Hemangiosarcoma and its Cancer Stem Cell Sub-Population are Effectively Killed by a Toxin Targeted through Epidermal Growth Factor and Urokinase Receptors

Abstract

Targeted toxins have the potential to overcome intrinsic or acquired resistance of cancer cells to conventional cytotoxic agents. Here, we hypothesized that EGFuPA-toxin, a bispecific ligand - targeted toxin (BLT) consisting of a deimmunized Pseudomonas exotoxin (PE) conjugated to epidermal growth factor and urokinase, would efficiently target and kill cells derived from canine hemangiosarcoma (HSA), a highly chemotherapy resistant tumor, as well as cultured hemangiospheres, used as a surrogate for cancer stem cells (CSC). EGFuPA-toxin showed cytotoxicity in four HSA cell lines (Emma, Frog, DD-1, and SB) at a concentration of ≤100 nM, and the cytotoxicity was dependent on specific ligand-receptor interactions. Monospecific targeted toxins also killed these chemoresistant cells; in this case, a “threshold” level of EGFR expression appeared to be required to make cells sensitive to the monospecific EGF-toxin, but not to the monospecific uPA-toxin. The IC₅₀ of CSCs was higher by approximately two orders of magnitude compared to non-CSCs, but these cells were still sensitive to EGFuPA-toxin at nanomolar (i.e., pharmacologically relevant) concentrations, and when targeted by EGFuPA-toxin, resulted in death of the entire cell population. Taken together, our results support the use of these toxins to treat chemoresistant tumors such as sarcomas, including those that conform to the cancer stem cell model. Our results also support the use of companion animals with cancer for further translational development of these cytotoxic molecules.

INTRODUCTION

Although current therapy protocols may result in prolonged survival, tumor recurrence and disease progression constitute major causes of mortality in many cancer patients. Bispecific ligand-targeted toxins (BLT) have the potential to overcome the major causes of treatment failure, such as acquired or intrinsic resistance to conventional therapies like chemotherapy and radiation therapy. The ligands are designed to bind specific receptors that are uniquely or highly expressed by tumor cells as compared to normal cells. This improves tumor specificity by directing the ligands towards the tumor, thus reducing toxic side effects associated with cancer chemotherapy and radiation therapy. Additionally, binding specific receptors enhances the potency of ligand-targeted toxins^1–3. In fact, as few as 1000 molecules of Pseudomonas exotoxin (PE) delivered to the cytosol results in tumor cell death, and this can be achieved using picomolar concentrations of BLTs^{1, 4}. Moreover, the potency and specificity of BLTs has been previously demonstrated through their in vivo activity, with acceptable safety profiles, against human breast cancer, brain cancer, and blood-derived tumors^{1, 3, 5–7}.

In this study, we tested a BLT called EGFuPA-toxin, designed to simultaneously target the epidermal growth factor receptor (EGFR), which is upregulated in a variety of cancers, and the urokinase receptor (uPAR), which is expressed on sarcomas, endothelial cells and tumor vasculature^8–11. EGF and the amino acid terminal fragment (ATF) of uPA were conjugated to a truncated Pseudomonas exotoxin A (PE38), shown previously to have potent anticancer activity via inhibition of protein synthesis¹². To enhance its potency, PE38 was modified by adding a Lys-Asp-Glu-Leu (KDEL) C-terminus signal to prevent secretion from the luminal endoplasmic reticulum. Finally, the toxin was deimmunized via mutation of seven B-cell epitope-encoding sequences, identified by Onda and Pastan¹³, to permit multiple in vivo treatments without generating an anti-toxin immune response.

EGFuPA-toxin makes a promising potential chemotherapeutic agent because in addition to targeting the EGFR, it also targets uPAR-expressing sarcomas, as well as endothelial cells lining the tumor vasculature. Canine hemangiosarcoma (HSA) is a tumor derived from blood vessel forming cells, and thus has been proposed as a model to study tumor angiogenesis^{14, 15}. This tumor has also been shown to express both EGFR^{16, 17} and uPAR¹⁶ (genome-wide gene expression profiles are available as GEO SuperSeries GSE15086). Canine HSAs are highly resistant to conventional therapy¹⁸, an observation that extends to HSA-derived cell lines in vitro.^a Thus, canine HSA provides a model of a highly resistant sarcoma, as well as a model for studying the effects of agents targeting cells that comprise the tumor vasculature. Furthermore, HSA appears to conform to the cancer stem cell (CSC) model, showing a hierarchy with a subpopulation of cells (CSCs) that show greater resistance to chemotherapeutic drugs compared to non-CSCs^a. The CSC model provides one potential explanation for the observed drug resistance seen in this disease.

Here, we examined the cytotoxicity of EGFuPA-toxin against HSA cell lines, as well as against cultures of cells enriched for HSA CSCs, to assess the feasibility of targeting these highly chemoresistant tumors and cells comprising the tumor vasculature. EGFuPA-toxin resulted in cytotoxicity against all of the HSA cell lines examined, including the enriched cancer stem cells at low nanomolar to picomolar concentrations, which should have pharmacologic relevance^1–3. Given the frequency of hemangiosarcoma in domestic dogs and infrequency of the homologous disease in people, further clinical development of this agent could be done in companion animals as prelude to trials in human patients. Therefore, this work supports additional clinical development of EGFuPA-toxin, as well as the potential to use companion animal tumors for further translational development of this compound.

MATERIALS & METHODS

Construction of Targeted Toxins, Protein Expression, Refolding, and Purification

DNA shuffling and cloning techniques were used to synthesize and assemble the genes encoding the single chain bispecific EGFuPA-toxin (EGFATFKDEL 7mut), as described previously². The gene consisted of an NcoI restriction site, an ATG initiation codon, human EGF, and the 135 ATF portion of uPA. Five amino acids of the PE38 molecule (RELK) were replaced with Lys-Asp-Glu-Leu (KDEL) at the C- terminus to prevent secretion from the luminal endoplasmic reticulum. As detailed elsewhere, the final toxin was deimmunized by the modification of 8 amino acids³. Genes for monospecific targeted toxins EGF-toxin (EGFKDEL) and uPA- toxin (ATFKDEL) were created using the same techniques. A negative control bispecific toxin BIC3 (CD3CD3KEL), targeted towards the T-cell surface marker CD3, was made by replacing the DT390 portion of the CD3 molecule with PE38KDEL¹⁹. Proteins were expressed and purified from inclusion bodies by using a Novagen pET (Merck KGaA, Darmstadt, Germany) expression system²⁰ and then dissolved in phosphate buffered solution (PBS).

Cell Culture

Canine HSA cell lines Emma, DD-1, Frog, and SB, and the feline mammary adenocarcinoma cell line K12 were maintained as adherent cultures at 37°C in 5% CO₂ atmosphere as described previously^{15, 16, 21}. For hemangiosphere formation, SB and DD-1 cells were grown under serum-free conditions that favor non-adherent cell growth and sphere formation^{22, 23}, and are henceforth referred to as SBS and DD-1S. The corresponding cell lines grown as a monolayer are referred to as SBM and DD-1. Hemangiospheres were dissociated enzymatically once a week into single-cell suspensions using Accutase (Sigma-Aldrich, St. Louis, MO); cells were allowed to re-form spheres under the same conditions for maintenance or re-plated as single cells for experiments. Human Jurkat cells (T-cell leukemia) and canine Nike cells (histiocytic sarcoma cells) were maintained as described ^{24, 25}.

Flow Cytometry

Cells were harvested using Accutase and resuspended in staining buffer (PBS with 2% FBS). Cells were incubated with 11.8 μg of canine IgG (Jackson ImmunoResearch Laboratories, Inc.) to block nonspecific binding and then incubated with 2 μg of anti-MHC class I antibody (VMRD, Pullman, WA), 12.5 μl of EGF Alexa Flour 488 (2 μM, Molecular Probes), or 2.5 μl of Native Human LMW Urokinase-FITC (133 μM, Cellsciences, Canton, MA). Cells were analyzed using a FACSCalibur cytometer (BD Biosciences, San Jose, CA) and data were analyzed with FlowJo (v9.5.2) software (Treestar, Ashland, OR). Viable cells were determined by exclusion of 7-Aminoactinomycin D (7AAD).

Cytotoxicity Assays

Cells were incubated overnight prior to addition of each toxin (EGFuPA-toxin, EGF-toxin, uPA-toxin, BIC3) at the concentration indicated in the results section. Conditions were performed in triplicate using 5,000 cells per well in 100 μl of culture medium. For antibody blocking assays, cells were incubated in 96 well plates for 24 hours preceding the addition of 0.1 nM EGFuPA-toxin and varying concentrations of anti-human EGF antibody (R&D Systems, Minneapolis, MN) or varying concentrations of anti-uPA antibody (American Diagnostica Inc., Stamford, CT), as indicated. The mouse leukocyte specific antibody anti-Ly5.2 was included at the same concentrations as a negative control²⁶. For high throughput screening (HTS), 1,500 cells (40 μl) were incubated in duplicate 384 well plates for 72 hr. The Sigma Aldrich library of pharmacologically active compounds (LOPAC) was used to assess cytotoxic responses using a robotic system (University of Minnesota Institute for Therapeutic Discovery and Development Core). In total, 1,280 compounds were screened for cytotoxicity against Frog cells and 640 were screened against Emma and SB cells. Each compound was added to a final concentration of 10 μM; viability was measured at 72 hr using the Alamar Blue assay as a readout. More than 100 of these have known activity as anti-metabolites, DNA damaging agents, or protein synthesis inhibitors.. EGFuPA-toxin (1.6 nM) was used in these assays as a control. Competitive binding assays were performed by incubating cells at 4°C with increasing concentrations of either EGF (Invitrogen, Camarillo, CA) or native human urokinase alone, or added together at the same concentration (i.e., twice as much ligand as was added when using either one alone). After 30 min, 0.5 nM EGFuPA-toxin was added. Viability was assessed after 72 hr of culture using CellTiter96 AQueous kit (Promega, Madison, WI) as described by the manufacturer.

Statistics

Descriptive statistics were done for each experimental condition (mean, SD, range). For viability assays, fractional viability was calculated as the percentage of live cells compared to control at the same time points. Where needed, IC₅₀ values were calculated using linear regression and the line of best fit.

RESULTS

Canine HSA and feline mammary adenocarcinoma cells express EGFR and uPAR

Since EGFuPA-toxin requires binding to EGFR and uPAR, we confirmed that HSA cells expressed the cognate receptor. Expression of uPAR and EGFR mRNA by HSA cells was documented previously using gene expression arrays and RT-PCR, respectively^{16, 17}. However, these studies did not evaluate surface protein expression; thus, we used flow cytometry to quantify the expression of both receptors in live cultured cells. Figure 1a shows expression of EGFR by four HSA cell lines (Emma, Frog, DD-1, SBM) and by K12 feline mammary adenocarcinoma cells that overexpress EGFR²¹ as controls. Our results show that fluorescein-labeled EGF reproducibly bound to all four HSA cell lines. Although expression levels were relatively low, the populations were largely unimodal and the histogram legends show shifts in mean fluorescence intensity (reflecting an overall change in expression for EGFR across the population) without defining arbitrary gates for “positive” and “negative” populations. Each of the four HSA cell lines also showed “shoulders” consisting of ~0.5–10% of the cells with higher expression. Consistent with our previous data²¹, K12 cells expressed high levels of surface EGFR. In contrast, Figure 1b shows binding of fluorescein-labeled uPA was bimodal and the histograms show gates that define the subsets in each of the HSA cell lines and in K12 cells (~7–25%) that expressed surface uPAR.

Figure 1. HSA and K12 cells express EGFR and uPAR.

(a) Expression of surface EGFR was assessed on HSA cells (Emma, Frog, DD-1, SBM) and control feline mammary carcinoma cells (K12) using flow cytometry with AlexaFluor488-labeled EGF. EGFR staining is shown by solid black curves and controls are displayed as grey filled curves. MFI = mean fluorescent intensity. (b) Expression of surface uPAR was evaluated using flow cytometry with FITC-labeled uPA. uPAR staining is shown by the solid black curves and controls are displayed as grey filled curves. Gates represent percent positive cells.

Canine HSA cells are highly chemoresistant to traditional chemotherapeutics but are sensitive to a BLT targeted by EGF and uPA

Chemoresistance is a problem in many solid tumors. In the case of canine HSA, it is unclear if the perceived chemoresistance is due to the conservative therapy regimens or an intrinsic property of the HSA cells themselves. We used an HTS platform to identify compounds with cytotoxic activity against the HSA cell lines Emma, Frog, and SB. Each cell line was sensitive to <10% of the compounds tested, as defined by reproducible inhibition of growth >50% compared to controls, and overlap of sensitivity profiles among the cell lines was even more restricted. Table 1 shows that among the cytotoxic chemotherapy drugs from the LOPAC, only idarubicin showed activity against all three cell lines. Interestingly, cyclosporin A and Bay 11–705, which inhibit the NFAT and NFκB transcriptional pathways respectively, each showed profound growth inhibition of canine HSA cells, suggesting additional areas for therapeutic application. All three cell lines were resistant to carboplatin and cisplatin (<35% inhibition), and while Emma and SB cells were sensitive to 5-azacytidine and 5-fluorouracil, Frog cells were not sensitive to either of these compounds. Other agents among the 1,280 LOPAC that showed only modest cytotoxic activity (<50% inhibition) against Frog cells in the HTS assay included the antiproliferative agents paclitaxel, vinblastine, and mitoxantrone, as well as the targeted protein kinase inhibitors SU4312 and SU 5416. In conventional MTS assays, SB and Frog cells also showed moderate resistance to doxorubicin (IC₅₀s ~100 nM and 1 μM, respectively)^a. Together, these data confirmed that HSA cells have relatively high intrinsic resistance to cytotoxic drugs, and highlighted the need to develop approaches to overcome such resistance.

Table 1.

Sensitivity to Select Antiproliferative Agents by Canine HSA Cell Lines¹

	Emma	SBM	Frog
Compound	Sensitivity (% cell death at 10μM concentration)
5-azacytidine	Yes (85%)	Yes (85%)	No
5-Fluorouracil	Yes (60%)	Yes (81%)	No
BAY 11–705	Yes (100%)	Yes (100%)	Yes (100%)
Carboplatin	No (<50%)	No (<50%)	No (<50%)
Chlorambucil	No (<50%)	Yes (60%)	No
Cisplatinum	No	No	No
Cyclosporin	Yes (80%)	Yes (95%)	Yes (100%)
Cyclophosphamide	No	No	No
Idarubicin	Yes (100%)	Yes (100%)	Yes (100%)

¹Viability was determined using an Alamar Blue assay in cells treated with compounds from the Sigma Aldrich LOPAC.

Next, we examined in vitro cytotoxicity of the EGFuPA-toxin against Emma, Frog, DD-1, and SBM cell lines. As shown in Figure 2, EGFuPA-toxin showed substantial dose-dependent cytotoxicity against all the HSA cell lines with IC₅₀s ranging from 0.01–1.0 nM. The EGFuPA-toxin showed comparable cytotoxicity in the HTS platform (65 ± 2.2% inhibition of Frog cells, 87 ± 1% inhibition of Emma cells, and 88 ± 1% inhibition of SB cells when used at 1.6 nM). As expected, EGFuPA-toxin was effective against K12 cells, which overexpress EGFR. Cytotoxicity was a specific event: EGFuPA showed no cytotoxicity against canine histiocytic sarcoma cells and the negative control, BIC3 toxin directed against the CD3 T-cell receptor complex, had no effect on hemangiosarcoma cell lines at a concentration of 100 nM whereas it predictably killed human Jurkat cells (a T-cell leukemia) with an approximate IC₅₀ of 1 nM (data not shown).

Figure 2. Canine HSA tumor cell lines are sensitive to EGFuPA-toxin.

(a) Canine HSA cell lines (Frog, DD-1, Emma, SBM) and control feline mammary carcinoma cells (K12) were cultured for 72 hr with increasing concentrations of EGFuPA-toxin. Viability was measured in triplicate samples using the Cell Titer 96® Aqueous Non-Radioactive Cell Proliferation Assay (MTS), and data were normalized to 100% in the absence of BLT. Data are representative of two or more independent experiments. Error bars represent normalized standard deviation. The ranges of IC₅₀s are as follows: Frog (0.007–0.01nM, standard deviation 0.002), DD-1 (0.07– 0.3nM, standard deviation 0.11), Emma (.02–0.2nM, standard deviation 0.07), SBM (0.003– 0.01nM, standard deviation 0.03), and K12 (.0037.03nM, standard deviation 0.01).

EGFuPA-toxin requires binding to cognate EGFR and uPAR to induce cytotoxicity

To verify that the BLT was acting specifically through cognate receptors, we examined the effect of blocking receptor interactions using anti-ligand antibodies or excess unlabeled ligands. We selected a concentration of EGFuPA-toxin (0.1 nM) that killed between ~45% and 99% of the cells. Figure 3a shows that in all HSA cell lines and in control cell line, K12, blocking the EGF portion of the EGFuPA-toxin with anti-EGF antibody or the uPA portion with the anti-uPA antibody prevented cell death in response to EGFuPA-toxin in a dose-dependent fashion. The inhibition of cell killing was a specific event, as addition of an irrelevant antibody (anti-Ly5.2) did not inhibit killing of either cell line by the EGFuPA-toxin.

Figure 3. Blocking EGFR and uPAR attenuates BLT cytotoxicity.

(a) HSA cell lines and K12 cells were incubated with 0.1 nM EGFuPA-toxin in the presence of varying concentrations of blocking antibodies against EGF, uPA, or Ly5.2 (irrelevant isotype control) as shown. Viability was measured in duplicate samples using an MTS assay. A value of 100% represents maximal viability in the absence of BLT. One representative experiment of two or more done is shown for each cell line. (b) Emma and (c) K12 cells were incubated with unlabeled ligands at the indicated concentrations at 4°C for 30 min. In the condition labeled EGF + uPA, the ligands were added at the same concentration used in each individual condition. EGFuPA-toxin was added at 0.5 nM and the experiment was carried out for 48 hr. Viability was measured in duplicate samples using the MTS assay. A value of 100% represents maximal viability in the absence of BLT. One representative experiment of three done is shown for each cell line.

We then determined if blocking the interaction of the toxin and the cognate receptors with excess ligand would have a similar effect. Figure 3b shows that, in Emma cells, blocking receptor-toxin interactions with both uPA and EGF ligands (present together, each at 1,000 times molar excess of the toxin concentration) had the greatest effect to restore cell growth, blocking 60% of EGFuPA-toxin cytotoxicity (triangles). At the same concentration (1000-fold excess), uPA alone prevented almost 40% of cell death (squares), whereas EGF alone was minimally effective (circles). Figure 3c demonstrates in K12 cells, the combination of uPA and EGF ligands (each present at 1,000-fold excess) abrogated ~70% of cell death (triangles), but unlike in the HSA cells, the addition of uPA alone, even at 1000-fold excess (squares), did not prevent cell death of K12 cells. As we saw in the HSA cells, the addition of EGF alone, even at 1000-fold excess (circles), also did not prevent K12 cell death.

Previous studies have shown that BLTs are more toxic than their monospecific counterparts^{1, 3, 5, 6, 20}; however, our data from the antibody blocking experiments suggested that, at least in K12 cells, the effect of the BLT might have been due primarily to the EGF component. Therefore, we examined how effective the BLT EGFuPA-toxin was at inducing cytotoxicity of HSA (Emma) cells and K12 cells as compared to monospecific EGF and uPA toxins. Figure 4 shows that in Emma cells, uPA-toxin was as effective as EGFuPA-toxin at inducing cell death, showing ≥50% cytotoxicity at a concentration of 0.1 nM. This occurred despite the very low expression of uPARs. Somewhat surprisingly, EGF-toxin was at least 100 times less active, requiring a concentration of 10 nM to achieve >50% cytotoxicity. In K12 cells, the monospecific EGF-toxin was as effective as the BLT EGFuPA-toxin, showing almost complete cytotoxicity (>98%) at a concentration of 0.1 nM. In K12 cells, the uPA-toxin required about 2 orders of magnitude greater concentration to achieve >99% cytotoxicity.

Figure 4. HSA and mammary carcinoma cells show differential sensitivity to monospecific versus bispecific BLTs.

Emma and K12 cells were cultured for 72 hr with or without varying concentrations of EGF-toxin, uPA-toxin, or EGFuPA-toxin as indicated. Viability was measured in duplicate samples using the MTS assay, and data were normalized to 100% in the absence of BLT. One representative experiment of two or more done is shown for each cell line.

Canine hemangiospheres express EGFR and uPAR and are sensitive to EGFuPA-toxin

The use of targeted toxins may help circumvent chemoresistance in tumors that show a hierarchical organization. In such tumors, CSCs show greater resistance to conventional cytotoxic drugs than non-CSCs^a, and they may be uniquely responsible for therapy failures and relapse²⁷. We examined the effect of EGFuPA-toxin on non-adherent tumor spheres (hemangiospheres), which contain an enriched fraction of drug resistant cells and the putative CSC fraction, to assess the capacity of EGFuPA-toxin to target and kill CSCs. As shown in Figure 5a, hemangiospheres derived from the SB cell line (SBS), showed a unimodal population distribution of EGFR expression resembling that seen in the monolayer cells, but EGFR levels were ~5-fold higher (compare to SBM cells in Figure 1). Interestingly, enrichment for hemangiospheres resulted in a unimodal population of uPAR expression, in contrast to the bimodal expression seen in the monolayer cell lines. Figure 5b shows that the EGFuPA-toxin effectively killed SBS cells, albeit the IC₅₀ was approximately 2 orders of magnitude higher than that of SBM cells. DD1-S was also sensitive to EGFuPA-toxin, albeit these enriched DD-1 sphere cells also were more resistant than their monolayer counterpart (Figure 5c). It is nonetheless noteworthy that hemangiospheres were sensitive to EGFuPA-toxin at concentrations that we expect to be pharmacologically achievable based on in vivo data^1–3.

Figure 5. CSCs from HSA express higher levels of EGFR and uPAR and are sensitive to EGFuPA_toxin-mediated cytotoxicity.

(a) Expression of EGFR and uPAR were measured as in Figure 2 on SB non-adherent hemangiospheres (SBS). (b) A two-fold dose response cytotoxicity assay was carried out for 72 hours on SB cells grown as a monolayer (SBM) and SB non-adherent hemangiospheres (SBS), as well as, (c) DD-1 cells grown as a monolayer (DD-1) and as non-adherent hemangiospheres (DD-1S). Viability was measured in triplicate samples using the MTS assay. A value of 100% represents maximal viability in the absence of BLT; error bars represent intra-experimental standard deviations. One representative experiment of three done is shown for each cell line.

DISCUSSION

Here, we showed for the first time that EGFuPA-toxin induces cytotoxicity of highly chemoresistant sarcoma cells. Our data demonstrate that canine HSA cell lines, which exemplify this class of tumors, express low levels of EGFR and uPAR proteins on the cell surface, and that EGFuPA-toxin effectively killed four independent HSA cell lines, as well as hemangiospheres enriched for CSCs. Cytotoxicity using the EGFuPA-toxin was specific, as blocking the interactions of the EGF and uPA ligands decreased the effectiveness of the BLT to kill HSA cells, and the BLT caused significant cell death at picomolar to low nanomolar concentrations, which have pharmacological relevance^1–3.

Although sarcomas are rare in humans, they can be extremely aggressive and some are highly refractory to conventional therapies, creating a significant unmet medical need for new treatment options^{28, 29}. In contrast to humans, where sarcomas make up less than 2% of diagnosed cancers, these tumors are commonly diagnosed in companion animals³⁰, providing an abundant source of samples with high value for comparative studies. Given the paucity of viable human samples, canine tumors can be leveraged as a resource to study important questions that would be challenging to address in humans. In particular, canine HSA is molecularly similar to idiopathic angiosarcoma in humans³¹, and it represents a prototypical, intrinsically chemoresistant tumor for which there are limited chemotherapeutic treatment options³². HSAs also show hierarchical organization with the CSC subpopulation acting as a major factor contributing to chemoresistance^a,³³.

Our data confirm previous results showing reproducible expression of EGFR by HSAs¹⁷. EGFR expression is not generally associated with endothelial cells, so it is unclear if this represents retention of a primitive lineage determinant or if it is a common trait of phenotypic infidelity associated with this tumor. The relatively low, but detectable expression of uPAR by these cells was perhaps more predictable based on gene expression profiling. Expression of surface uPAR was confirmed by flow cytometry, which revealed that expression of this receptor is restricted to a subset of the population and shows slight variation, but not substantive differences across cell lines.

Intriguingly, we found that receptor density was not necessarily correlated with sensitivity to EGFuPA-toxin. For example, expression of these receptors did not specifically correlate with sensitivity between HSA cells from conventional monolayer cultures and HSA CSCs enriched in non-adherent, 3-dimensional spheroid cultures. In monolayer cultures, cytotoxicity from uPA-toxin was as effective as EGFuPA-toxin in HSA cells, however, the enriched CSC population showed a higher level of uPAR expression compared to the monolayer cells, yet was more resistant to EGFuPA-toxin. Additionally, the EGF-toxin showed less cytotoxic activity despite the apparently greater density of EGF binding sites on the cell surface. Although EGF-toxin did not efficiently kill HSA cells, antibody blocking of the EGF ligand alone abrogated cell death of HSA cells induced by EGFuPA-toxin. This could be explained by steric hindrance or because the interaction between EGF on the BLT and EGFR on the cells was necessary to stabilize binding of EGFuPA-toxin to uPAR. The notion that cytotoxicity of EGFuPA-toxin in HSA cells was principally mediated by the uPA-uPAR interaction is supported by the observation that excess uPA, but not excess EGF alone, reduced the number of HSA cells killed by the BLT.

Concomitantly, the results showing that the combination of both ligand competitors further reduced killing of HSA cells by EGFuPA-toxin suggest that EGFR indeed stabilized and augmented binding and internalization of the toxin in these cells.

We interpret these data to suggest that a “threshold” level of EGFR expression is necessary to render cells sensitive to the effects of EGF-targeted toxins. That is, high-density expression of EGFRs, like that seen in epithelial tumors with EGFR amplification, is sufficient to allow targeting with EGF alone. In contrast, low-density expression of EGFRs, like that seen in tumors with no EGFR amplification, might require a second ligand to enhance or mediate targeting^{1, 20}. Thus, in the case of sarcomas where EGFR expression is unlikely to reach a sufficiently high threshold for EGF targeting, the main activity of the EGF component of the EGFuPA-toxin may be to stabilize uPA-uPAR interactions. This has important clinical implications, as it indicates the addition of EGF to the BLT will more specifically target sarcomas that express even low levels of EGFRs. This also may increase the relative concentration of EGFuPA-toxin in the tumor microenvironment, potentially enhancing cytotoxic effects against tumor neovasculature² while at the same time lowering the probability of adverse side effects from targeting normal cells expressing physiologic uPAR, as seen during instances such as wound healing and vascular repair^{34, 35}.

Interestingly, uPA expression in the monolayer cell lines was bimodal (only some cells in the population expressed uPAR), whereas upon enrichment for hemangiospheres, uPAR expression became unimodal. Yet, treatment of monolayer cells with monospecific uPA-toxin resulted in >50% cytotoxicity at 0.1 nM and 100% cytotoxicity at 1 nM. The effects of EGFuPA-toxin required both uPAR and EGFR expression, as the toxin did not affect cell lines lacking these molecules. Therefore, we believe it is unlikely that HSA cells lacking surface uPAR are dying through off-target effects. It is possible, albeit unlikely, that uPAR expression levels in the negative population are too low to be detected by flow cytometry. Therefore, we favor the explanation that the uPAR-expressing cells are responsible for maintaining the overall population, either through self-renewal and expansion or by providing survival and growth signals through cell-cell interactions or by production of soluble factors. These two possibilities are not mutually exclusive and they are consistent with those of CSCs, suggesting that the uPAR expressing cells might include some or all of the CSCs³⁶.

The paradoxical resistance of uPAR⁺ SBS cells suggests that mechanisms of resistance in CSCs are operative even under conditions of exposure to lethal toxins like PE38, but unlike conventional cytotoxic drugs, the resistance to PE38 can be overcome at therapeutically achievable doses^1–3. Moreover, CSCs generally represent a small subset of the tumor population, so the effective concentrations that kill virtually all the cells in monolayer cultures are expected to be more relevant for in vivo situations. Clinically, this endows BLTs with a unique effect against tumors by attacking a multitude of critical components ranging from differentiated cells to CSCs and tumor vasculature, thus expanding the range of tumors that can be treated using this approach.

In summary, our data show that EGFuPA-toxin can help to overcome several problems commonly seen in therapeutic development: potency, off-target effects, and chemoresistance. Additionally, we have identified a subset of the HSA population that when killed by EGFuPA-toxin results in death of the entire population, consistent with targeting of CSCs. Our findings support the development of EGFuPA-toxin as a therapeutic agent for aggressive sarcomas that fail conventional therapy, and the use of companion animals, in which these spontaneous tumors occur with high frequency, to accelerate this process.

Manuscript Category

3.1.8 Cancer Therapy. Reports on new advances in cancer therapy in humans are welcome, especially the results of well-designed randomized trials involving novel therapy strategies or those implicating molecular response indicators to classic therapeutics. If the authors are describing the results of a randomized controlled trial, we recommend use of the style guidelines in describing the study population (see JAMA 1996;276:637–639). If the authors are describing the results of observational studies of therapy, the standards applicable in observational studies in epidemiology should be followed (see above). IJC does not publish case reports.

Impact Statements

• -Sarcomas are aggressive, highly chemoresistant tumors with few treatment options. These data show for the first time that a bispecific ligand-targeted toxin composed of a Pseudomonas exotoxin conjugated to epidermal growth factor and urokinase (EGFuPA-toxin) induces cytotoxicity of highly chemoresistant sarcoma cells, including a subpopulation of cancer stem cells.
• -Sarcomas are infrequent in humans and thus pose a research challenge. However, this challenge can be tackled using domestic dogs, a species that develops sarcomas spontaneously and with high incidence. The similarities between human and canine sarcomas make dogs a valuable resource for therapeutic development. Our findings support the use of companion animal tumors for translational development of EGFuPA-toxin, as a means to accelerate further clinical development of this agent in humans.

Glossary

Abbreviations

(ATF)

Amino Terminal Fragment

(BLT)

Bispecific ligand-targeted toxin

(CSC)

Cancer Stem Cell

(MTS)

Cell Titer 96® Aqueous Non-Radioactive Cell Proliferation Assay

(EGF)

Epidermal Growth Factor

(EGFR)

Epidermal growth factor receptor

(HSA)

Hemangiosarcoma

(HTS)

High-throughput screening

(KDEL)

Lys-Asp-Glu-Leu

(PE)

Pseudomonas Exotoxin

(PE38)

Pseudomonas Exotoxin A

(SBM)

SB cells grown as a monolayer

(SBS)

SB non-adherent hemangiospheres

(uPA)

Urokinase

(uPAR)

Urokinase receptor

(LOPAC)

Library of pharmacologically active compounds