Hyaluronidase-Expressing Salmonella Effectively Targets Tumor-Associated Hyaluronic Acid in Pancreatic Ductal Adenocarcinoma

Abstract

In pancreatic ductal adenocarcinoma (PDAC), the extracellular matrix (ECM) surrounding cancer cells forms a barrier that often limits the ability of chemotherapeutic drugs and cytotoxic immune subsets to penetrate and eliminate tumors. The dense stromal matrix protecting cancer cells, also known as desmoplasia, results from the overproduction of major ECM components such as collagens and hyaluronic acid (HA). Although candidate drugs targeting ECM components have shown promise in increasing penetration of chemotherapeutic agents, severe adverse effects associated with systemic depletion of ECM in peripheral healthy tissues limits their use at higher, more effective doses. Currently, few strategies exist that preferentially degrade ECM in tumor tissue over healthy tissues. In light of this, we have developed an attenuated, tumor-targeting Salmonella typhimurium (ST) expressing functional bacterial hyaluronidase (bHs-ST), capable of degrading human HA deposited within PDAC tumors. Our data show that bHs-ST (1) targets and colonizes orthotopic human PDAC tumors following systemic administration and (2) is efficiently induced in vivo to deplete tumor-derived HA, which in turn (3) significantly increases diffusion of ST within desmoplastic tumors. BHs-ST represents a promising new tumor ECM-targeting strategy that may be instrumental in minimizing off-tumor toxicity while maximizing drug delivery into highly desmoplastic tumors.

INTRODUCTION

PDAC is predicted to become the second leading cause of cancer-related death in western countries and, thus, is considered a major public health concern (1). Currently, the approved drug combination of gemcitabine and nanoparticle albumin–bound paclitaxel (Abraxane®) and the highly toxic FOLFIRINOX for advanced PDAC have been shown to improve patient survival compared to gemcitabine alone (2,3). However, these toxic regimens may only extend survival for a matter of months at the expense of decreased quality of life and increased potential for complications later in life (4). The continued lack of early detection methods, poor treatment efficacy and resistance contribute to a meagre 5-year survival rate (for all stages) of only 7% (5). It is imperative that new agents and combination treatment strategies are developed to improve overall survival for patients with PDAC.

HA, also known as hyaluronan, is a component of PDAC stroma that is expressed at extremely high levels in the ECM, resulting in a biophysical barrier that significantly increases interstitial fluidic pressure, compresses blood vessels and hinders effective drug delivery (6,7). While PDAC tumors have the greatest incidence of HA overexpression in patients (>95%), other cancer types such as breast and prostate cancer also express high levels (8,9), and metastatic lesions have been found to have similar levels of HA to the primary tumor (10). Thus, agents to degrade tumor-derived HA, and other overexpressed ECM components, to improve drug delivery and efficacy, has been an area of extensive research (11–14). Recently, clinical trials with a pegylated form of the human hyaluronidase PH20 (PEGPH20) reported significant increases in progression-free and overall survival when combined with Abraxane® and gemcitabine compared to gemcitabine alone (15). However, because hyaluronidase is delivered systemically and activity is not restricted to only tumor tissue, significant adverse events have been observed relating to HA depletion in joints and other organs, requiring lower doses or co-administration with additional agents to minimize these stresses (7,16,17). Oncolytic viruses, specifically adenoviruses, encoding human PH20 are among the few agents that can cause tumor-specific ECM degradation (18).

While eukaryotic hyaluronidases have been categorized as “hyaluronidases”, numerous studies have confirmed a broader target range to include degradation of chondroitin, chondroitin sulfates, and collagen (19,20). While degradation of these additional ECM components in the tumor could be beneficial to maximizing drug delivery to tumor cells, this may further contribute to systemic toxicity as these components are also abundant in non-malignant, healthy peripheral tissues. Bacterial hyaluronidases (bHs) have long been studied as a cost effective alternative to bovine and human hyaluronidases, due to greater ease in purification and a single specificity to HA depending on the bHs used (21). Nearly all bHs are expressed by gram-positive bacteria which include the genera Streptococcus, Streptomyces and Clostridium and, like in eukaryotes, act primarily as tissue remodelling or “spreading” factors (22). Although bHs have been purified and shown to have comparable or higher activity than eukaryotic hyaluronidases, similar potential toxicity in humans exists when delivered systemically as bHs, and the bacteria from which they were isolated, are not tumor-specific. Several genera of gram-negative bacteria, however, have been extensively studied for their ability to colonize, replicate in, and regress solid tumors, with attenuated strains of Salmonella typhimurium (ST) showing the most promise (23–25). Many studies have shown that these attenuated ST strains are highly tumor-specific and are easily cleared from non-tumor tissues with ratios from 250:1 to 9000:1 ST found within the tumor versus peripheral organs such as the liver (26). Previous work attempting to express human hyaluronidases on the surface of gram-negative bacteria (i.e. E. coli) resulted in auto display of hyaluronidase with significantly low activity likely due to defective post-translational modifications and folding (27–29). In contrast, extensive work by Pavan et al. describes the successful E.coli expression and purification of bHs from Streptomyces koganeinsis with activity comparable to or higher than bovine and human hyaluronidases and a unique specificity for only HA (21). However, whether bHs was secreted or auto displayed in E.coli was not determined.

In this study, we have developed and characterized various attenuated ST strains expressing bHs from S. koganeinsis (bHs-ST). We show that attenuated ST is capable of auto-displaying functional bHs that can effectively degrade purified and tumor-derived HA. We also confirm that bHs-ST, when delivered systemically, is capable of preferentially colonizing orthotopic human PDAC tumors in mice and can cause remarkable degradation of tumor-derived HA resulting in enhanced diffusion of ST throughout the tumor. This is the first microbial-based, tumor-specific, ECM-degrading strategy that could significantly improve efficacy of therapies for PDAC and other desmoplastic tumor types.

MATERIALS AND METHODS

Animals and cell lines

NSG mice were obtained from breeding colonies housed at the City of Hope (COH) Animal Research Center and, for all studies, handled according to standard IACUC guidelines. The PANC-1 and PC-3 cell lines were obtained from ATCC® (CRL1469™, CRL1435™) in 2017. Cells were frozen in liquid nitrogen at low passage and used within 20 passages of receipt from ATCC. Mycoplasma testing of cell lines was preformed following the protocol from Christian Praetorius (BiteSizeBio) derived from Uphoff and Drexler (30). Thawed cells were tested for mycoplasma routinely prior to use in experimentation in vitro or prior to implantation in mice. PC-3 cells were maintained in RPMI media containing 10% FBS, 2mM L-glutamine and pen/strep. PANC-1 cells were maintained at ≤80% confluency in DMEM containing 10% FBS, 2mM L-glutamine and pen/strep.

ST strains and generation of bHs-ST

YS1646 was obtained from ATCC® (202165™). Other attenuated strains were kind gifts obtained from Roy Curtiss III (χ8429, χ8431, χ8768), B.A.D Stocker (SL7207) and Michael Hensel (MVP728) (31–35). YS1646 was cultured in modified LB media containing MgSO₄ and CaCl₂ in place of NaCl. All other strains were cultured in Miller LB media (Fisher BioReagents). The S. koganeinsis bHs amino acid sequence (UniProt, A0A0U2E2) was used to synthesize an S. typhimurium codon-optimized cDNA (Biomatik) inserted in-frame into a 6xHis-EGFP-pBAD bacterial expression vector (kind gift from Michael Davidson, Addgene #54762) using XhoI/EcoRI sites which removes the EGFP insertion. In-frame insertion of bHs into the pBAD vector adds a 6XHis tag to the N-terminus of the protein and is predicted to generate a membrane-bound bHs upon induction with L-arabinose. This plasmid can be acquired through Addgene, plasmid #134259. χ8768-LUX was generated using the pAKlux2 plasmid (kind gift from Attila Karsi, Addgene #14080). Plasmids were electroporated into ST strains (1mm gap cuvettes, settings: 1.8kV, 186 ohms), spread onto LB plates containing 100 μg/mL ampicillin and incubated overnight at 37°C. Glycerol stocks were generated for pBAD-bHs-positive clones identified by colony PCR and restriction digest of plasmid preparations.

Bacterial growth, viability and analysis of bHs expression

ST clones electroporated with pBAD-bHs were cultured in media with or without 2% (w/v) L-arabinose at 37°C, 225 rpm for time intervals ranging from 3hr - 24hr. Growth kinetics were monitored through absorbance readings at 600 nm (Genesys 30, Thermo Scientific) every 1–2 hrs, up to 24 hrs. 6XHis-tagged bHs expression was detected in bacterial lysates by western blot and localization of bHs was detected by immunofluorescence using a primary monoclonal mouse anti-6XHis antibody (Proteintech). For immunofluorescence, uninduced and induced ST grown for ~3 hours were fixed with 4% paraformaldehyde at room temperature (RT) for 30 minutes, and permeabilized with 0.1% Triton-X 100/PBS pH=7.2 at RT for 30 minutes followed by lysozyme (Sigma, 100 μg/mL final concentration in 5mM EDTA) at RT for 45 minutes. Fixed/permeabilized bacteria were incubated with primary antibody (1:100) for 30 minutes with shaking in a humidified 37°C incubator followed by incubation with FITC-conjugated anti-mouse secondary (1:200, Abcam) and DAPI for 30 minutes with shaking in a humidified 37°C incubator.

Hyaluronan-BSA LB (HBL) plate and turbidimetric assays

HBL plates for evaluating hyaluronidase activity were generated as previously described. Briefly, LB agar plates containing final concentrations of 0.4 mg/mL HA (Sigma, H-1504), 1% bovine serum albumin fraction V (Sigma) and 100 μg/mL ampicillin (Sigma) were used for plating uninduced and induced ST strains (10⁶ colony forming units (CFU)/5 μL drop) at 37°C for 16–24 hrs. Plates were then flooded with 2N glacial acetic acid. Clear zones were observed against a background of opaque precipitated BSA conjugated to the undigested HA. For turbidimetric quantification of HA degradation in culture media over time, the cetyltrimethylammonium bromide (CTAB) turbidimetric method was used (36). In brief, LB media containing 0.4 mg/mL HA and 100 μg/mL ampicillin, with or without 2% L-arabinose, was used to culture bHs-ST strains (2 mL starting volume) over 24 hrs at 37°C, 225 rpm. HA content (absorbance) in culture media (100 μL aliquot) was measured every 2–3 hrs by addition of 2.5% CTAB reagent (25 μL, Sigma) and absorbance read at 600 nm.

Orthotopic and subcutaneous tumor implantation

Previously published methods were used for orthotopic implantation of PANC-1 cells into the pancreas of NSG mice (37). Briefly, while anesthetized and using sterile techniques, a small incision was made in the skin and peritoneal lining and the pancreas externalized. Using a 27 gauge needle, approximately 2 × 10⁶ PANC-1 cells in a volume of 50 μL Matrigel (BD Biosciences) were injected into the body of the pancreas. The pancreas was then reinserted into the peritoneal space and inner and outer incisions were closed using absorbable sutures and staples, respectively. Analgesics were administered pre- and post-surgery. For subcutaneous (s.c.) injections, 2 × 10⁶ PANC-1 cells in a volume of 50 μL Matrigel were injected s.c. into the right flank of NSG mice.

ST administration, induction and therapeutic studies in PANC-1 tumor-bearing NSG mice

NSG mice with palpable orthotopic PANC-1 tumors (>250 mm³) were intravenously injected with 2.5×10⁶ cfu χ8768-LUX or χ8768-bHs. For therapeutic studies, mice with s.c. PANC-1 tumors (>150 mm³) were injected with 2.5×106 cfu χ8768-bHs. Actively growing χ8768-LUX is constitutively bioluminescent and was used to evaluate χ8768 colonization of PANC-1 tumors in vivo using intravital imaging (LagoX, Spectral Imaging). For all experiments, two days after administrating χ8768-bHs, mice were administered 240 mg L-arabinose or PBS intraperitoneally (i.p). For gemcitabine treatment, mice were given 40 mg/kg i.p. two days later and continued treatment 2 times per week thereafter. Tumor diameters were measured using a digital caliper.

Immunohistochemistry/immunofluorescence (IHC/IF) to detect HA, ST and pan-cytokeratin

Prior to incubation with bHs-ST in vitro, sections of PANC-1 tumors were de-paraffinized and rehydrated. Uninduced and induced χ8768-bHs (10⁸ CFU), PBS or bovine hyaluronidase (Sigma) were incubated on tissue sections overnight in a humidified 37°C incubator. Following treatment, specimens were incubated with a biotinylated HA binding protein (HABP, Sigma) at 5 μg/mL final concentration for 2 hours at 37°C. Slides were then washed, incubated with streptavidin-HRP at RT for 1 hr and visualized with a DAB kit (Vectastain). H&E and Masson’s Tricrhome were used for staining. PANC-1 tumor sections, skin and decalcified joints from NSG mice treated intravenously with χ8768-bHs (uninduced and induced). Sections were de-paraffinized and rehydrated and stained overnight with 2.5 μg/mL HABP, 1:100 anti-ST antibody (Santa Cruz, sc-52223), 1:100 anti-pan-cytokeratin (AE1/AE3) or according to H&E and Masson’s Trichrome protocols used by the Pathology Research Services Core (City of Hope). Streptavidin-PE (Vector), anti-mouse-Cy5 (Invitrogen), or anti-mouse-HRP 1:250 were then used to visualize HA, ST or pan-cytokeratin by brightfield or fluorescence microscopy (Zeiss Observer II), in addition to DAPI for visualizing nuclei during fluorescence imaging. Tiling was performed at 5X or 10X, while higher resolution images for ST, HA and DAPI were done at 100X (oil).

Blood Vessel/Duct measurements

Ten to twelve fields at 10x for H&E and trichrome stained slides from PANC-1 tumors treated with χ8768-bHs and induced with L-arabinose or uninduced in vivo were imaged using a Leica DMi8 Microscope. In each field, the largest diameter of vessels/ducts containing red blood cells and/or surrounded by collagen were measured using the Leica LasX software. A Mann-Whitney test was performed on values using the Prism 7.2 software from GraphPad.

RESULTS

Tightly-regulated expression of bHs by attenuated ST strains

In order to circumvent potential toxic effects of constitutive bHs expression on attenuated ST strains, we employed a tightly-regulated inducible expression system. Inducible expression in ST is possible through the use of a construct containing the P_BAD promoter of the araBAD (arabinose) operon and the gene encoding the positive and negative regulator of this promoter, araC (38). An ST codon-optimized bHs sequence, based on the amino acid sequence of the well-characterized S. koganeinsis bHs, was synthesized and cloned into a previously described pBAD vector to generate pBAD-bHs (39). A single plasmid preparation of pBAD-bHs was used for electroporation into various attenuated strains of ST (Supplementary Table 1). Colony polymerase chain reaction (PCR) was performed for each transformed strain (≥8 colonies) to detect for retention of the bHs transgene (Figure 1A). All ampicillin-resistant colonies examined for YS1646 and MVP728 were completely negative for the bHs transgene in the absence of L-arabinose, suggesting loss of the transgene independent of induced protein expression. Of note, both YS1646 and MVP728 are derived from the same parental strain ATCC 14028. Culturing of pBAD-bHs-positive colonies in uninduced and induced (+2% L-arabinose) conditions, followed by coomassie blue (CB) staining and western blot (WB) of pellet lysates, revealed expression of His-tagged bHs at the correct molecular weight (27 kD) as well as tight regulation of protein expression (Figure 1B). No bHs was detected in culture media by CB or WB (Supplementary Figure 1A), suggesting that bHs is not secreted by these ST strains following induction.

Figure 1. Transgene stability and expression of Streptomyces koganeinsis hyaluronidase (bHs) by attenuated ST strains.

(A) Colony polymerase chain reaction (PCR) to detect for the bHs transgene contained within an inducible pBAD vector (pBAD-bHs) transformed into indicated ST strains. Representative colony PCRs shown from ≥8 colonies per transformed strain. A positive PCR control using ST-specific attB primers was performed for each colony. E. coli (BL21) transformed with pBAD-bHs is used as a positive PCR control for bHs and negative control for ST attB. (B) ST strains retaining the bHs transgene were cultured in Luria Broth (LB) containing 0% (uninduced) or 2% (induced) L-arabinose and cultured for 3 hours at 37°C. Lysates from ~5×10⁷ colony forming units (CFUs) were run on a 4–20% polyacrylamide gradient gel and subjected to coomassie blue staining (CB) and western blot analysis against an amino terminal His-tag fused to bHs (α-His). Predicted bHs size ~27 kDa (arrow). L = protein ladder. (C) ST strains encoding His-tagged bHs were cultured in LB media containing 2% L-arabinose and immuno-stained (α-His) to determine localization of bHs (green). Nuclei are stained with DAPI (blue). All data presented are representative of ≥3 experiments.

Using HHMTOP, PSORTb and CellP-Loc subcellular localization prediction tools, bHs is predicted to be anchored to the cytoplasmic membrane at its N-terminus, while the active region (residues 66–247) is localized to the outer membrane/extracellular space (Supplementary Figure 1B) (40–42). To determine the subcellular location of bHs expressed by the various ST strains, we performed immunofluorescent staining to observe the 6XHis-tag fused to the N-terminus of the bHs protein (Figure 1C). His-tagged bHs expressed by induced χ8429-bHs and χ8768-bHs reveals clear bHs localization outside of the bacterial cytoplasm, defined by DAPI staining of genomic DNA. In contrast, bHs expressed by SL7207-bHs, and to a smaller extent in χ8431-bHs, is localized to the cytoplasm, suggesting impaired transport and formation of inclusion bodies in these attenuated strains. Altogether, these data confirm that expression of the ST codon-optimized bHs transgene is tightly regulated using an inducible pBAD system and that the expressed bHs protein can be auto displayed on the bacterial surface of certain attenuated ST.

Growth kinetics and viability of bHs-expressing ST strains

While tight regulation of transgene expression is important for minimizing toxicity during initial growth stages, sufficient growth and viability following induction will also be critical to maximizing bHs activity. Thus, we determined growth kinetics of each of the bHs-ST-expressing strains over 24 hours in non-inducing and inducing (±2% L-arabinose) conditions. SL7207 alone is already known to have dramatically reduced growth kinetics compared to wild type ST, reaching a maximum optical density (O.D.) 2–3 fold lower than other attenuated strains (Figure 2). Under induced conditions, the maximum O.D. for SL7207-bHs is significantly reduced to under 1 (Figure 2A), whereas other attenuated ST strains could reach maximum O.D.s 3-fold higher (Figure 2B–D). Both inherently poor growth kinetics of unmodified SL7207 as well as cytoplasmic accumulation of bHs (Figure 1C), could contribute to the significantly reduced growth kinetics of SL7207-bHs following induction. Interestingly, while χ8431-bHs showed mixed localization of bHs by immunofluorescence, induced growth kinetics were indistinguishable from χ8768-bHs and χ8429-bHs (Figure 2D, Supplementary Figure 1C). These data suggest that χ8768, χ8431 and χ8429 have greater viability upon bHs induction compared to SL7207, which could potentially translate into more extensive HA degradation.

Figure 2. Growth kinetics and viability of bHs-expressing ST strains.

Optical density readings (OD₆₀₀) for uninduced (solid blue circles) and induced (open red circles) ST strains (A) SL7207, (B) χ8429 and (C) χ8768 transformed with the pBAD-bHs construct. Cultures were done in triplicate and error bars represent standard error of the mean. (D) Growth curves of induced bHs-expressing strains are compared. *p<0.05 by ANOVA. (E) Bacterial cells from uninduced (–) and induced (+2% L-arabinose) cultures of SL7207-bHs and χ8768-bHs were stained at indicated time points (4 and 24 hours) with acridine orange (live, green) and ethidium bromide (dead, orange/red) and imaged by fluorescence microscopy at 100X magnification. Scale bar = 10 μm. All data are representative of ≥3 experiments.

To further investigate bacterial viability after induction, we performed live/dead staining using a mixture of acridine orange (AO) and ethidium bromide (EB), respectively, during log phase (4 hr) and stationary phase (24 hr) of uninduced and induced cultures (43). As shown in Figure 2E and Supplementary Figure 1D, the percentage of viable bacterial cells after induction of SL7207-bHs is significantly lower than χ8768- and χ8429-bHs, as indicated by highly EB-positive SL7207-bHs as early as 4 hr and continuing 24 hr after induction. These results further emphasize the deleterious toxic effects of bHs expression on the viability of attenuated strains such as SL7207 but also highlight ST strains capable of auto displaying bHs and remaining viable during and long after initiation of induction.

BHs-ST strains degrade purified HA

To test the functionality of bHs expressed by the various attenuated ST strains, we employed HA agar plate clearing and liquid culture turbidimetric assays (36). For plate clearing assays, HA and BSA are mixed into LB agar plates (HBL plates). Addition of 2N acetic acid to HBL plates containing intact HA will form a white precipitate with BSA, while areas of HA degradation will remain clear. BHs-expressing strains were pre-induced for 3 hours in LB media containing 0 to 4% L-arabinose and then spotted (1×10⁸ CFU/5 μL) onto HBL plates overnight. After flooding HBL plates with 2N acetic acid, zones of clearing were observed for χ8768-, χ8431- and χ8429-bHs, but not SL7207-bHs (Figure 3A). Interestingly, χ8431-bHs, which had exhibited intermediate surface display of bHs (Figure 1C), also demonstrated intermediate HA degradation. These data suggest that ST strains that efficiently display bHs on their surface and exhibit greater viability are far more effective at degrading pure HA.

Figure 3. Functional analysis of bHs-expressing ST strains.

(A) bHs-expressing strains were cultured in LB broth containing indicated percentages of L-arabinose for 3 hours at 37°C. 1×10⁸ CFUs were plated onto LB-hyaluronan-BSA agar plates overnight and then flushed with 2N acetic acid. Plates were imaged on a black background to visualize areas of clearing, indicating hyaluronan breakdown. (B) 1×10⁸ CFUs of ST-bHs strains were added to LB containing 0% or 2% L-arabinose and 0.4 mg/mL HA. The cetyltrimethylammonium bromide turbidimetric method (CTM) was used to determine rate of HA breakdown over 24 hours (OD₆₀₀) for pBAD-bHs-transformed SL7207, (C) χ8768 and (D) χ8729. Error bars = standard error of the mean. All data are representative of ≥3 experiments.

To determine the kinetics of HA degradation by the various bHs-ST strains, we cultured each under uninduced and induced conditions in LB media containing HA (0.4 mg/mL) and measured HA content over a 24 hour period using the cetyltrimethylammonium bromide (CTAB) turbidimetric method (39). At each time point, high molecular weight HA in culture media was precipitated with CTAB and optical density determined at a wavelength of 600 nm (Figure 3B–D). We observed higher overall rates of HA degradation over the 24 hr period by χ8768- and χ8429-bHs (~0.15 O.D. units/hr), an intermediate rate for χ8431-bHs (~0.10 O.D. units/hr) (Supplementary Figure 1E) and no degradation by SL7207-bHs, which recapitulates activity observed for each strain on HBL plates. The highest rate of degradation was observed for χ8768-bHs during the first 12 hours of induction (~0.3 O.D. units/hr), whereas χ8429-bHs and χ8431-bHS showed two times less activity (~0.15 O.D. units/hr). Overall, these data indicate that the χ8768-bHS strain is most effective in degrading purified HA within hours of induction.

χ8768-bHs effectively degrades tumor-derived HA

Based on its viability following induction and ability to efficiently degrade purified HA, we selected χ8768-bHs to further determine if bHs-expressing ST could degrade tumor-derived HA. We utilized the human pancreatic cancer line PANC-1, which we confirmed expresses high levels of HA when grown orthotopically in NSG (immune-deficient) mice. We first performed in vitro HA degradation experiments whereby fixed PANC-1 tumor sections were incubated with pre-induced χ8768-bHs. Overnight incubation of PANC-1 tumor sections with pre-induced χ8768-bHs resulted in dramatic degradation of HA compared to sections incubated with PBS or uninduced χ8768-bHs (Figure 4), with no observable loss or change in tumor cell content. We performed similar degradation experiments using the PC-3 prostate cancer cell line, which secretes high levels of HA while in culture, and also observed considerable depletion of HA by induced χ8768-bHs with no change in tumor cell density (Supplementary Figure 2). These results strongly suggest that χ8768 expressing bHs degrades HA directly and not through a mechanism involving depletion of HA-expressing tumor cells.

Figure 4. Induced χ8768-bHs effectively depletes tumor-derived hyaluronan.

χ8768-bHs was grown for 3 hours in LB media containing 0% (uninduced) or 2% (induced) L-arabinose. 1×10⁸ CFUs were then co-incubated with de-paraffinized serial sections of PANC-1 tumor tissue overnight. HA was detected using biotinylated HA-binding protein (HABP) followed by incubation with streptavidin-HRP and ImmPACT DAB substrate. Serial sections incubated with PBS serve as an HA-positive control and specificity of HABP was confirmed through overnight incubation with 10 U/mL bovine hyaluronidase (Bov). Hs). Scale bar = 75 μm. All images are representative of ≥3 experiments.

Systemically-delivered χ8768-bHs degrades HA in xenograft PDAC tumors

We next determined the ability of χ8768-bHs to colonize and deplete HA in orthotopically implanted PANC-1 tumors when delivered intravenously into mice. We first verified that the χ8768 strain was capable of colonizing PANC-1 tumors utilizing a constitutive bacterial reporter gene construct encoding the bioluminescent LUX operon (44). When recombinant χ8768 encoding LUX (χ8768-LUX) was injected intravenously into NSG mice bearing orthotopically implanted PANC-1 tumors (>250 mm³) we observed bioluminescence localized to the area of the tumor, which was detected on day 3 and was undetectable on day 5 (Supplementary Fig. 3A). To further verify tumor-specific colonization by χ8768-LUX, we isolated tumor, spleen and liver (noted areas of ST accumulation) (26) following i.v. injection and measured bioluminescence for each tissue type. Indeed, χ8768-LUX was highly concentrated in tumor tissue while completely absent in both spleen and liver (Supplementary Fig. 3B). These results suggest that χ8768 is capable of colonizing orthotopic PANC-1 tumors after systemic administration and that tumor-specific colonization is achieved by day 2, which would represent an ideal time point for induction of bHs activity.

Thus, NSG mice with orthotopic PANC-1 tumors (>250 mm³) were intravenously (i.v.) administered 2.5×10⁶ CFUs of χ8768-bHs and then induced 2 days later by a single intraperitoneal (i.p.) injection of 240 mg of L-arabinose per mouse. PANC-1 tumors were excised 24 after induction and sectioned and stained for both HA and ST. As shown in Figure 5A, tumors from mice receiving only χ8768-bHs (uninduced) were characterized by limited (punctate) ST colonization and high HA deposition remaining in areas of ST colonization. In contrast, tumors from mice administered χ8768-bHs and L-arabinose (induced) showed a dramatic degradation of HA, particularly within areas colonized by χ8768-bHs (Figure 5B), with no observable loss of tumor cell density based on DAPI (overlay) and pan-cytokeratin staining (Supplementary Figure 3C). Remarkably, greater diffusion of χ8768-bHs (p<0.05, t-test) was also observed under induced conditions (Figure 5C, D). We observed diffusion of χ8768-bHs from a number of duct-like structures into highly dense (DAPI-positive) tumor tissue of induced mice, which were predominantly devoid of HA (Supplementary Figure 3D). In uninduced mice, these structures were less prevalent but when observed, ST were found within the duct but had not diffused into the surrounding tissue which displayed high HA staining. Similar to observations made in vitro, these in vivo results support a mechanism whereby HA depletion occurs predominantly through direct degradation by induced X8768-bHs and not through reductions in tumor cell density. Through H&E staining, we identified vessel-like structures which are easily detected by the presence of red blood cells (Supplementary Figure 3E). Tumors of induced mice displayed a predominance of larger, open vessel-like structures compared to a majority of smaller, closed structures in uninduced mice (Supplementary Figure 3F, p<0.0001, t-test), reminiscent of the confined, punctate immunofluorescent staining of ST colonization in these tumors (Figure 5C). Altogether, these data strongly suggest that χ8768-bHs is capable of effectively degrading tumor-derived HA in vivo to facilitate delivery of agents as large as ST, in which a single bacterium can measure 2.5 um in width, 5 μm in length, and reach a molecular weight in the hundreds of gigadaltons (45–49).

Figure 5. Systemic delivery of χ8768-bHs effectively degrades HA within xenograft PANC-1 tumors.

Uninduced χ8768-bHs (2.5×10⁶ CFU) was injected intravenously (i.v.) into NSG mice bearing orthotopic PANC-1 tumors (>250 mm³). After 48 hours, mice were then administered (A) PBS (uninduced) or (B) 240 mg L-arabinose (induced) by intraperitoneal (i.p.) injection. Tumors were isolated 16 hours later, sectioned and stained for ST (green) and HA (red) for subsequent immunofluorescence imaging at 10X and 100X magnification. Blue, nuclear staining using DAPI. 10X scale bars = 200 μm, 100X scale bars = 10 μm. (C) Tile-scanning was performed on entire tumor sections at 10X magnification. Sections were stained for ST (green) and DAPI (blue) for subsequent immunofluorescence imaging. Scale bars = 2 mm. Representative tumors are shown for uninduced and induced groups. (D) Percent area of tumor colonized by χ8768-HAse under uninduced and induced conditions based on immunofluorescence. Percentage calculated using: (Area occupied by χ8768-Hase (green)/Total tumor area (DAPI)) × 100%. Areas (μm²) were determined using Image-Pro Plus (Media Cybernetics) analysis software. *p<0.05, t-test. All data are representative of ≥3 experiments.

χ8768-bHs potentiates the anti-tumor effects of gemcitabine treatment

We next evaluated the therapeutic benefits of HA-degradation, by induced χ8768-bHs, when used in combination with gemcitabine chemotherapy. We utilized a subcutaneous (s.c.) PANC-1 xenograft model in order to measure tumor growth during treatment. We first confirmed the ability of χ8768-bHs to deplete HA in s.c. PANC-1 tumors and also determined the duration of HA depletion following induction. As shown in Figure 6A, we observed a greater presence of χ8768-bHs and specific depletion of HA in PANC-1 tumors 3 and 7 days after induction that is not present under uninduced conditions or in other HA-rich tissues, such as the skin and joints (Supplementary Figure 4A), under induced conditions. Only minimal punctate χ8768-bHs tumor colonization was observed on day 3 under uninduced conditions, which is completely absent by day 7 and on day 14 (Figure 6A, Supplementary Figure 4B). However, under induced conditions, χ8768-bHs and HA depletion were still present in PANC-1 tumors on day 7 and waned by day 14. Similar to previous observations, there was no loss of tumor cell density (based on pan-cytokeratin and DAPI staining) or collagen degradation (trichrome) in HA-depleted areas (Supplementary Figure 4B, C). These data suggest that systemically-delivered χ8768-bHs is capable of depleting HA specifically in PANC-1 tumors for an extended period following induction.

Figure 6. χ8768-bHs potentiates the anti-tumor effects of gemcitabine treatment in PANC-1 tumor xenografts.

Uninduced χ8768-bHs (2.5×10⁶ CFU) was injected i.v. into NSG mice bearing subcutaneous (s.c.) PANC-1 tumors (>150 mm³). After 48 hours, mice were then administered PBS (uninduced, U) or 240 mg L-arabinose (induced, I) by i.p. injection. (A) S.c. tumors and skin (n=4) were isolated 3, 7 and 11 days post-i.p. injection (dpi), sectioned and stained for HA (red) and ST (green) for subsequent immunofluorescence (IF) imaging at 5X magnification. Trichrome staining of serial sections for same tissue sample also shown to left of IF images. Representative images shown. Arrows indicate area of ST/HA overlap. Scale bars = 50 um. (B) After 2 dpi, groups of tumor-bearing mice (n=6) were administered either gemcitabine (40 mg/kg) or diluent control (0.9% saline) by i.p. route, followed by additional administrations twice per week. PBS only group did not receive pre-treatment with χ8768-bHs. Tumors were measured weekly using a digital caliper. **p<0.01, ***p<0.001, ANOVA with Tukey’s post hoc test. (C) Mouse body weights were measured on indicated days following gemcitabine or control treatment and are presented as a percentage of initial body weight. n.s., not significant.

We next determined if HA depletion by induced χ8768-bHs could improve the efficacy of gemcitabine in the PANC-1 model. Mice with established s.c. PANC-1 tumors (n=6, >150 mm³) were administered χ8768-bHs and were induced or left uninduced 48 hours later by i.p. injection of L-arabinose or PBS, respectively. For mice receiving gemcitabine, 40 mg/kg were administered 2 days post-i.p. injection (dpi) and twice per week thereafter. As shown in Figure 6B, tumor growth rates for mice having received induced χ8768-bHs in combination with gemcitabine were significantly reduced (weeks 9–11, p<0.01, 0.001, ANOVA) compared to groups receiving χ8768-bHs alone or uninduced χ8768-bHs in combination with gemcitabine. Additionally, no significant change from initial body weight was observed in mice receiving χ8768-bHs alone compared to PBS (Figure 6C). Only negligible weight loss (<10%) was observed in groups receiving gemcitabine, with no discernable added weight loss caused by χ8768-bHs pre-treatment under induced conditions. Overall, these data indicate that χ8768-bHs treatment can deplete HA in tumors for an extended period of time to improve efficacy of gemcitabine with little to no added toxicity.

DISCUSSION

Hyaluronidase administration has been shown to enhance the efficacy of gemcitabine and nAb-paclitaxel in PDAC tumor models and has had some clinical benefit in PDAC patients (50,51). However, the high risk of adverse effects associated with systemically delivered hyaluronidase still presents major concerns due to ECM degradation in healthy tissues. To reduce risk, lower doses of hyaluronidase must be given, which may not necessarily maximize the therapeutic efficacy of chemotherapy. BHs-ST is the first example of a microbial-based, ECM-degrading agent that focuses its enzymatic activity strictly to tumor tissue, potentially maximizing HA degradation and therapeutic drug delivery. We established that bHs is anchored to the surface of ST, reducing the likelihood of systemic off-tumor effects resulting from circulating bHs. Previous studies have also shown that bHs expressed by S. koganeinsis has a unique specificity to HA, further reducing the risk of degrading other major ECM components in healthy tissue. Furthermore, we and others have shown replication of attenuated ST in tumor tissue, suggesting that an initial input of bHs-ST could be amplified in tumors for a few days, before induction of bHs expression, to cause maximal HA degradation and minimal off-target effects. The ability of attenuated ST to extensively and preferentially colonize tumors also increases the window for therapeutic intervention and delivery. The types and sizes of therapeutic agents that show enhanced delivery and whether multiple administrations are possible will be ongoing studies to determine the overall utility of bHs-ST. For greater feasibility, an autonomous version of bHs-ST will likely need to be developed to avoid additional induction steps. Furthermore, transient tumor colonization by ST, as observed in Figure 6A and Supplementary Figure 3A, would be a limitation of our current strategy, requiring multiple doses and/or exquisite timing of induction for optimal HA depletion in each individual. The use of tumor-specific promoters, such as hypoxia-inducible promoters, to drive bHs expression is one potential solution to overcome this (52).

In some cases, ST-based cancer therapy has been shown to regress tumors in pre-clinical models, and this regression is heavily dependent on the ability of ST to colonize tumors (53,54). The first attenuated ST to enter clinical trials, VNP20009, was administered to patients with metastatic melanoma and head and neck cancer (25). Virtually no tumor regression was observed in any patients receiving VNP20009, and only a small number of patients had tumors colonized by ST. Our ability to express bHs in ST could significantly increase tumor colonization, as shown for induced bHs-ST, and could potentially increase efficacy of ST-based therapies. In this study, increased distribution of induced χ8768-bHs within PANC-1 tumors was not shown to be therapeutic alone, which is likely a limitiation of using an immune-compromised xenograft model. Indeed, previous work utilizing PEGPH20 in immune competent mice significantly improved colonization and anti-tumor efficacy of our ST-based therapeutic targeting indoleamine 2,3-dioxygenase (shIDO-ST) (23). Thus, χ8768-bHs alone could prove to be more tumor-specific and therapeutic, compared to hyaluronidase, in an immunocompetent setting (17). Overall, these observations suggest that delivery of relatively large particles such as bacteria could be enhanced through the use of ECM-degrading enzymes. Therefore, bHs-ST pre-treatment could easily be combined with virotherapy, cell- or antibody-based therapies to improve their delivery.

In addition to increasing interstitial fluidic pressure, which prevents blood flow and diffusion of therapeutic agents, HA also plays a large part in tumor progression as a signaling molecule that activates intracellular pathways and promotes motility and metastasis (55,56). Indeed, clinical data reveal a better prognosis for patients with lower deposition of HA (10,57). Thus, targeting HA alone may be sufficient to cause primary tumor regression as seen in PC3 prostate tumors (58) or prevent metastasis. However, pre-clinical and clinical trials have shown that targeting hyaluronan alone does little to decrease tumor growth and metastasis (7,59). A few key studies in PDAC have even shown that ECM depletion may be detrimental to overall survival as the ECM may act as a physical barrier to tumor cell dissemination, and the increased vascularization can increase tumor cell survival (60). Thus, the beneficial effect of hyaluronidase treatment for tumors will be in its ability to improve delivery of therapeutic drugs or antibodies (59), and for PDAC, the aforementioned data may indicate that ECM depletion must go hand-in-hand with chemotherapy or other anti-cancer treatment to avoid possible unfavorable effects.

The number of therapies being developed to treat cancer, which includes nanoparticles, antibody-based therapies, chemotherapeutic combinations, viruses and bacteria, and T-cell therapies, continues to rise and, therefore, strategies to improve penetration of these agents is also required. Currently, only a small number of ECM-targeting agents can claim to improve intratumoral drug delivery, with the caveat of systemic ECM degradation and associated adverse events (7,14). We have developed an ECM-degrading agent that is more tumor-specific and, thus, can increase the number of potential therapeutic combinations to maximize efficacy while minimizing toxicity. In this work, we show that bHs-ST is capable of targeting tumor-derived HA in PDAC tumors and increases penetration of large particles. While bHs-ST may only be effective in patients with HA-high tumors, such as those selected for treatment with PEGPH20 (50), the ST platform described in this work could be used to develop other ECM-targeting strategies with more universal application and benefit to PDAC patients.