A novel Carcinoembryonic Antigen (CEA)-Targeted Trimeric Immunotoxin shows significantly enhanced Antitumor Activity in Human Colorectal Cancer Xenografts

Immunotoxins are chimeric molecules, which combine antibody specificity to recognize and bind with high-affinity tumor-associated antigens (TAA) with the potency of the enzymatic activity of a toxin, in order to induce the death of target cells. Current immunotoxins present some limitations for cancer therapy, driving the need to develop new prototypes with optimized properties. Herein we describe the production, purification and characterization of two new immunotoxins based on the gene fusion of the anti-carcinoembryonic antigen (CEA) single-chain variable fragment (scFv) antibody MFE23 to α-sarcin, a potent fungal ribotoxin. One construct corresponds to a conventional monomeric single-chain immunotoxin design (IMTXCEAαS), while the other one takes advantage of the trimerbody technology and exhibits a novel trimeric format (IMTXTRICEAαS) with enhanced properties compared with their monomeric counterparts, including size, functional affinity and biodistribution, which endow them with an improved tumor targeting capacity. Our results show the highly specific cytotoxic activity of both immunotoxins in vitro, which was enhanced in the trimeric format compared to the monomeric version. Moreover, the trimeric immunotoxin also exhibited superior antitumor activity in vivo in mice bearing human colorectal cancer xenografts. Therefore, trimeric immunotoxins represent a further step in the development of next-generation therapeutic immunotoxins.

The right side shows the schematic representation of the native proteins displaying their different domains arrangement. In both cases (cDNAs and proteins), structural and functional motifs are highlighted with colours: α-factor secretion signal peptide (S, black); scFvCEA (V L , dark blue, and V H , light blue), L18 linkers ( 18 , dark gray); TIE XVIII , human non-colagenous trimerization domain (TIE, light gray); α-sarcin (red) and histidine-tag (T, yellow). Given its trimeric structure, native IMTXTRICEAαS should display a theoretical size of 159 kDa, meanwhile for monomeric IMTXCEAαS the expected mass value should be 45 kDa. Western blot analysis using rabbit anti-α-sarcin serum (left) or commercial anti-histidine tag antibody (right). α-Sarcin designates 0.1 µg of the fungal natural protein used as a control. MW corresponds to prestained Bio-Rad Precision Plus protein molecular weight standards. Images correspond to full-length gels and blots acquired and analyzed using the Gel Doc XR Imaging System and Quantity One 1-D analysis software (BioRad) or ChemiDoc-It (UVP) and VisionWorks LS, respectively. Different exposures of the blots are presented in Supplementary Fig. S1.
The molecular size in solution of native IMTXCEAαS and IMTXTRICEAαS was determined by FPLC size exclusion chromatographic analysis using a Superdex200 10/300 GL column. As can be observed in Fig. 3B, the elution profile of IMTXCEAαS showed a mass of 45 kDa which corresponds to its expected monomeric globular structure, while the elution profile for IMTXTRICEAαS showed a single symmetric elution peak corresponding to the expected theoretical mass of 159 kDa due to its native trimeric arrangement. This value was also confirmed by laser scattering analysis which results suggested a globular structure with a diameter of 9-10 nm (Fig. 3C), in full accordance with the previously predicted theoretical size of the trimeric format in solution 8 . functional in vitro characterization. The first step in the functional characterization of both immunotoxins was to evaluate the ability of their targeting domain (MFE23 scFv) to bind CEA, either expressed on the cells surface or as a protein immobilized on plastic. According to flow cytometry results, both immunotoxin constructs were able to recognize CEA-positive SW1222 cells but not the CEA-negative HeLa cells (Fig. 4A). The trimeric immunotoxin recognize and bind CEA-positive SW1222 cells more efficiently than the monomeric design (Fig. 4B). In accordance with the flow cytometry analysis, ELISA binding titration curves revealed how the trimeric immunotoxin showed higher binding to immobilized CEA than the monomeric version (Fig. 4C).
Next, we analyzed the highly specific ribonucleolytic activity of the α-sarcin moiety. IMTXCEAαS and IMTXTRICEAαS were able to release the characteristic α-fragment, produced by the specific cleavage of a single phosphodiester bond at the rRNA sarcin-ricin loop (SRL), as efficiently as α-sarcin alone (Fig. 4D). Thereby, these results proved that the ribonucleolytic α-sarcin activity was preserved in both immunotoxins.
In addition, both constructs exhibited high structural stability and maintained their functional integrity in conditions mimicking a physiological context. After incubation of the purified proteins in FBS for 96 hours at 37 °C, western blot analysis showed that both immunotoxins kept their full molecular integrity (Fig. 5A) for at least 48 hours. Furthermore, their ribonucleolytic activity (Fig. 5B) and CEA binding ability, as assessed by ELISA (Fig. 5C) were also preserved during that period of time when compared to freshly prepared control samples. It  Rabbit reticulocytes assays were made in order to test the ribonucleolytic activity of α-sarcin within both constructs. The gels show the release of α-fragment, highlighted by a black arrow, produced by the specific SRL cleavage. In the example shown, 2 and 12 pmol were assayed for both immunotoxins and fungal wild-type α-sarcin. C, negative control where the protein sample was replaced by buffer. Images correspond to full-length gels and blots acquired and analyzed using the Gel Doc XR Imaging System and Quantity One 1-D analysis software (BioRad). Gels presented in the figure were cropped from from different gels or from different parts of the same gels. Original full-length gels are presented in Supplementary Fig. S2 www.nature.com/scientificreports www.nature.com/scientificreports/ is important to note that even after 96 hours of incubation the binding activity of both constructs was still around 90% of the original value.
Once checked that both targeting domains and toxic payload not only kept their functional activities within the purified immunotoxins but also were stable for reasonably long periods of time in physiologically-relevant conditions, inhibition of protein synthesis was studied. After 72 hours of incubating CEA-positive SW1222 and CEA-negative HeLa cells with the immunotoxins, 3 H-Leu incorporation was calculated as a measurement of protein biosynthesis (Fig. 6). As expected inhibition of protein synthesis was only observed within antigen-positive cells. These experiments allowed the calculation of an IC 50 value for both immunotoxins as the molar concentration of protein needed to produce a 50% of protein synthesis inhibition. As shown in Fig. 6, IMTXTRICEAαS was significantly more effective in inhibiting protein synthesis in CEA-positive SW1222 (IC 50 = 6 nM) than monomeric IMTXCEAαS (IC 50 = 60 nM).
In vivo antitumoral effect. Finally, the antitumor activity of two different doses (25 or 50 µg) of IMTXCEAαS and IMTXTRICEAαS was assayed in nude mice bearing CEA-positive human colorectal cancer xenografts (Fig. 7A). Both immunotoxins strongly inhibited tumor growth in a dose-dependent manner as compared to the control group by the end of treatment (day 14). Remarkably, we observed in the IMTXCEAαS groups a rapid tumor growth rebound after the last injection, whereas mice receiving IMTXTRICEAαS showed long-lasting antitumor effect. In fact, at day 25 mean tumor volumes in the IMTXCEAαS groups were around three times larger than in the IMTXTRICEAαS ones ( Fig. 7A,C). Consequently, mice treated with IMTXTRICEAαS showed a higher survival rate (i.e time to reach 2000 mm 3 ; P = 0.0001), as shown in the www.nature.com/scientificreports www.nature.com/scientificreports/ Kaplan-Meier curves (Fig. 7B). Importantly, no signs of toxicity were observed at the doses used, considering body weight and external appearance of the mice (Fig. 7D).

Discussion
Chimeric proteins combining antibody specificity for a user-defined TAA expressed on the cell surface and toxin-derived cytotoxicity (immunotoxins) are important tools for novel cancer therapy strategies. However, next-generation designs are necessary in order to overcome the limitations that currently restrict their therapeutic use, because of suboptimal tumor targeting, tumor penetration and biodistribution, or systemic side effects of non-targeted toxins 11,30 . Recently, the use of α-sarcin as an suitable toxic moiety in recombinant immunotoxins against colorectal cancer has been reported. This immunotoxin showed potent and specific antitumor effectiveness, due to a decrease in cancer cell proliferation and tumor angiogenesis and increased in apoptotic activity, while avoiding unspecific and cytotoxicity against normal cells 25,46 . At the same time, toxin accumulation within targeted cells was considerably increased by the high-affinity binding of the scFv antibody domain 16,25 .
Interestingly, novel multivalent antibody formats appear to increase immunotoxin tumor targeting capacity compared to the monovalent version. Indeed, trimeric immunotoxins are multivalent constructs showing increased functional affinity (avidity). Therefore, the use of these improved antibody formats for the design of next-generation immunotoxins seems to be a promising alternative. Hypothetically, this approach would further increase the toxic payload within the targeted cells and, consequently, it would reduce the amount of immunotoxin necessary to obtain identical antitumor effect along to the risk of undesired non-specific side effects.
As a proof of concept, we have generated, purified and characterized two different α-sarcin-based constructs. The main goal driving this approach was to assess the impact of the trimerbody format in an immunotoxin context, in terms of antigen binding and cytotoxicity. Both immunotoxins, were correctly produced and purified to homogeneity from P. pastoris extracellular media, after methanol induction. Both proteins showed the expected size, according to their monomeric or trimeric format, and CD spectra consistent with their globular β-sheet content. They were also able to specifically bind to CEA on the cell surface and to internalize into the target cells. Furthermore, they kept α-sarcin highly specific ribonucleolytic activity, needed for ribosome inactivation, and www.nature.com/scientificreports www.nature.com/scientificreports/ were stable and functional in physiological-like conditions. Therefore, it seemed safe to conclude that the immunotoxins, both the monomeric and trimeric versions, were properly folded, according to their structural and functional properties.
Their specific cytotoxic activity was evaluated in vitro against SW1222 cells, used as a model of CEA-positive colorectal cancer cell line, and against HeLa cells as off-target cell line. Although both immunotoxins inhibited protein biosynthesis specifically in CEA-positive cells, IMTXTRICEAαS promoted more efficient protein synthesis inhibition than the monomeric version. Thereby, the application of the trimerbody technology platform to α-sarcin immunotoxins improved its cytotoxic effect, at least in vitro. The observed reduction of the IC 50 value for the trimeric immunotoxin would be attributed to its multivalence, increased avidity and higher total amount of toxin load inside the target cells.
The effect observed with both anti-CEA immunotoxins in vitro was also confirmed in vivo using nude mice bearing subcutaneous SW1222 xenografts. Both CEA-targeted immunotoxins caused a significant inhibition of tumor growth. In accordance with the in vitro protein biosynthesis inhibition assays, IMTXTRICEAαS also promoted higher inhibition of tumor growth. Furthermore, this effect was maintained for longer time, with a higher survival rate, once the treatment had finished. Mice treated with IMTXTRICEAαS showed a very slow tumor growth rate till at least day 30, when mice were sacrificed. The improved effect of IMTXTRICEAαS in vivo could also be attributed to a more favorable pharmacokinetics, in terms of half-life and tumor retention. In fact, it has just been reported a faster clearance of the monomeric MFE23 scFv compared to the corresponding trimerbody, which showed higher tumor accumulation in CEA-positive tumor-bearing mice 56 . Thus and as might be expected from its improved in vivo tumor targeting properties, the trimerbody-based immunotoxin showed to be much better therapeutic drug in terms of its antitumor effectiveness. Taking into account that multi-targeted therapies are currently considered as cutting-edge anti-tumoral strategies 58 , we could design trimeric immunotoxins with binding domains directed against different TAAs, increasing tumor specificity and avoiding tumor escape due to antigen-loss, to enhance the effectiveness of these molecules 59 .
Different therapeutic strategies and clinical trials have been developed using MFE-23, including the design of diabodies 60 , immunocytokines, BITEs (bispecific T cell engagers) 61,62 , antibody-directed prodrug therapy strategies 63 and an anti-CEA chimeric antigen receptor (CAR). In this sense, T cells transduced to express this CAR, efficiently killed CEA+ cells in vitro, and inhibited the growth of CRC tumors in vivo, but unfortunately the trial was prematurely closed due to lack of prolonged CAR T cell persistence and acute respiratory toxicity 64 . Given that no MFE23-based therapeutic strategy is being pursued currently for the treatment of CEA+ solid tumors, we considered the existence of a window of opportunity for the design and characterization of MFE23-based immunotoxins.
In summary, the results herein presented not only confirm the potential of CEA-targeted ribotoxin-based immunotoxins, but also represent a further step in the development of next-generation biopharmaceuticals. To our knowledge, we have generated the first immunotoxin with trimeric format, and demonstrated its enhanced cytotoxic activity in vitro and therapeutic effect in vivo compared to the classical monomeric version. The modular design of the construct makes easy the substitution of MFE23 by other scFv directed against different TAA, so we can envision an array of ribotoxin-based immunotoxins for the treatment of a variety of cancers.

Methods plasmid construction.
Plasmids encoding α-sarcin 25 , scFvMFE23 and the TIE XVIII domain 8,27 , were previously obtained. The relevant cDNA sequences were amplified by PCR, including the restriction sites needed for cloning. The IMTXCEAαS or IMTXTRICEAαS cDNAs were cloned in pPICZαA (Invitrogen) for their extracellular production in the methylotrophic yeast P. pastoris BG11. The linker between the two domains of the monomeric immunotoxin was the tripeptide Gly-Gly-Arg. Regarding the trimeric immunotoxin, the sequence L18-TIE XVIII -L18 was inserted between the targeting and toxic domains. Finally, a six His-tag was added at the C-terminus of both immunotoxins for protein detection and purification (Fig. 1). The resulting plasmids were propagated in Escherichia coli DH5αF' and sequenced at the DNA sequence Genomics Unit of the Universidad Complutense.
protein production. Electrocompetent P. pastoris BG11 cells were electroporated with 10 µg of linearized plasmid after digestion with Pme I, as previously described 25,38,45 , using a Bio-Rad Gene pulser apparatus. The yeast containing the desired constructions were selected after plating the electroporation mixture in YPDS plates with increasing amounts of zeocin (100-750 µg/ml). Multiple independent clones were tested in order to find the most productive colonies. For these screenings, cells were grown for 24 hours at 30 °C in BMGY in 24-well plates, being harvested afterwards by 10 min and 1600 g centrifugation. They were then suspended in BMMY for protein production induction at 25 °C and 200 rpm shaking. Protein production was periodically monitored up to 96 hours, adding methanol every 24 hours to a final concentration of 0.5% (v/v) to maintain a steady induction. Protein secretion to the extracellular media over time was analyzed by 0.1% (w/v) sodium dodecyl sulfate (SDS)-15% (w/v) polyacrylamide gel electrophoresis (PAGE) and western blot. A rabbit anti-α-sarcin serum was used for specific detection of the ribotoxin moiety. Given the location of the poly-His tag at the C-terminal end, the integrity of the purified protein was also checked by Western blot using a commercially available anti-histidine tag antibody. Before performing large-scale production, the protein synthesis levels of selected colonies were checked. Those ones with the highest production level were selected for each construction and stored at −80 °C in glycerol media for later use. For large-scale production, growing was performed for 24 hours at 30 °C and 200 rpm using five two-litter baffled Erlenmeyer flasks, each containing 400 ml of BMGY. Induction was carried out in the same Erlenmeyer flasks containing 200 ml of BMMY. Temperature was reduced to 25 °C, (maintaining the same shaking speed). After 24 hours of induction, the media was completely renewed by centrifugation and the cultured induced in the same conditions for another 24 hours. The extracellular medium was collected by centrifugation and dialyzed against 50 mM sodium phosphate buffer, containing 0.1 M NaCl, pH 7.5.
Protein purification. Both immunotoxins were purified from the dialyzed extracellular medium using a Ni 2+ -NTA agarose column (HisTrap TM FF Columns, GE Healthcare). The medium was applied on the column at 1 ml/min using a peristaltic pump. The column was first washed with 50 mM sodium phosphate buffer, 0.1 M NaCl, pH 7.5, followed with the same buffer containing 20 mM imidazole. Finally, the immunotoxins were eluted rising imidazole to 250 mM. Fractions containing the purified protein were pooled and exhaustively dialyzed against the 50 mM sodium phosphate buffer, 0.1 M NaCl, pH 7.5.
Biophysical characterization. Immunotoxins structural characterization was performed as previously described 19,25,38,45 . Absorbance measurements were carried out on an Uvikon 930 spectrophotometer (Kontron Instruments). Far-UV circular dichroism (CD) spectra were obtained from immunotoxin samples at 0.15 mg/ ml in 50 mM sodium phosphate buffer, 0.1 M NaCl, pH 7.5, using a Jasco 715 spectropolarimeter and a scanning speed of 50 nm/min. Cells of 0.1-cm optical path were employed. Four spectra were averaged to obtain the final data.
To check the globular size of both immunotoxins in solution, FPLC was performed using a Superdex 200 column (GE Healthcare Life Sciences) in an AKTA purifier apparatus (GE Healthcare Life Sciences). With the same purpose, laser scattering measurements with the native trimeric immunotoxin in solution were also made at the Spectroscopy and Correlation Facility of the Universidad Complutense. Briefly, two different concentrations of IMTXTRICEAαS, 0.15 and 0.3 mg/ml, were filtered through 0.22 µm filters and analyzed at 25 °C for 60 seconds. The resulting size distribution curves were recorded for size determination, using the average value of ten spectra. Data were analysed using Zetasizer Ver. 7.02 software.

Scientific RepoRtS |
(2019) 9:11680 | https://doi.org/10.1038/s41598-019-48285-z www.nature.com/scientificreports www.nature.com/scientificreports/ Ribonucleolytic activity assays. The highly specific ribonucleolytic activity of α-sarcin was tested against ribosomes from a rabbit cell-free reticulocyte lysate as previously described 25,45 . The release and detection of a characteristic 400 nt rRNA fragment, known as α-fragment, shows unambiguously the α-sarcin activity against the SRL With this purpose, the lysate was first diluted 3-fold with 40 mM Tris-HCl, pH 7.5, containing 40 mM KCl and 10 mM EDTA. Then, 50 μl aliquots of this dilution (containing 5-6 pmol of ribosomes approximately) were incubated for 15 min at room temperature with different amounts of the tested proteins. The reaction was stopped by addition of 250 μl of 50 mM Tris-HCl, pH 7.4, containing 0.5% (w/v) SDS, followed by strong vortexing. Then RNA phenol/chloroform extraction was performed and the RNA was precipitated from the aqueous phase by adding isopropanol. Finally, the resulting pellet was washed with 70% (v/v) −20 °C ethanol, air dried and resuspended in 10 μl of DEPC H 2 O. α-Fragment release was visualized, after heating the samples at 90 °C for 5 min, by electrophoresis on denaturing 2% agarose gels and ethidium bromide staining.
cell Lines growth and culture. HeLa cells (human cervix adenocarcinoma; CCL-2), obtained from the American Type Culture Collection (Rockville, MD, USA), were used as tumoral CEA-negative cells, whereas colon carcinoma SW1222 cells, provided by Dr. Carl Batt under the partnership Cornell University-Ludwig Institute of Cancer Research, were used as model for the positive ones. HeLa cells were grown as described 47,65 , in Dulbecco's modified Eagle's medium; meanwhile SW1222 cells were maintained in RPMI-1640 medium 25 . Both media contained 300 mg/ml of L-glutamine, 50 U/ml of penicillin, and 50 mg/ml of streptomycin, and were supplemented with 10% fetal bovine serum. Incubation was performed at 37 °C in a humidified atmosphere (CO 2 :air, 1:19, v-v). Harvesting and propagation of cultures were routinely performed by trypsinization. The number of cells was determined using a hemocytometer. flow cytometry. Trypsinized cells were distributed into aliquots of 3 × 10 5 cells and washed three times with phosphate buffered saline (PBS). These aliquots were incubated with different concentrations of purified immunotoxins, using scFvMFE23 as a positive control, for 1 h at room temperature with gentle agitation. A second incubation was performed in the dark with anti-His-Alexa488 (Sigma) diluted 1/100. Between incubations and after the final one, the cells were sedimented by centrifugation (1200 g, 4 °C, 10 min) and washed with PBS for three times. Flow cytometry was performed on a FACScan (Becton Dickinson) and data were analyzed using the WinMDI software.
eLiSA. The ability of IMTXCEAαS, IMTXTRICEAαS and their scFv counterparts to bind human CEA was studied by ELISA as previously described 27 . Plates (Nunc A/S. Roskilde, Denmark) were coated with CEA (0.25 µg/well), washed and blocked with 5% BSA in PBS. Then, 100 µl with the corresponding concentration of the different constructions were added and incubated for 1 hour at room temperature. After three washes, the wells were incubated for one more hour at room temperature with an anti His-tag antibody (BioRad). Washes were repeated and HRP-conjugated goat anti-mouse antibody was added for another hour at room temperature. Then, the plate was washed and developed with the corresponding substrate. Antigen-binding titration was performed with serial dilutions of the purified immunotoxins. Three independent replicates were conducted to calculate the average values. protein biosynthesis inhibition. As previously described [47][48][49] , protein biosynthesis inhibition is the assay routinely used to evaluate the toxic effect of ribotoxins. To evaluate the cytotoxic effect of both immunotoxins, cells were seeded into 96-well plates (1 × 10 4 cells/well) in culture medium and maintained under standard culture conditions for 24 hours. Then 200 µl of free-FBS fresh medium with different immunotoxin concentrations were added. Following 72 hours of incubation at 37 °C, the medium was removed and replaced with fresh one supplemented with 1 mCi per well of L-[4,5-3 H]-Leucine (166 Ci/mmol; Amersham, UK). After 6 hours, medium was removed and cells were fixed with 5.0% (w/v) trichloroacetic acid. The resulting pellet was finally washed three times with cold ethanol and dissolved in 200 µl of 0.1 M NaOH containing 0.1% SDS. Its radioactivity was counted on a Beckman LS3801 liquid scintillation counter. The results were expressed as percentage of the radioactivity incorporated to calculate IC 50 values (protein concentration inhibiting 50% protein synthesis). Three independent replicates per two assays were conducted to calculate the average IC 50 values. Serum stability assays. For testing the structural and functional stability of the new recombinant designs, both immunotoxins were incubated in 60% (v/v) FBS during 96 hours at 37 °C in sterile conditions at a final concentration of 100 nM. Every 24 hours an aliquot was removed and quickly frozen at −80 °C until the entire study was completed. As a control, serum-exposed immunotoxins were frozen immediately to represent a zero time point. Identical aliquots of the different time-points samples were tested for their capability to bind CEA by ELISA and their ribonucleolytic activity. In addition, the immunotoxin molecular integrity was tested by anti-α-sarcin western blot. www.nature.com/scientificreports www.nature.com/scientificreports/ 1:1 PBS-Matrigel (BD Biosciences) mixture. Once the tumor volume reached 50-100 mm 3 , mice were injected intraperitoneally either with PBS or the different immunotoxins. Seven doses, every 48 hours, of PBS or the two different amounts of immunotoxin (25 or 50 µg) were given.
Tumors were routinely measured with an external caliper, and volume was calculated as (width/2) 2 × (length/2). Mice were also weighted throughout the experiment. At the end of the treatment (15 th day), or before in case of potential suffering, animals from control group were sacrificed and tumors were collected for further analysis. After the last treatment dose (15 th day) in immunotoxins-treated groups, the evolution of tumor growth was followed in four mice randomly selected from each group, while the other two mice were sacrificed for comparison with the control group if necessary. Statistical analysis. ANOVA with a post hoc analysis by the Student-Newman-Keuls' test was used to compare variations in the mean tumor sizes at different treatment time points in each experimental group. Differences between experimental groups were considered statistically significant at P < 0.05. All values were expressed as arithmetic means ± sem (standard error of the media). equipment and settings. Gels images from Figs 2A, 4D and 5B were acquired and analyzed using the Gel Doc XR Imaging System and Quantity One 1-D analysis software (BioRad).
Blots images from Figs 2B and 5A were acquired and analyzed using ChemiDoc-It (UVP) and VisionWorks LS analysis software.
If processing in brightness and contrast of gel and blot images has been made it was applied to the entire image including controls. No high-contrast gels or blots has been displayed. When necessary, cropped gels and juxtaposing images were displayed to improve the clarity and conciseness of the presentation, being explicited in the figure.
SigmaPlot -Scientific Data Analysis and Graphing Software (Systat Software Inc.) was used for statystical analysis an graphing of experimental data in Figs 3, 4A-C, 5C, 6 and 7.

Data Availability
All data generated or analysed during this study are included in this published article.