomly separated into three post-excision window groups: 1, 3, and 10 minutes. For each post-excision window group, 10 biopsies were treated with H2O2 to perturb cell viability. For all treated/untreated biopsies, half (n=5) were stained with 100 µM HS-27 for 1 minute and the other half were stained with PI (cell viability indicator) for 15 minutes. A customized fluorescence microscope [27] was used to image biopsies with ex:488 nm/em:525 ± 7 nm for HS-27 and ex:555 nm/em:615 ± 7 nm for PI with integration time of 500ms.
2.5 Hsp90 diffusion kinetics calculations
Thin slices of murine mammary tumors (4T1) (1.3 mm to 0.8 mm, n=5) harvested from 4T1 flank tumors were trimmed to fit 12-mm customized wells with porous polycarbonate membranes in contact with 100 µM HS-27 or HS-217 (inactivated version of HS-27) [25], with a sample size of 4 for each group. A customized fluorescent microscope (see above) was used to image the top surface of the specimen over 60 minutes, using ex:488 nm/em:525 ± 7 nm. The average fluorescence intensity over each time point was calculated and fit to a previously validated diffusion coefficient calculation model [28].
2.6 Validation of diffusion kinetics
4T1 murine breast tumors were stained with 100 µM HS-27 (n = 5) or HS-217 (n = 6) for 1 minute. The stained tumor specimens (without washing) were embedded into optimal cutting temperature compound (OCT compound) and flash frozen using liquid nitrogen. The frozen specimens were sectioned into 30-µm slices with a cryostat microtome (Microm HM 505 E, Walldorf, Germany). Five slices were obtained from each specimen and five fields-of-view (FOVs) were acquired using conventional fluorescence microscopy with ex:488 nm/em:525 nm. For all FOVs, the penetration depth was determined using the intensity profile from the tumor edge to the tumor center with a cut-off of 500 grayscale based on the average of background fluorescence of unstained tissue. Penetration depths from each FOV for each slice were averaged for all specimens. All images were analyzed with ImageJ.
2.7 Ex vivo Hsp90 fluorescence imaging with a contact and a non-contact imaging system
4T1 biopsies acquired from mammary tumors as described above were stained with either 100 µM HS-27 (n=4) or HS-217(n=4) for 1 minute. The stained biopsies were rinsed with PBS to remove residual stain. Tumor-mimicking Hsp90 liquid phantoms (2, 5, 10, and 25 µM) were prepared by mixing HS-27 with Polystyrene spheres (07310, Polysciences, Warrington, Pennsylvania) as the scatterer in 1X PBS solution. The reduced scattering level for all fluorescence phantoms was 20 cm−1, previously determined by averaging the scattering level from 390 nm to 650 nm [29]. A fiber-based micro endoscope (HRME) [30,31] and a customized microscope as described in the previous section were used to image surface HS-27 fluorescence of mammary tumor biopsies and tumor-mimicking phantoms using protocols previous described [24,25]. The customized fluorescence microscope [27] used ex:488 nm/em:525 ± 7 nm and the HRME utilized an LED light source centered at 455 nm (20 nm FWHM), a 475nm dichroic mirror, and a 500 nm long-pass filter as emission filter [30]. All images were calibrated using a fluorescence standard (USF 210-010, LabSphere) by dividing each pixel within the image by the mean pixel intensity from the fluorescence calibration slide.
2.8 Data processing and statistical analysis
The statistical significance between two groups was determined using a two-sided Student’s t-test. The statistical significance among three or more groups was determined using one-way analysis of variance (ANOVA) with Tukey’s test as the post-hoc test. Tukey’s test and ANOVA were performed with JMP. The statistical significance of survival curves comparing two or three groups was determined by the Kolmogorov-Smirnov (KS) test. The R2 and p-values of linear regressions were determined using least-squares regression. The KS test and least-squares regression were performed with the MATLAB Statistics Toolbox. A 95% confidence interval was used for all tests, with a p-value less than 0.05 considered as statistically significant.
3. Result
3.1 Membrane permeabilization due to cell death increases HS-27 fluorescence
Fluorescence imaging of HS-27 was performed within a 1-5-minute window that was previously used in the clinic [24]. To assess the effect of viability on HS-27 fluorescence, NMuMG, murine mammary epithelial cells, with no surface Hsp90 expression, were treated with 35 mM hydrogen peroxide (H2O2) for 15 minutes to induce apoptosis [26]. Figure 1(A) shows representative images of NMuMG cells stained with either HS-27 or PI with and without the viability challenge with H2O2, indicating increased HS-27 uptake in H2O2-treated cells. Quantification of HS-27 and PI fluorescence revealed a significant increase in both HS-27 and PI staining in the H2O2-treated group (Fig. 1(B)), suggesting decreased cell viability increases HS-27 fluorescence independent of surface Hsp90 expression.
Fig. 1. Membrane permeabilization due to cell death increases HS-27 fluorescence: Cell death was induced by incubation with 35 mM hydrogen peroxide (H2O2) for 15 minutes. Both H2O2 (-) and H2O2 (+) NMuMG (non-tumorigenic) murine mammary epithelial cells were stained with either Hoechst 33342 (blue fluorescence) and 100 µM HS-27 (green fluorescence) for 15 minutes or Hoechst 33342 (blue fluorescence) and Propidium Iodide (PI) (red fluorescence) for 15 minutes. Images were acquired with confocal microscopy. A) Representative images of H2O2 (-) and H2O2 (+) NMuMG with HS-27 and PI staining (n=5). Scale bars are 100 µm. H2O2 (-) (top) showed minimal HS-27 and PI uptake, while both HS-27 and PI signals were significantly increased in the H2O2 (+) (bottom) group. B) The normalized fluorescence was quantified per cell and shown as an average. HS-27 and PI fluorescence are significantly higher in H2O2 (+) NMuMG cells (p<0.01). For all experiments, p-values were determined by Student’s t-test. ** indicates p-value is less than 0.01
Next, we next investigated how the length of the post-excision window affects tissue viability and HS-27 fluorescence. In a previous clinical study, we determined that the time between biopsy excision and Hsp90 imaging can be as long as 10 minutes [24,25]; therefore, we assessed the impact of 1-, 3-, and 10-minute post-excision on PI and HS-27 fluorescence using biopsies obtained from mice bearing 4T1 murine breast tumors. Representative HS-27 and PI fluorescence images of H2O2(-) and H2O2(+) 4T1 biopsies with 1-, 3-, and 10-minute post-excision windows are shown in Fig. 2(A). A slight increase in HS-27 and PI fluorescence is observed in the 10-minute post-excision window as compared to the 1-, and 3- post-excision windows; however, as indicated in Fig. 2(B), there were no significant differences in HS-27 or PI fluorescence among all post-excision groups, indicating that the tissue remains viable within a post-excision window of 10 minutes. Survival curves in Fig. 2(C) show that there are significant changes in both HS-27 and PI signals between viable (H2O2(-)) and non- viable tissue (H2O2 (+)) for each post-excision window.
Fig. 2. Tissue remains viable ex vivo within a 10-minute post-excision window. Ex vivo 4T1 (murine mammary tumor) cells were stained with either 100 µM HS-27 (green fluorescence) for 1 minute or Propidium Iodide (PI) (red fluorescence) for 15 minutes. Tissue degradation was induced by incubation with 35 mM hydrogen peroxide (H2O2) for 15 minutes. Images were acquired with conventional fluorescence microscopy with a 488 nm laser and a 525 ± 7 nm emission filter for HS-27 and a 525 nm laser and a 615 ± 7 nm emission filter for PI. A) Representative images of H2O2 (-) (left) and H2O2 (+) (right) ex vivo biopsy stained with HS-27 (left two panels) and PI (right two panels) with a 1-, 3-, and 10-minute post-excision window (n=5). B) Comparison of survival curves of H2O2 (-) HS-27 fluorescence (top) and H2O2(-) PI fluorescence (bottom) for different 1-, 3-, and 10-minute post-excision windows. There were no statistically significant differences among 1-, 3-, and 10-minute post excision windows in both HS-27 and PI uptakes. C) Survival curves of PI signals (top) and HS-27 signals (bottom) for 1-, 3-, and 10-minute post-excision windows, treated with (red) and without (black) hydrogen peroxide. There were statistically significant increases in both HS-27 and PI uptake between H2O2 (-) and H2O2 (+) groups. Statistical significance between survival curves was determined by the Kolmogorov-Smirnov (KS) test. For all groups n = 5 biopsies. Survival curves show the mean ± SEM.
3.2 Non-specific diffusion into the tissue traps increases non-specific fluorescence
Another factor that may influence fluorescence imaging of ex vivo tissue specimens using HS-27 is fluorescence from HS-27 that nonspecifically diffuses superficially beneath the tissue surface that cannot be simply removed by washing. To quantify staining penetration depth as a function of staining concentration and time, we investigated the diffusion kinetics of both an Hsp90-specific fluorescence probe (HS-27) and an inactivated analog (HS-217) using a two-compartment transport model of diffusion [28,32]. This is particularly important with respect to the experimental protocol that should be used (staining mechanism, time, and concentration), and the type of imaging device used. Since HS-27 specifically binds to tumor cells, it should have a shorter depth of penetration in tissue than HS-217. The setup and methodology of the two-compartment diffusion model are shown in Fig. 3(A). Representative HS-27 and HS-217 fluorescence images captured at the top surface of the tumor slice over 100 minutes are shown in Fig. 3(B) and were used to create curves for HS-27 or HS-217 diffusion through tumor tissue over time. Representative diffusion curves are shown in Fig. 3(C) and Fig. 3(D). The diffusion coefficients of HS-217 and HS-27 in the water were estimated to be 1.5 × 10-6 cm2/sec based on their molecular weights (1021.20 and 993.14) and an average molecular radius of 1.5nm [17,25] using Stokes-Einstein equation. The diffusion coefficients of HS-217 and HS-27 in the tumor were computed as 11.55 ± 3.63 E-7 cm2/sec and 6.72 ± 1.14 E-7 cm2/sec, respectively, demonstrating an approximately two-fold increase in diffusion rate for non-specific HS-217. The 1-minute penetration depth (used in the clinic) of HS-217 and HS-27 based on the diffusion coefficient were 366.75 ± 46.96 µm and 248.39 ± 25.12 µm, respectively. The molecular weight of HS-27 and HS-217 is comparable; HS-27 is 993.14 and HS-217 is 1021.20 g/mol.
Fig. 3. Non-specific fluorophores diffuse deeper in tissue than tumor-specific fluorophores. A) Scheme of diffusion study design. Thin slices (0.8 to 1.3 mm) of murine mammary tumor (4T1) tissue (n=4) were placed on top of a porous polycarbonate membrane in contact with 100 µM of HS-27 or HS-217 for 100 minutes. A customized fluorescence microscope with a 488 nm laser and a 525 ± 7 nm emission filter was used to image the top surface of the tumor specimen over 4 hours. The average fluorescence intensity for each time point was calculated and fit into the two-compartment transport model. B) Representative HS-27 and HS-217 fluorescence images over time with background subtraction from the 0-minute fluorescence image. C) Representative HS-217 curve fitting results from the two-compartment diffusion model. The diffusion coefficient (DT) of HS-217 was computed as 11.55 ± 3.63 E-7 cm2/sec. D) Representative HS-27 curve fitting results from the two-compartment diffusion model. The diffusion coefficient of HS-27 (DT) in the tumor was computed as 6.72 ± 1.14 E-7 cm2/sec. R2 and p-values were determined by the least-squares regression.
To validate HS-27 and HS-217 diffusion coefficients computed from the two-compartment diffusion model, the modeled 1-minute penetration depths of HS-27 and HS-217 were compared to the actual penetration depths of both probes in 4T1 murine breast tumors stained with 100 µM HS-27 (n = 5) or HS-217 (n = 6) for 1 minute. Stained tumor specimens were flash frozen and sectioned into 30-µm slices. Representative images of HS-217 and HS-27 sections are shown in Fig. 4(A). The average 1-minute staining penetration depths of HS-27 and HS-217 were 194.55 ± 56.95 µm and 328.45 ± 85.54 µm, respectively. Figure 4(B) illustrates a box-whisker plot of the modeled and sectioned HS-27 and HS-217 penetration depths from individual samples. For both modeled and sectioned penetration depths, a statistical significance is observed between HS-27 staining and HS-217 staining (p<0.05). For both HS-27 and HS-217, no statistically significant difference was found between modeled and sectioned penetration depths. Student's t-test was used to determine the statistical significance.
Fig. 4. Modeled HS-27 and HS-217 penetration depths are similar to sectioned penetration depths. Ex vivo 4T1 murine breast tumor specimens were stained with either 100 µM HS-27 or 100 µM HS-217 for 1 minute. The specimens were immediately flash frozen and sectioned into 30 µm slices. Five field-of-views (FOVs) of each slice were acquired with conventional fluorescence microscopy. A) Representative images of the section of HS-217 (top) and HS-27 (bottom) stained tumor specimen B) The averaged penetration depths were determined from each specimen and compared to the modeled penetration depths. For both modeled and sectioned penetration depths, HS-27 is statistically different from HS-217 (p-value < 0.01). For either HS-27 or HS-217, there were no statistically significant differences between modeled and sectioned penetration depths. ANOVA with Student’s t-test as post-hoc test was used for statistical analysis. * indicates a p-value of less than 0.05.
3.3 Contact-based imaging introduces more variability in fluorescence imaging
In previous clinical studies [23–25], a fiber-based micro endoscope with a field of view of 750 µm [30] was used to image surface HS-27 fluorescence of human breast CNBs. The micro endoscope was placed in contact with the biopsies during the imaging protocol and then lifted and moved across the sample to acquire an image from the entire 10-20 mm biopsy specimen [33]. We compared the contact and non-contact imaging approaches and studied the factors that could potentially affect the variability of HS-27 fluorescence. To evaluate the impact of contact-based imaging strategies, we performed non-contact imaging using a conventional fluorescence microscope and compared the results to that of the micro endoscope. Figure 5(A) shows representative images of a tumor-mimicking HS-27 phantom imaged with each system. The image obtained with the non-contact system shows finer details within the phantom than the contact system. Figure 5(B) shows representative images of HS-27- and HS-217-stained biopsies taken with both the contact and non-contact systems. Qualitatively, HS-27-stained biopsies showed greater fluorescence than HS-217 stained biopsies with both systems, as expected. The box-whisker plot, shown in Fig. 5(C), shows that there is a significant difference between HS-27 and HS-217 mean fluorescence for contact vs. non-contact imaging; however, the p-value for the for non-contact imaging was two orders of magnitude better than that for the contact imaging system (0.005 vs 0.03, respectively). Additionally, contact imaging showed a four-fold increase in variance for both HS-217 and HS-27 fluorescence compared to non-contact imaging (coefficient of variation of 0.416 for HS-27 and 0.207 for HS-217 for contact imaging vs 0.095 for HS-27 and 0.057 for HS-217 for the non-contact imaging) shown in Table 1.
Fig. 5. Contact imaging introduces more variability in fluorescence signals compared to non-contact imaging. A) Representative images of tumor-mimicking HS-27 phantoms captured with a fiber-based contact imaging system (top panel) or a conventional non-contact fluorescence microscope (bottom panel) B) Ex vivo 4T1 (murine mammary tumor) tissue samples stained with either 100 µM HS-27 (left) or HS-217 (right) for 1 minute were imaged with either a fiber-based contact imaging system (top panel) or a conventional non-contact fluorescence microscope (bottom panel). The scale bars are 500 and 200 µm, respectively. C) Distribution of means of HS-27 and HS-217 of both imaging systems. The p-value between HS-27 and HS-217 for the contact imaging system was 0.03 and it was 0.005 for the non-contact imaging system. The coefficient of variation of HS-27 for non-contact and contact imaging were 0.095 and 0.416, respectively. The coefficient of variation of HS-217 for non-contact and contact imaging were 0.057 and 0.207, respectively. Statistical significance was determined by Student’s t-test. * indicates a p-value of less than 0.05 and ** indicates a p-value of less than 0.01.
Table 1. Statistics of HS-27 and HS-217 for Contact and Non-contact Imaging Systems
4. Discussion
Hsp90 is ubiquitously overexpressed on the surface of breast cancer cells across disease stages and receptor subtypes [20,21,18,17]. Our group has previously demonstrated the diagnostic value of imaging Hsp90 expression on ex vivo biopsies using the FITC-tethered Hsp90 inhibitor HS-27 [24,25]. In our previous clinical study, contrast between cancer and benign tissue spanned a wide range. We hypothesized that this was due to systematic errors from both staining and imaging. In this study, we investigated three factors that could influence the potential sources of variability in ex vivo Hsp90 molecular imaging: reduced tissue viability, non-specific diffusion, and contact-based imaging. Together we used these three factors as the evaluation metrics to systematically examine the efficacy of Hsp90 imaging of breast tissue biopsies. From this study, we have established that decreased cell viability in degraded tissue, non-specific diffusion of Hsp90, and contact-based imaging systems with limited FOVs contribute to non-specific fluorescence in ex vivo imaging. Before translating Hsp90 imaging into a rapid point-of-care diagnostic tool, we must carefully control post-excision time to minimize tissue degradation, optimize staining time and concentration to limit diffusion, and develop a portable, non-contact strategy for imaging the entire specimen.
Other groups have evaluated the efficacy of ex vivo tissue staining using similar approaches. Johnson et al. performed viability analysis using a vital dye on excised tissue imaged with confocal microscopy and found that compromised viability undermined the quality of staining, and image collection [34]. Consistent with those results, we found that compromised viability led to increased HS-27 fluorescence, resulting in false representation of surface Hsp90 expression. In our tissue viability study, we used hydrogen peroxide to induce cell death in 4T1 tissue biopsies serving as the positive control of the experiments and applied propidium iodide (PI) to specimens to assess its effect of viability on Hsp90 staining in tissue. We demonstrated that excised biopsy specimens remain viable for quantitative Hsp90 imaging within a 10-minute post excision window, ensuring the feasibility of Hsp90 imaging in a clinical setting. We also recognized that, for viability control in our experimental design, hydrogen peroxide is not the only method to compromise cell viability in tissue; other groups have reported that a chemical induction method, such as staurosporine, can also be used as a viability challenge in vitro and ex vivo [35].
Other studies that evaluated ex vivo tissue staining protocols also mentioned the relevance of probe diffusion to the uniformity of staining in a qualitative way [36,37]. In our diffusion study, we established a mathematical diffusion model based on previous work [28,32] that input staining time and concentration to provide a systematic way to evaluate the uniformity and estimate staining penetration depth with different staining protocols. A shorter staining penetration depth indicated less unspecific staining from ex vivo tissue and the mathematical diffusion model guided design of staining protocols. In our diffusion study, many factors challenged the accuracy of the diffusion model. Although the experimental setup only allowed diffusion occurring along the z-axis, we observe diffusion non-uniformly along the x-axis and y-axis. Hotspots of fluorescence signals were observed in specific regions of the tumor specimen, indicating that there were variabilities in diffusion kinetics within the specimen. One possible explanation is that tissue heterogeneity may contribute to the fluorescence variability and hence influence the prediction of our computational model. More advanced diffusion models can be used in future studies to further investigate heterogeneous diffusion.
Previously, our group used a fiber-based optical system to acquire Hsp90 fluorescence from the surface of CNBs. By comparing the contact and non-contact imaging systems, we demonstrated that non-contact imaging reduces the variance and improves upon the contrast of ex vivo molecular imaging. One hypothesis for this improvement in imaging is that contact imaging introduces variable pressure to the ex vivo tissue specimen during probe placement. Previous studies have shown that probe pressure can alter fluorescence signal. [38,39] Therefore, the significantly higher variability of contact imaging system may make it challenging to differentiate ER/PR+ tumor and TNBC tumor specimens (which have a weaker Hsp90 fluorescence than HER2+ breast cancer) and benign tissues. We will therefore focus future efforts on developing a non-contact imaging system for Hsp90 molecular imaging. By incorporating a wide-field, high-resolution, non-contact imaging system with tumor-specific molecular probes, we hope to improve the sensitivity of our Hsp90 imaging platform to allow more feasibility for clinical translation to margin assessment.
The use of our Hsp90 platform will not be limited to breast cancer diagnosis. Elevated Hsp90 expression has been associated with poor prognostic outcome and lower survival rate of breast cancer patients [16,20,40], indicating a potential use for our diagnostic platform to provide prognostic information. Additionally high Hsp90 expression has been shown to correlate with TNM cancer staging [41], indicating the potential of Hsp90 to inform prognosis and clinicopathological parameters. Moreover, high Hsp90 expression has been discovered in other types of cancer, such as melanoma [42], leukemia [43], and prostate cancer [44]. Elevated Hsp90 expression was also found in the serum of non-small cell lung cancer patients, and the Hsp90 inhibitor AT13387 has been under clinical trial phase I/II for treating stage IV and recurrent non-small lung cancer patients [45,46]. Another type of Hsp90 inhibitor has been reported as part of a potential combination therapy against colorectal cancer [47].
An Hsp90-based imaging platform has the potential to provide rapid point of care diagnosis of breast cancer diagnoses in LMICs. Breast cancer is the most commonly diagnosed cancer in women and is the second leading cause of cancer mortality worldwide [48], with low and middle-income countries (LMICs) bearing more than half of the breast cancer burden [49–51]. Most women in LMICs present with late-stage disease, which is associated with poor prognosis and high mortality rates [51,52]. Delays in diagnosis in LMICs can be attributed to both a shortage of specialists, including radiologists and pathologists, and equipment [49,50]. Therefore, there is an unmet need for a standardized, automated procedure to image and diagnose breast CNBs. Our Hsp90 imaging strategy is primed for combination with low-cost, point-of-care imaging technologies such as ultrasound [53] as an alternative to traditional pathology analysis in LMICs.
In this study, we have carefully evaluated the sources of non-specific fluorescence in Hsp90 staining and imaging protocols to facilitate clinical translation. Our imaging platform will potentially serve as an enhancement to traditional pathological examination used in margin assessment to reduce the turn-around time during BCS and minimize the rate of second surgeries. Additionally, portability and low-cost features of our platform could ultimately be used in the low-resource settings to address a lack of resource to perform pathology for breast cancer diagnostics.
Funding
National Institute of Biomedical Imaging and Bioengineering (5R01EB028148-02, 5R21EB025008-02).
Acknowledgements
This work was supported by generous funding from NIH (5R21EB025008-02, 5R01EB028148-02). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Disclosures
Dr. Ramanujam has founded a company called Zenalux Biomedical and she and other team members have developed technologies related to this work where the investigators or Duke may benefit financially if this system is sold commercially.
Data Availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. M. C. Lee, K. Rogers, K. Griffith, K. A. Diehl, T. M. Breslin, V. M. Cimmino, A. E. Chang, L. A. Newman, and M. S. Sabel, “Determinants of breast conservation rates: reasons for mastectomy at a comprehensive cancer center,” Breast J. 15(1), 34–40 (2009). [CrossRef]
2. M. Morrow, R. Jagsi, A. K. Alderman, J. J. Griggs, S. T. Hawley, A. S. Hamilton, J. J. Graff, and S. J. Katz, “Surgeon recommendations and receipt of mastectomy for treatment of breast cancer,” JAMA 302(14), 1551–1556 (2009). [CrossRef]
3. R. G. Pleijhuis, M. Graafland, J. de Vries, J. Bart, J. S. de Jong, and G. M. van Dam, “Obtaining adequate surgical margins in breast-conserving therapy for patients with early-stage breast cancer: current modalities and future directions,” Ann. Surg. Oncol. 16(10), 2717–2730 (2009). [CrossRef]
4. M. S. Moran, S. J. Schnitt, A. E. Giuliano, J. R. Harris, S. A. Khan, J. Horton, S. Klimberg, M. Chavez-MacGregor, G. Freedman, N. Houssami, P. L. Johnson, and M. Morrow, “Society of Surgical Oncology-American Society for Radiation Oncology consensus guideline on margins for breast-conserving surgery with whole-breast irradiation in stages I and II invasive breast cancer,” Int. J. Radiat. Oncol., Biol., Phys. 88(3), 553–564 (2014). [CrossRef]
5. A. Qorbani, F. Fereidouni, R. Levenson, S. Y. Lahoubi, Z. T. Harmany, A. Todd, and M. A. Fung, “Microscopy with ultraviolet surface excitation (MUSE): A novel approach to real-time inexpensive slide-free dermatopathology,” J. Cutan. Pathol. 45(7), 498–503 (2018). [CrossRef]
6. A. K. Glaser, N. P. Reder, Y. Chen, E. F. McCarty, C. Yin, L. Wei, Y. Wang, L. D. True, and J. T. C. Liu, “Light-sheet microscopy for slide-free non-destructive pathology of large clinical specimens,” Nat. Biomed. Eng. 1(7), 0084 (2017). [CrossRef]
7. N. Nishio, N. S. van den Berg, S. van Keulen, B. A. Martin, S. Fakurnejad, N. Teraphongphom, S. U. Chirita, N. J. Oberhelman, G. Lu, C. E. Horton, M. J. Kaplan, V. Divi, A. D. Colevas, and E. L. Rosenthal, “Optical molecular imaging can differentiate metastatic from benign lymph nodes in head and neck cancer,” Nat. Commun. 10(1), 5044 (2019). [CrossRef]
8. R. W. Gao, N. T. Teraphongphom, N. S. van den Berg, B. A. Martin, N. J. Oberhelman, V. Divi, M. J. Kaplan, S. S. Hong, G. Lu, R. Ertsey, W. Tummers, A. J. Gomez, F. C. Holsinger, C. S. Kong, A. D. Colevas, J. M. Warram, and E. L. Rosenthal, “Determination of tumor margins with surgical specimen mapping using near-infrared fluorescence,” Cancer Res. 78(17), 5144–5154 (2018). [CrossRef]
9. K. S. Samkoe, B. D. Bates, J. T. Elliott, E. LaRochelle, J. R. Gunn, K. Marra, J. Feldwisch, D. B. Ramkumar, D. F. Bauer, K. D. Paulsen, B. W. Pogue, and E. R. Henderson, “Application of fluorescence-guided surgery to subsurface cancers requiring wide local excision:literature review and novel developments toward indirect visualization,” Cancer Control 25(1), 107327481775233 (2018). [CrossRef]
10. A. R. Pantel and D. A. Mankoff, “Molecular imaging to guide systemic cancer therapy: Illustrative examples of PET imaging cancer biomarkers,” Cancer Lett. 387, 25–31 (2017). [CrossRef]
11. T. Hussain and Q. T. Nguyen, “Molecular imaging for cancer diagnosis and surgery,” Adv. Drug Delivery Rev. 66, 90–100 (2014). [CrossRef]
12. H. Lee, W. Akers, K. Bhushan, S. Bloch, G. Sudlow, R. Tang, and S. Achilefu, “Near-infrared pH-activatable fluorescent probes for imaging primary and metastatic breast tumors,” Bioconjugate Chem. 22(4), 777–784 (2011). [CrossRef]
13. B. W. Pogue, E. L. Rosenthal, S. Achilefu, and G. M. van Dam, “Perspective review of what is needed for molecular-specific fluorescence-guided surgery,” J. Biomed. Opt. 23(10), 100601 (2018). [CrossRef]
14. W. S. Tummers, J. M. Warram, N. S. van den Berg, S. E. Miller, R. J. Swijnenburg, A. L. Vahrmeijer, and E. L. Rosenthal, “Recommendations for reporting on emerging optical imaging agents to promote clinical approval,” Theranostics 8(19), 5336–5347 (2018). [CrossRef]
15. B. W. Maloney, D. M. McClatchy, B. W. Pogue, K. D. Paulsen, W. A. Wells, and R. J. Barth, “Review of methods for intraoperative margin detection for breast conserving surgery,” J. Biomed. Opt. 23(10), 1–19 (2018). [CrossRef]
16. Q. Cheng, J. T. Chang, J. Geradts, L. M. Neckers, T. Haystead, N. L. Spector, and H. K. Lyerly, “Amplification and high-level expression of heat shock protein 90 marks aggressive phenotypes of human epidermal growth factor receptor 2 negative breast cancer,” Breast Cancer Res. 14(2), R62 (2012). [CrossRef]
17. J. J. Barrott, P. F. Hughes, T. Osada, X. Y. Yang, Z. C. Hartman, D. R. Loiselle, N. L. Spector, L. Neckers, N. Rajaram, F. Y. Hu, N. Ramanujam, G. Vaidyanathan, M. R. Zalutsky, H. K. Lyerly, and T. A. Haystead, “Optical and Radioiodinated Tethered Hsp90 Inhibitors Reveal Selective Internalization of Ectopic Hsp90 in Malignant Breast Tumor Cells,” Chem. Biol. 20(9), 1187–1197 (2013). [CrossRef]
18. J. J. Barrott and T. A. J. Haystead, “Hsp90, an unlikely ally in the war on cancer,” FEBS J. 280(6), 1381–1396 (2013). [CrossRef]
19. L. B. Crowe, P. F. Hughes, D. A. Alcorta, T. Osada, A. P. Smith, J. Totzke, D. R. Loiselle, I. D. Lutz, M. Gargesha, D. Roy, J. Roques, D. Darr, H. K. Lyerly, N. L. Spector, and T. A. J. Haystead, “A fluorescent Hsp90 probe demonstrates the unique association between extracellular Hsp90 and malignancy in vivo,” ACS Chem. Biol. 12(4), 1047–1055 (2017). [CrossRef]
20. L. Whitesell and S. L. Lindquist, “HSP90 and the chaperoning of cancer,” Nat. Rev. Cancer 5(10), 761–772 (2005). [CrossRef]
21. F. Zagouri, E. Bournakis, K. Koutsoukos, and C. A. Papadimitriou, “Heat shock protein 90 (hsp90) expression and breast cancer,” Pharmaceuticals 5(9), 1008–1020 (2012). [CrossRef]
22. F. Zagouri, T. N. Sergentanis, A. Nonni, C. A. Papadimitriou, N. V. Michalopoulos, P. Domeyer, G. Theodoropoulos, A. Lazaris, E. Patsouris, E. Zogafos, A. Pazaiti, and G. C. Zografos, “Hsp90 in the continuum of breast ductal carcinogenesis: evaluation in precursors, preinvasive and ductal carcinoma lesions,” BMC Cancer 10(1), 353 (2010). [CrossRef]
23. B. T. Crouch, J. Duer, R. J. Wang, J. Gallagher, A. Hall, M. S. Soo, P. Hughes, T. A. J. Haystead, and N. Ramanujam, “Optimizing fluorescently-tethered Hsp90 inhibitor dose for maximal specific uptake by breast tumors,” Proc. SPIE 10478, 1047804 (2018). [CrossRef]
24. B. T. Crouch, J. Gallagher, R. Wang, J. Duer, A. Hall, M. S. Soo, P. Hughes, T. Haystead, and N. Ramanujam, “Exploiting heat shock protein expression to develop a non-invasive diagnostic tool for breast cancer,” Sci. Rep. 9(1), 3461 (2019). [CrossRef]
25. B. Crouch, H. Murphy, S. Belonwu, A. Martinez, J. Gallagher, A. Hall, M. S. Soo, M. Lee, P. Hughes, T. Haystead, and N. Ramanujam, “Leveraging ectopic Hsp90 expression to assay the presence of tumor cells and aggressive tumor phenotypes in breast specimens,” Sci. Rep. 7(1), 17487 (2017). [CrossRef]
26. J. Xiang, C. Wan, R. Guo, and D. Guo, “Is hydrogen peroxide a suitable apoptosis inducer for all cell types?” BioMed Res. Int. 2016, 1–6 (2016). [CrossRef]
27. C. Zhu, A. F. Martinez, H. L. Martin, M. Li, B. T. Crouch, D. A. Carlson, T. A. J. Haystead, and N. Ramanujam, “Near-simultaneous intravital microscopy of glucose uptake and mitochondrial membrane potential, key endpoints that reflect major metabolic axes in cancer,” Sci. Rep. 7(1), 13772 (2017). [CrossRef]
28. O. Chuchuen, J. R. Maher, M. G. Simons, J. J. Peters, A. P. Wax, and D. F. Katz, “Label-free measurements of tenofovir diffusion coefficients in a microbicide gel using Raman spectroscopy,” J. Pharm. Sci. 106(2), 639–644 (2017). [CrossRef]
29. N. Rajaram, A. F. Reesor, C. S. Mulvey, A. E. Frees, and N. Ramanujam, “Non-invasive, simultaneous quantification of vascular oxygenation and glucose uptake in tissue,” PLoS One 10(1), e0117132 (2015). [CrossRef]
30. T. J. Muldoon, M. C. Pierce, D. L. Nida, M. D. Williams, A. Gillenwater, and R. Richards-Kortum, “Subcellular-resolution molecular imaging within living tissue by fiber microendoscopy,” Opt. Express 15(25), 16413–16423 (2007). [CrossRef]
31. J. L. Mueller, J. E. Gallagher, R. Chitalia, M. Krieger, A. Erkanli, R. M. Willett, J. Geradts, and N. Ramanujam, “Rapid staining and imaging of subnuclear features to differentiate between malignant and benign breast tissues at a point-of-care setting,” J. Cancer Res. Clin. Oncol. 142(7), 1475–1486 (2016). [CrossRef]
32. O. Chuchuen, J. R. Maher, M. H. Henderson, M. Desoto, L. C. Rohan, A. Wax, and D. F. Katz, “Label-free analysis of tenofovir delivery to vaginal tissue using co-registered confocal Raman spectroscopy and optical coherence tomography,” PLoS One 12(9), e0185633 (2017). [CrossRef]
33. C. M. Focke, T. Decker, and P. J. van Diest, “The reliability of histological grade in breast cancer core needle biopsies depends on biopsy size: a comparative study with subsequent surgical excisions,” Histopathology 69(6), 1047–1054 (2016). [CrossRef]
34. S. Johnson and P. Rabinovitch, “Ex vivo imaging of excised tissue using vital dyes and confocal microscopy,” Curr. Protoc. Cytom .Chapter 9, 39 (2012). [CrossRef]
35. L. Hofmann, D. Hose, A. Grießhammer, R. Blum, F. Döring, S. Dib-Hajj, S. Waxman, C. Sommer, E. Wischmeyer, and N. Üçeyler, “Characterization of small fiber pathology in a mouse model of Fabry disease,” eLife 7, e39300 (2018). [CrossRef]
36. J. Bini, J. Spain, K. Nehal, M. Rajadhyaksha, V. Hazelwood, and C. DiMarzio, “Confocal mosaicing microscopy of human skin ex vivo: spectral analysis for digital staining to simulate histology-like appearance,” J. Biomed. Opt. 16(7), 076008 (2011). [CrossRef]
37. M. R. Ankle and P. S. Joshi, “A study to evaluate the efficacy of xylene-free hematoxylin and eosin staining procedure as compared to the conventional hematoxylin and eosin staining: An experimental study,” J. Oral Maxillofac. Pathol. 15(2), 161–167 (2011). [CrossRef]
38. L. Lim, B. Nichols, N. Rajaram, and J. Tunnell, “Probe pressure effects on human skin diffuse reflectance and fluorescence spectroscopy measurements,” J. Biomed. Opt. 16(1), 011012 (2011). [CrossRef]
39. A. Nath, K. Rivoire, S. Chang, D. Cox, E. Atkinson, M. Follen, and R. Richards-Kortum, “Effect of probe pressure on cervical fluorescence spectroscopy measurements,” J. Biomed. Opt. 9(3), 523–533 (2004). [CrossRef]
40. L. Zhong, X. Peng, G. E. Hidalgo, D. E. Doherty, A. J. Stromberg, and E. A. Hirschowitz, “Antibodies to HSP70 and HSP90 in serum in non-small cell lung cancer patients,” Cancer Detect. Prev. 27(4), 285–290 (2003). [CrossRef]
41. K. Kryeziu, J. Bruun, T. K. Guren, A. Sveen, and R. A. Lothe, “Combination therapies with HSP90 inhibitors against colorectal cancer,” Biochim. Biophys. Acta, Rev. Cancer 1871(2), 240–247 (2019). [CrossRef]
42. A. C. Society, “Cancer Facts and Figures 2019,” (2019).
43. S. Dey, “Preventing breast cancer in LMICs via screening and/or early detection: The real and the surreal,” WJCO 5(3), 509–519 (2014). [CrossRef]
44. M. M. Rivera-Franco and E. Leon-Rodriguez, “Delays in breast cancer detection and treatment in developing countries,” Breast Cancer: Basic Clin. Res. 12, 1178223417752677 (2018). [CrossRef]
45. L. A. Gutnik, B. Matanje-Mwagomba, V. Msosa, S. Mzumara, B. Khondowe, A. Moses, R. E. Kohler, L. A. Carey, C. N. Lee, and S. Gopal, “Breast cancer screening in low- and middle-income countries: a perspective from Malawi,” JGO 2(1), 4–8 (2016). [CrossRef]
46. S. Sayed, R. Lukande, and K. A. Fleming, “Providing pathology support in low-income countries,” JGO 1(1), 3–6 (2015). [CrossRef]
47. S. K. Vashist, P. B. Luppa, L. Y. Yeo, A. Ozcan, and J. H. T. Luong, “Emerging technologies for next-generation point-of-care testing,” Trends Biotechnol. 33(11), 692–705 (2015). [CrossRef]
48. E. Pick, Y. Kluger, J. M. Giltnane, C. Moeder, R. L. Camp, D. L. Rimm, and H. M. Kluger, “High HSP90 expression is associated with decreased survival in breast cancer,” Cancer Res. 67(7), 2932–2937 (2007). [CrossRef]
49. M. C. Diehl, M. O. Idowu, K. Kimmelshue, T. P. York, L. W. Elmore, and S. E. Holt, “Elevated expression of nuclear Hsp90 in invasive breast tumors,” Cancer Biol. Ther. 8(20), 1952–1961 (2009). [CrossRef]
50. B. Becker, G. Multhoff, B. Farkas, P. J. Wild, M. Landthaler, W. Stolz, and T. Vogt, “Induction of Hsp90 protein expression in malignant melanomas and melanoma metastases,” Exp. Dermatol. 13(1), 27–32 (2004). [CrossRef]
51. P. Flandrin, D. Guyotat, A. Duval, J. Cornillon, E. Tavernier, N. Nadal, and L. Campos, “Significance of heat-shock protein (HSP) 90 expression in acute myeloid leukemia cells,” Cell Stress Chaperones 13(3), 357–364 (2008). [CrossRef]
52. E. F. Burgess, A.-J. L. Ham, D. L. Tabb, D. Billheimer, B. J. Roth, S. S. Chang, M. S. Cookson, T. J. Hinton, K. L. Cheek, S. Hill, and J. A. Pietenpol, “Prostate cancer serum biomarker discovery through proteomic analysis of alpha-2 macroglobulin protein complexes,” Proteomics: Clin. Appl. 2(9), 1223–1233 (2008). [CrossRef]
53. N. C. Institute, “Erlotinib hydrochloride and Hsp90 Inhibitor AT13387 in treating patients with recurrent or metastatic non-small cell lung cancer”, retrieved 01/04, 2016, https://clinicaltrials.gov/ct2/show/NCT02535338?term=hsp90+inhibitor+AND+cancer&rank=4.
