Targeted Hsp70 fluorescence molecular endoscopy detects dysplasia in Barrett’s esophagus

As recent data indicated that Hsps play a key role in EAC development, we analyzed Hsp70 by immunohistochemistry (IHC) in 12 patients with BE, low-grade dysplasia (LGD), and/or EAC. A significantly higher level of Hsp70 expression was found in human EAC (2.67 ± 0.21 A.U.) as compared to LGD (1.38 ± 0.18 A.U.) and most importantly non-dysplastic BE (0.22 ± 0.15 A.U., Fig. 1A, C). Similarly, we observed strong Hsp70 expression in the transgenic L2-IL1b mouse model, evaluated by counting positive cells in BE region (Fig. 1B). At 9–12 months of age, L2-IL1b mice develop low-grade to high-grade dysplastic lesions in the setting of BE-like metaplasia at the squamocolumnar junction (SCJ), enabling us to study the associated changes [11, 33]. Notably, patients (p = 0.0001) and L2-IL1b mice showed a statistically significant increase in the Hsp70 expression during stepwise progression from non-dysplastic BE to LGD to high-grade dysplasia (HGD) and EAC (Fig. 1C and D), suggesting that Hsp70 expression correlates closely with disease progression.

We next evaluated Hsp70 expression and uptake of Hsp70-TPP in patient-derived 3D organotypic organoids isolated from BE, LGD, or HGD/EAC. In line with the IHC findings, we found an increased expression of mHsp70 in organoids derived from HGD/EAC (> 70%) in comparison with organoids derived from LGD (33%) and BE without dysplasia (36%) (Fig. 2A). Given our in vitro findings suggesting Hsp70 as a potential viable therapeutic target, we investigated potential toxic effects of Hsp70-TPP by treating BE organoids for 48 h with Hsp70-TPP labeled with carboxyfluorescein (CF). No adverse effects on organoid growth or proliferation were detected in these experiments (Fig. 2B).

We then performed time-lapse microscopy of human organoids from a patient with dysplastic BE after treatment with 100 µg/ml Hsp70-TPP-CF (Fig. 2C, Movie S1 A and B). Internalization of Hsp70-TPP in the epithelial cell layer of the organoid could be observed within 10 min after incubation (Fig. 2C, Fig. S1). Quantification of the fluorescence signal in the organoid (Fig. 2D) revealed a progressive signal increase within the organoid after the treatment. We excluded autofluorescence of Matrigel by acquiring pre- and post-treatment images of the FITC channel (Fig. 2E). A fast Hsp70-TPP uptake was also observed in human organoids from LGD after an incubation for 10 min with 200 µg/ml Hsp70-TPP-CF (Fig. S1), consistent with the Hsp70 expression in LGD organoids assessed by flow cytometry. Taken together, these results confirm membrane expression of Hsp70 by dysplastic esophageal cells within organoids and indicate that Hsp70-TPP produce a strong epithelial cell-specific signal without cytotoxicity.

Based on the findings that local Hsp70 expression increased gradually during esophageal carcinogenesis and with the demonstrated ability of Hsp70-TPP to penetrate organoids, we studied the utility of labeled Hsp70-TPP to diagnose dysplasia in our BE mouse model during endoscopy. Hsp70-TPP-Cy5.5 was injected intravenously (i.v.) into L2-IL1b mice (6, 9, and 12 months old) 24 h prior to post-euthanasia fluorescence upper endoscopy. Fluorescence endoscopy was performed using an in-house custom-made endoscopy system (0.8-mm flexible fiberscope with 6000 pixels imaging plane, see Fig. 3 for other details and Movie S2). Following endoscopy, fluorescence imaging of the excised stomach and esophagus was performed (Fig. 3). Hsp70-targeted fluorescent signal accumulated predominantly in dysplastic lesions at the SCJ (Fig. 4A) and the lesions were also visible in the near-infrared (NIR) channel (Fig. 4A, middle panel). Overlay images (Fig. 4A, bottom panel) showed that Hsp70-TPP-Cy5.5 accumulated in low-, mid-, and mid-/high-grade dysplasia lesions at the SCJ but not metaplastic areas, indicating that fluorescence endoscopy could indeed detect Hsp70 expression even at early dysplastic stage. No NIR signal was detected in a non-injected control IL1b mouse (data not shown). In addition, ex vivo wide-field imaging of the excised stomach and esophagus showed a specific uptake of the Hsp70-TPP-Cy.5.5 specifically in dysplastic lesions of L2-IL1b mice. Of note, the signal intensity was extremely low in the surrounding non-dysplastic epithelium, confirming the specificity of fluorescence endoscopy (Fig. 4B, arrows). Moreover, we observed an increased median target/tumor-to-background ratio (TBR) by endoscopy in injected L2-IL1B mice with mid-/high-grade lesions compared with low-grade lesions (Fig. 4C). However, endoscopy (p = 0.6057) and wide-field imaging (p = 0.4871) showed no statistically significant difference between the TBR of L2-IL1b mice with mid-/high-grade and early/low-grade lesions by t test (Fig. 4C). Therefore, we conclude that Hsp70-TPP-Cy5.5 can specifically label different stages of dysplastic lesions but not metaplasia or normal tissue in a pre-clinical setting and can be successfully detected using fluorescence molecular endoscopy.

In order to translate the benefits of Hsp70-specific dysplasia detection into the clinical setting, we investigated the performance of the labeled Hsp70-TPP peptide applied topically to primary human specimens. In preparation to this experiment, we validated the binding of Hsp70-TPP in a xenograft subcutaneous mouse model. In this model, specific binding of Hsp70-TPP-Cy5.5 to the exposed tumor cells was observed following topical administration of the peptide (Fig. S2A). The mean penetration depth of the peptide-tracer, as measured on 1-mm slices from the tumors, was 0.52 ± 0.05 mm (Figure S2) after 5 min of application. Next, Hsp70-TPP-Cy5.5 was applied topically to primary resected human esophageal specimens from two EAC patients undergoing esophagectomy without prior chemotherapy. Hsp70-TPP-Cy5.5 was topically applied to the resected fresh tissue, prior to formalin fixation, including tumor (ROI1, Fig. 5A, B upper panel, and C) and non-tumor areas (ROI2 and ROI3, Fig. 5A). After 5 min of incubation, the quantified fluorescence signal (Fig. 5B, upper panel, right) showed a higher fluorescence intensity in tumor areas (ROI1) compared with non-tumorigenic esophagus (ROI2) and stomach (ROI3). Of note, adjacent BE tissue did not show any signal, but two additional previously unrecognized lesions were identified in a macroscopically normal appearing area of the resected specimen of patient 1 (P1) after Hsp70-TPP-Cy5.5 application (Fig. 5A and B, lower panel). The lesions showed a high and specific TBR, underlining the potential for improved fluorescence-guided biopsy protocols (Fig. 5B, lower panel, right). In addition, a high TBR (≥ 4) was observed in the tumor areas of both patients (Fig. 5C, right). Finally, histopathological analysis of all esophageal specimens confirmed the diagnosis of EAC with expression of Hsp70 by IHC (Fig. 5D and E). Importantly, non-dysplastic BE regions of the excised esophageal samples did not show any signal, providing evidence of high specificity of Hsp70-TPP-Cy5.5 for the detection of dysplasia and EAC.

The objective of the present study was to evaluate the feasibility of using Hsp70 as a predictive marker of EAC by ex vivo fluorescence molecular endoscopy in mouse models of BE and ex vivo imaging on human tissues. The correlation of Hsp70 expression with EAC progression was first evaluated by immunohistochemistry (IHC) on dysplasia and EAC tissues from 12 patients and IHC Hsp70 data were then compared with our L2-IL1b mouse model of BE at different time points (6, 9, and 12 months; n = 3–6 per group). We then evaluated Hsp70 expression in vitro using patient-derived organotypic BE cultures with different degree of dysplasia (n = 3/4) by flow cytometry and treated them with the Hsp70-specific 14-mer Tumor-Penetrating Peptide (Hsp70-TPP) labeled with carboxyfluorescein/CF. After treatment, we used time-lapse microscopy on organoids culture with different degree of dysplasia (n = 2) to visualize peptide internalization. In addition, we evaluated the growth and proliferation of BE organoids isolated from 3 different patients in duplicates before and after 24 and 48 h of treatment with Hsp70-TPP. Experiments were then performed using our mouse model of BE L2-IL1b and endoscopic evaluation for Hsp70 was performed 24 h after the injection of Hsp70-TPP labeled with Cy5.5 by a custom-made fluorescence molecular endoscopy system. One non-injected mouse (L2-IL1b mouse, female, 12 months old) served as control. The endoscopy was performed using 6-, 9-, and 12-month-old male and female mice, with a sample size of 2–3 to 6 per age group.

3D organoids were isolated from esophageal biopsies of patients with BE, dysplasia, or HGD/EAC using a slightly modified protocol [50]⁠. Briefly, the tissue was cut in 1–2-mm pieces and incubated with Accutase (ThermoFisher) for 20 min at RT with gentle shaking. After the incubation, EDTA-PBS without calcium and magnesium (DPBS, 2 mM) was added followed by PBS plus 10% FBS and 1% penicillin–streptomycin (PS) and the sample was centrifuged at 500 rcf for 7 min at 4 °C. After centrifugation, the pellet was resuspended in 50 μl of Matrigel (Corning no. 356231) and seeded in a 24-well plate with L-WRN-derived media plus growth factor supplements. Human organoids derived from BE without dysplasia (Hu108, Hu115, and Hu109), BE with dysplasia (Hu60), or EAC (Hu75) were then maintained at 37 °C in a humidified atmosphere of 5% CO₂ and 95% air and media was changed every 2–3 days. Experiments were then performed before 6 passages.

Hsp70-TPP peptide was found to specifically bind membrane-bound as well as cytosolic residing Hsp70 by mimicking the proteins’ oligomerization domain, mediating its in vivo tumor specificity in many different solid tumors [27]. Previous toxicity studies on murine tumor models revealed no toxic side effects of Hsp70-TPP peptide [26, 27]. In the present study, we investigated for the first time the cytotoxicity of Hsp70-TPP-CF on human tissues. BE organoids (n = 3; passage 3–6) were incubated for 48 h with 200 µg/ml of the peptide added to the culture media. For this experiment, the organoid viability and proliferation were assessed before and after an incubation time of 24 and 48 h by measuring size and number of viable organoids. Experiments were performed in duplicate. To measure the diameter of the organoids, pictures (5–10 images per organoid type per each time point) were taken with an inverted microscope (Zeiss Axiovert 200 M) and the size of each organoid was measured with ImageJ/Fiji software [51], using the mean of 3 diameters. The average size of all three organoid types was then calculated at each time point by using the mean size of 11 to 27 organoids for each organoid type. Data are represented as the mean ± SEM. The number of organoids for each type counted at 24 and 48 h after incubation was normalized by dividing it to the organoid count at time 0. Data are represented as the mean of normalized data from all three organoid types ± SEM.

For the time-lapse microscopy, organoids (passage 3–6) were incubated with cell recovery solution (Corning) for 10 min on ice, in order to dissolve the embedding Matrigel. PBS plus 10% FBS was then added to each well and the whole content was placed in a Falcon tube on ice for 30 min. After the Matrigel was dissolved, the organoids were washed and subsequently incubated with Hsp70-TPP-CF (100–200 μg/ml diluted in PBS). After an incubation of 10 min and washing, the pelleted organoids were resuspended in 50 μl of Matrigel and seeded in a 2-well or 4-well chamber slide (Lab-Tek II Chamber Slide). Time-lapse microscopy was then started immediately and performed over a course of 1–8 h using an epifluorescence inverted microscope (Zeiss AxioObserver Z1). Images were acquired in bright field and FITC channel every 60–120 s with automatic exposure time. Time-lapse videos were then processed and brightness and contrast of the whole images were further adjusted using ImageJ/Fiji software.

To confirm the expression of membrane-associated Hsp70 in malignant cells of the human organoids, flow cytometry was performed on single cell suspensions from human organoids as follows. Untreated, viable organoids (passage 3–6) were washed with PBS plus 10% FBS and subsequently disassembled by incubation in EDTA-trypsin (Sigma-Aldrich, St. Louis, MO, USA), supplemented with 0.5U/ml collagenase (Roche Diagnostics, Mannheim, Germany) for 10 min at 37 °C. After enzyme deactivation with PBS plus 10% FBS, the single cell suspension from organoids was centrifuged again and the pellet was incubated with isotype-matched (IgG1) control antibody (BD Biosciences) or FITC-conjugated cmHsp70.1 antibody for 30 min at 4 °C protected from the light as described previously [43]⁠. Only viable cells (propidium-iodine negative) were then analyzed using a FACSCalibur flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA). Flow cytometry data were then quantified by using the mean fluorescence intensity and/or the percentage of Hsp70-positive cells, obtained after subtracting the contribution from cells bound to the IgG1 control antibody.

All animal experiments were performed with protocols approved by the District Government of Bavaria and in accordance with the German Animal Welfare and Ethical Guidelines of the Klinikum rechts der Isar, TUM (Munich, Germany). All procedures were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Genetic mouse models over-expressing IL1b (L2-IL1b) were generated as previously reported and backcrossed to C57BL/6 J mice [11, 33]⁠. Mice were fed with water and standard chow diet (Ssniff, V1124-000) ad libitum and were genotyped between 6 and 8 weeks of age. For the endoscopy experiments, L2-IL1b mice of 3 different age groups (6, 9, and 12 months; n = 2/3–6 per group) were used.

For histology, L2-IL1B mice of 3 different age groups ( 6, 9, and 12 months) representing disease progression were sacrificed by isoflurane overdose . The esophagus and stomach of L2-IL1B mice were resected, formalin-fixed (10%), paraffin-embedded, and sectioned. Slides were then processed for H&E staining. Histological scoring was performed by an experienced mouse pathologist using a blinded scoring system, using previously established criteria for the influx of immune cells per high-power field, metaplasia, and dysplasia in mice [11, 33]. Inflammation represents a score of all immune cells within a defined area of tissue around the squamocolumnar junction (SCJ), predominantly made up of neutrophil myeloid cells. Metaplasia was assessed through identification of mucous-producing cells per gland, and the number of glands with mucous-producing cells in the BE area. Dysplasia was assessed by evaluating cellular atypia in the presence of low- and high-grade dysplasia within each gland. For final scoring, the histological scoring of each mouse was averaged with the macroscopic scoring of the lesions, performed by evaluating tumor extension into the cardia and tumor size as reported by Munch et al. [33].

For human tissues, BE, LGD, and HGD/EAC tissues from esophagectomy or endoscopic resection specimens from 12 patients were fixed in formalin, embedded in paraffin blocks, and stained for H&E and Hsp70 IHC. Images were then acquired with an Aperio Slide Scanner (Leica Biosystems, Wetzlar, Germany).

Quantification of Hsp70 was performed in the mouse samples by counting the Hsp70 + cells (brown) in the entire BE region. In human samples, the expression of Hsp70 was semi quantitatively assessed in the cells of the whole BE, dysplasia, and/or EAC region by an experienced gastroenterologist as the following: 0—no expression, 1—low expression, 2—moderate expression, and 3—high expression.

The high sensitivity fluorescence/color endoscope that was built for this project is based on an imaging system developed by our group [47]. Specifically, the imaging setup was designed to offer video-rate simultaneous color and near-infrared (NIR) fluorescence endoscopy. The system employs a multipurpose flexible endoscope for small animal imaging (i.e., 0.8-mm outer diameter, 6000 pixels, Micrendo-Fiberskop, SCHÖLLY FIBEROPTIC GMBH, Denzlingen, Germany). White light illumination for color imaging is provided by a 250-W halogen lamp (KL-2500 LCD, Schott AG, Mainz, Germany) while the fluorescence excitation is provided by a fiber-coupled continuous wave (CW) laser diode emitting at 670 nm (SLD1332V, Thorlabs, Newton, NJ, USA). The power at the distal end of the endoscope complies with the American National Standards Institute (ANSI) and the European Standards (EN) limits for the maximum permissive exposure in skin (40 mW/cm² measured at distance < 2 mm). Both light sources are coupled into a multimode bifurcated fiber-bundle (Leoni FiberOptics, Neuhaus-Schierschnitz, Germany) connected to the light guide of the fiberscope. The acquired optical signal is divided by a beam splitter (FF685-Di02, Semrock, Rochester, NY, USA) into a visible and a NIR channel. The visible channel is recorded by a high-resolution color charge-coupled device (CCD) camera (pixelfly qe, PCO AG, Kelheim, Germany), through relay lenses (MAP10100100-A, Thorlabs), while a highly sensitive electron-multiplying CCD (EMCCD) is used for the detection of the NIR channel (DV897DCS-BV, Andor Technology, Belfast, Northern Ireland). In order to enable the use of the system with various fluorophores, a dichroic mirror rotating holder has been installed (CDFW5/M, Thorlabs), while sliding holders (CFS1/M, Thorlabs) allow for the interchangeability of the filters so that optimal imaging is achieved under any fluorophore.

For fluorescence endoscopy, male and female L2-IL1b mice (7 males, 4 females) with different grades of dysplasia were used. Mice were i.v. injected with Hsp70-TPP-Cy5.5 (OEM manufactured by Thermo Fisher; dosage: 100 µg, equal to 45 nmol per animal) and sacrificed 24 h post-injection through anesthetic overdose. Immediately after the sacrifice, endoscopy was then performed by advancing the endoscope into the esophagus while air was gradually introduced to allow esophagus distension. Videos were recorded and stored in AVI file format. The TBR from the fluorescence endoscopy images was quantified by averaging the pixel intensity values within two ROIs (tumor and background ROI) and taking their ratio. The selection of the ROIs was performed by defining the fluorescence signal from BE and macroscopically visible dysplastic lesions and the surrounding normal esophageal mucosa was defined as the background.

To verify the in vivo biding capabilities of Hsp70-TPP-Cy5.5 to target malignant lesions and to validate the fluorescence molecular endoscopy, we used an ex vivo fluorescence imaging system to detect Hsp70-TPP-Cy5.5-derived fluorescence signals on exposed or excised specimen. For near-infrared epifluorescence imaging in tissue, specimens were illuminated using a 670-nm diode laser (B&W tek, DE, USA). The emitted fluorescence was guided through a 780/10 bandpass filter and captured with a back-illuminated EM-CCD camera (iXon DU888, Andor), as described previously [27]. Fluorescence images were sequentially acquired at exposure times of 0.2, 0.5, 1, and 2 s. Comparative analysis has been carried out at images taken with identical exposure times, objective apertures, and camera settings. Signal specificity was determined by calculating the ratio of the mean signal intensities of the relevant tissue and the adjacent normal tissue using ImageJ as previously reported [52]. Data are represented as single plotted values and/or mean ± SEM. Imaging procedures of the different types of specimen have been performed as follows.

To verify the applicability of Hsp70-TPP-Cy5.5 to specifically label malignant tissue using topical application, a subcutaneous (s.c.) xenograft human pancreas carcinoma (Colo357) model and the IL-1b overexpressing genetic (L2-IL1b) mouse model were used. Animals with xenograft tumors were euthanized when s.c. tumors reached a volume of 0.2 to 0.3 cm³ and tumors and adjacent tissues were excised. L2-IL1b mice received an endoscopic imaging after sacrificing them and the whole esophagus and stomach were spread flatly on a black coated aluminium foil before imaging. Xenograft s.c. tumors were spread flatly on a black coated aluminum foil and incubated with an aerosol of Hsp70-TPP-Cy5.5 at a concentration of 200 µg/ml in PBS, for 5 min at room temperature. After extensive rinsing of the tissue with PBS, fluorescence images from the exposed areas were taken, as described above. To quantify the penetration depth of Hsp70-TPP-Cy5.5 in malignancies, fresh s.c. tumors were subsequently cut transversally and longitudinally into 1-mm-thick slices directly after exposure to the Hsp70-TPP-Cy5.5 aerosol. Fluorescence imaging was performed, as described above.

For fluorescence imaging on human specimens, resected fresh human esophageal carcinomas and normal tissues were collected from two patients with EAC undergoing curative esophagectomy without prior chemotherapy. After collection, excised tissues were rinsed in PBS before topical spray application of Hsp70-TPP-Cy5.5 at a concentration of 200 µg/ml to the total area of the specimen (~ 420 cm²). After incubation for 5 min at RT, the specimens were extensively rinsed in PBS and imaged as described above. The specific binding of Hsp70-TPP-Cy5.5 was quantified with ImageJ by using the ratio between the mean fluorescence intensity of the tumor area and the mean fluorescence intensity in the normal tissue. After imaging, the tissues were fixed in 10% formalin, embedded in paraffin blocks, and sections were stained for H&E and Hsp70. IHC for Hsp70 was performed as reported in “Immunohistochemistry (IHC)” section.

Quantitative data were analyzed by one-way analysis of variance (ANOVA) followed by post hoc Tukey test, two tailed unpaired t test (quantification of endoscopy) and  nested t test (quantification of ex vivo wide-field imaging) . Statistical analysis of semi-quantitative data was performed by Kruskall-Vallis test and differences in the means were evaluated using Dunn’s multiple comparison test. All statistical analyses were performed with GraphPad Prism Software (San Diego, CA, USA) and data are represented as single plotted values and/or mean ± SEM. P ≤ 0.05 was considered significant.