A cocktail of antibodies to multiple tumor-specific neoantigens increases binding and inhibits tumor growth

Background: Antibodies that target a single tumor antigen fail to cure stage IV cancer patients due to tumor heterogeneity resulting in variable expression of antigen. Tumor cells with insufficient binding of antibody will not undergo antibody induced cytotoxicity. We describe targeting multiple tumor-specific antigens that resulted in homogeneous dense binding to mouse melanoma cells and significant tumor growth inhibition. Methods: Surface-related tumor-specific mutations on B16-F10 cells were identified. Peptides containing the single amino acid mutation were synthesized for 9 different neoantigens. Rabbits were vaccinated with each of these peptides and high affinity polyclonal antibodies to each peptide were obtained. The 9 antibodies were combined as a cocktail and mice with implanted B16-F10 cells were treated with and without PD1 inhibitor. Results: Even a single dose of the antibody cocktail inhibited tumor growth and prolonged survival. PD1 inhibitor alone had little effect on tumor growth. The antibody cocktail plus PD1 inhibition increased tumor response and 4 doses of the cocktail completely prevented tumor growth in 50% of the mice. Complete responses were durable. The complete responders were highly resistant to tumor re-challenge at 6 months. No adverse events were identified in the antibody treated mice. Conclusions: Multiple tumor-specific cell surface-related neoantigens were abundant in B16-F10 cells. Antibodies to 9 of these neoantigens had variable binding but when combined had dense homogeneous binding. Even one dose of this cocktail of 9 antibodies improved survival and when multiple does were combined with PD1 inhibition 50% of the mice were rendered permanently tumor free.

3 single overexpressed antigen. The unresolved problem with this approach is that within the heterogeneous cancer cell population there are cancer cells with low antigen expression. These cells will not bind sufficient antibody and will continue to grow. This explains why Herceptin, considered to be a prototype blockbuster antibody, only delays time to progression by 3 months. None of the patients are cured [1].
Antibodies are highly capable of initiating cytotoxicity of target cells. Decades ago it was shown that serum transfer from animals resistant to a tumor would induce tumor resistance in a recipient animal [2][3][4][5][6][7]. Superb electron and fluorescent microscopic images graphically demonstrated multiple types of effector cells engaged in antibody directed cell destruction [8,9]. Target heterogeneity was understood as early as 1962, and it was shown that resistance to antibodies could be overcome by mixtures of antisera demonstrating that "cytotoxic sensitivity is related to the surface concentration of reactive sites" [10]. Even a cocktail of 3 or 4 antibodies was shown to increase bioactivity against tumor cell lines [11]. After observing that combinations of monoclonal antibodies increased effectiveness the authors suggested that cocktails of antibodies may be useful in the clinic [12].
In this study, we have attempted to overcome tumor heterogeneity in the B16-F10 mouse model. We focused on surface related tumor-specific neoantigens. Such tumor-specific mutations are abundant and largely overlooked because they are highly variable between individuals, are not driver mutations, and they are not necessarily overexpressed. We generated rabbit antibodies to multiple tumor-specific neoantigens and combined them together as a cocktail. The cocktail substantially increased cell surface binding across the cell population. We demonstrate that this strategy overcomes tumor heterogeneity and results in substantial inhibition of tumor growth.

Methods 4
Cells and reagents: B16-F10 melanoma tumor cells (ATCC CRL-6475) were procured from American Type Culture Collection (ATCC, Manassas, VA). Dulbecco's Modified Eagle's Medium (DMEM) and Trypsin-EDTA solution were purchased from ATCC. All other reagents used in this study were of molecular or high purity grade. Rat anti-mouse PD1 (CD279) antibody (clone  and rat IgG2a isotype control of anti-PD1 antibody (clone 2A3) were purchased from Bio X cell (West Lebanon, NH). Alexa Fluor 568 conjugated goat-antirabbit antibody and cross-absorbed goat anti-mouse IgG (H+L)-horseradish peroxidase (HRP) conjugate were procured from Life Technologies (Carlsbad, CA).
Neoantigens: Using sequence data of B16-F10 mouse melanoma cells [13], we selected multiple cell-surface-related mutated proteins (neoantigens) that had a single amino acid change. The peptide length was kept to 11 amino acids, keeping mutated amino acid in or near the center ( Table 1). The peptides were analyzed for their B-cell response predictions [14] and immunogenicity by determining the homology of the selected mouse antigen epitope with the same in host species. Antibodies to the selected peptides were commercially supplied by GenScript (Piscataway, NJ).
Anti-neoantigen antibodies: The purified antibodies supplied by GenScript were titered against the mutated peptides for affinity estimation (EC50) using ELISA [15]. Following peptide immobilization in 96-well plates, the wells were blocked with 1% BSA in PBS. The plates were washed 3 times with PBS and incubated with serially (2-fold) diluted purified antibodies. Wells were washed again 3 times and incubated with HRP-conjugated antirabbit IgG antibodies. The binding was monitored 450 nm with 3,3',5,5'tetramethylbenzidine (TMB) reagent as an HRP substrate.
Immunofluorescence microscopy of anti-neoantigen antibodies binding to B16-F10 cells: B16-F10 whole cell slides and B16-F10 mouse tumor sections were fixed in 3% paraformaldehyde and washed with PBS (3 times, 3 min each) for immunofluorescence microscopy by the method described previously [16]. Briefly, the B16-F10 tumor cells and sections were blocked with Image-it Fx signal enhancer (Invitrogen, Carlsbad, CA) and were incubated with individual polyclonal anti-mutated peptide antibodies and their cocktail for 1 h at room temperature. Slides were rinsed 3 times in PBS and incubated for 30 min with Alexa Fluor 568-conjugated goat anti-rabbit antibody (Invitrogen). The slides were rinsed again in PBS, 3 times for 3 min each, and stained with 1000x diluted DAPI (Invitrogen) for 10 min. The slides were cover-slipped and examined (Ex 579 nm/Em 603 nm and Ex 359 nm/Em 461 nm) using Nikon TE2000-U inverted fluorescence microscope (Nikon Corp., Kangawa, Japan).

Animal tumor model:
The spontaneous C57BL/6-derived B16-F10 melanoma tumor cells were used as a model system for this study. It is a well-established and widely used syngeneic tumor model representing an aggressive tumor that is difficult to treat. Six to seven weeks old (date of birth ± 3 days) female mice of C57BL/6J strain were purchased from The Jackson Laboratory (Bar Harbor, ME). The mice were allowed to acclimate in the animal facility for one week before starting the experiment. The animal facility was       The animals of all the experiments were weighed 2 times/week and the tumor size were followed by measuring tumor volume (V = w 2 x l/ 2 ) every day starting from 7 days postimplantation using an electronic caliper. The survival time was calculated based on the death of an animal or euthanized following reaching to ≥2000 mm 3 tumor volume.
Following our IACUC approved protocol, the tumor-bearing animals were euthanized when tumor reached ≥2000 mm 3 or when animals exhibit signs of dehydration, difficulty walking, cachexia or other signs of physical distress, whichever comes first.

ELISA for binding of tumor-bearing mouse plasma antibodies to B16-F10 tumor
lysate: The lysate tumor preparation and lysate ELISA were done by the methods described previously [17]. Briefly, the lysate of B16-F10 tumors grown in untreated mice were prepared by the cycles of freezing/thawing and ultrasonication. Following 50μL cell lysate (1 mg/mL protein) immobilization and well-blockage with 1% casein-TBS (Thermo Statistical analyses: Tumor growth differences in the animals of different groups were analyzed by One-way and Two-way ANOVA followed by Tukey's multiple comparison test. Kaplan Meier plot were used for survival data presentation and the survival curve comparisons for determining the significance of difference in groups were analyzed using 9 Log-rank (Mantel-Cox) test and median survival values. Prism software (GraphPad, San Diego, CA) was used for some of the analyses.

Results
Polyclonal rabbit antibodies demonstrated high-affinity antibodies against each mutated peptide. Fig. 1 displays the results of characterization of these antibodies. Following a one-week acclimation period in our animal facility and prior to any treatment, the weights of each animal were obtained for baseline measure.

Animal Experiment 1:
The effects of inoculating different numbers of B16-F10 cells (10,000 -300,000) on tumor volume are presented in Supplementary Fig. S1. Tumor volumes increased more rapidly with higher numbers of tumor cells inoculated.
By day 21 DPI, animals died or were euthanized because tumor size reached 2000 mm 3 .
No unexpected weight changes were observed during this period. All treatment experiments used 300,000 tumor cells for implantation based on consistent rapid growth and lack of apparent side effects during tumor growth period. Also, during this short period of tumor growth, plasma antibodies to B16-F10 tumor lysates above background levels were not detectable (data not shown).

Animal Experiment 2:
Tumor growth and mouse survival data following antibody cocktail treatment are presented in Fig. 2. The results demonstrate tumor growth inhibition by the treatment with the 9-antibody cocktail ( Fig. 2A). Two-way ANOVA showed significant tumor  Fig. 5 shows that 6.5 months after a complete response to 4 doses of antibody cocktail and PD1i (Fig. 4), these mice were persistently resistant to reinoculation without any further treatment. It is highly unlikely than any of the rabbit antibodies were present at the time of re-inoculation. It appears that following the initial treatment and complete response, these mice developed a robust anti-melanoma adaptive immune response. Fig. 5 includes a set of normal untreated mice for comparison.

13
In this study, we targeted multiple tumor-specific proteins that were not overexpressed Selection of mutated targets was based upon published NGS sequence data of B16-F10 14 melanoma cells [13]. We did not attempt to select for driver mutations. We did not select based upon expression data. The only criteria were that the mutation was related to the cell surface by its reported presence in the membrane or as a secreted protein. For membrane proteins we restricted the selection to those with mutations not in the cytoplasmic or intramembrane portion. These selection criteria retrieved many possible targets. We designed short peptides representing the mutated section of the protein with the mutated amino acid in the center. The peptide design was based on our previous successful vaccination experiments with mice [15]. The set of peptides predicted to be immunogenic in the rabbit were then synthesized and used as vaccines. All of the 9 selected peptides produced high titers of antibodies. Affinity enrichment yielded 4 to 6 mg of high-affinity polyclonal antibodies. These simple selection criteria were surprisingly productive at producing desirable antibodies. The majority bound B16-F10 cells and when combined together produced remarkable tumor inhibition of a relatively aggressive tumor.
To assess the possible applicability of our approach for treating cancer patients we analyzed the single non-synonymous mutations of four breast cancer patients from published sequence data to determine how many mutations were membrane-associated [21]. The total number of gene mutations in these patients ranged from 15 to 221. We determined the cell location of each identified gene mutation using the human protein atlas project [22] or the human gene database through GeneCards [23]. We found that 27%-46% of all the nonsynonymous mutations reported in these breast cancer patients were proteins found on the plasma membrane. This information indicates that cancer cell surface mutations in breast cancer patients are not uncommon and will be potentially available for targeting with antibodies.
We observed no adverse events in the treated mice. Antibodies, even from different species, are not inherently toxic [24][25][26]. The strategy of using tumor-specific mutations may reduce cross-binding to normal proteins. Using a cocktail which diminishes the dose of each antibody may further diminish possible normal cell toxicity to any individual antibody. A normal cell that expresses a cross-reactive antigen would be exposed to a relatively low dose of that individual antibody. The likelihood of normal cells binding to additional antibodies from the cocktail becomes increasingly remote as the number of different antibodies in the cocktail increases.
Cell surface mutations that are not overexpressed or involved in maintaining the malignant phenotype have been largely ignored as potential targets for therapy. It appears that such mutations are in abundance. It also appears that there is considerable variation between patients in which proteins are mutated. These types of mutations would normally be considered of low value from the perspective of developing single pharmaceutical reagents that can be used for many patients.
However, the success rate for curing stage IV cancer patients using the conventional strategy of a single antibody targeting a single overexpressed protein is close to zero.
Despite published lists of desirable targets to help guide researchers, the search for a single overexpressed universal target in cancer patients is elusive likely does not exist [27,28].We propose a new approach that if successful will fundamentally change how patients will be treated with antibodies. This will be accomplished by using multiple antibodies that bind multiple tumor-specific antigens resulting in dense homogeneous binding of all of the cancer cells. Antibodies are highly bioactive and readily induce cancer cell cytotoxicity as long as antibody binding achieves a critical threshold. In clinical practice, however, monoclonal antibodies never functionally achieve critical threshold

Consent for publication: Not applicable.
Availability of data and material: Not applicable.
Competing interests: The authors declare that they have no competing interests. Authors' contributions: GS designed the studies, helped acquire the data, interpreted the data, assisted in writing the manuscript, and is personally accountable for this work. YS helped acquire the data, interpreted the data, assisted in writing the manuscript, and is personally accountable for this work. SP helped acquire the data, interpreted the data, assisted in writing the manuscript, and is personally accountable for this work. DK designed the studies, interpreted the data, assisted in writing the manuscript, and is personally accountable for this work. Tables   Table 1 List of the peptides that were derived from the selected mutated B16-F10 tumor cell proteins and used for rabbit polyclonal antibodies production.      Tumor growth in the mice that had a durable complete response to tumor challenge ( Fig. 4)