D2F2/E2 cells, a mouse mammary tumor line that has been stably transfected with a vector expressing the human HER2/neu gene and its parent cell line, D2F2 were a generous gift from Dr. Wei-Zen Wei, (Karmanos Cancer Institute, Wayne State University, Detroit, MI USA). Early passage cells were frozen and periodically thawed for experimental use or restocking. Mycoplasma testing was negative using the Impact III PCR profile from IDEXX (RADIL, Columbia, MO, USA). Anti-CTLA4 monoclonal antibody (mAb 9H10) was obtained commercially (BioXcell Fermentation/Purification Services #BE0131, West Lebanon, NH, USA) as was cyclophosphamide (Bristol-Myers Squibb Co., Princeton, NJ, USA). Mice were 8 to 20 weeks of age and weighed 20-25 g. Thy 1.2 BALB/c were obtained from Taconic (Hudson, NY). Animal studies were approved by the University of Pittsburgh institutional Animal Care and Use Committee (IACUC Protocol #: 21028761).

A replicating virus expressing the following properties was created from vector components as previously described (3): Preferential infection of cells expressing human HER2/neu, expression of mouse granulocyte-macrophage colony-stimulating factor, and expression of enhanced green fluorescent protein. Construction used vectors generously supplied by Dr. John K. Rose (Department of Pathology, Yale University, New Haven, CT, USA) and Genentech Inc., South San Francisco, CA, USA).

Female BALB/c mice weighing 20-25g and aged 8-20 weeks were implanted intraperitoneally (IP) with 2 x 10⁶ D2F2/E2 cells in 300 µl PBS. Viral immune-oncotherapy consisted of treatment on day 3 after peritoneal implant with IP rrVSV, 1 x 10⁸ ID, on day 4 with 200 µg anti-CTLA4 mAb IP and on day 5 with cyclophosphamide (CTX), ~100 mg/kg IP. Cure was defined by survival without any symptoms of disease for >100 days after tumor cell implantation. Control mice received the same viral immune-oncotherapy but were never implanted with tumor.

Memory T-cells were obtained from spleens, peritoneal cells or blood of cured animals and controls. Animals were sacrificed prior to spleen cell harvesting. Spleens were harvested, minced and ground through a 70 µM nylon cell strainer (#352350, BD Falcon, Franklin Lakes, NJ). Red blood cells were removed using RBC lysis buffer (Alfa Aesar, J62150AP). Peritoneal washings were performed by injecting 10 ml of sterile PBS into the peritoneum through a 16-gauge needle which was left in place. Two minutes later all the fluid that could be aspirated easily into the syringe was collected, ~ 9 ml. All cells were washed twice with PBS and re-suspended in PBS. Peripheral blood was collected by submandibular punch and dripping the blood into a heparinized tube. Mononuclear cells were separated using Lymphocytes Separation Medium. CD4 and CD8 T-cells were isolated by flow cytometry (see below).

Peritoneal tumors were established in host animals and treated 3 days later by adoptive transfer of splenocytes from cured animals. As previously described, host animals were pre-treated with cyclophosphamide one day before transfer of memory cells (4). Tumor challenge to cured or control mice was achieved by injecting 2 x 10⁶ D2F2/E2 cells IP.

Peritoneal cells were suspended in ice-cold PBS/0.1% BSA/0.2% Azide and stained with combinations of the following antibodies: CD4- APC-eFluor 780 (eBioscience 47-0041-82), CD8a- PE-Cyanine7 (eBioscience 25-0081-82), CD90.2-PE (BD 553006), Live-Dead fixable red Kit (Life Science, NC0836452) and H-2K(d)/TYLPTNASL, human HER2 p63 tetramer conjugated with APC (NIH Tetramer Core Facility at Emory University). Immunofluorescence was quantified using a LSR Fortesa (Becton Dickinson, Mountainview, CA, USA) and cell sorting was performed using a FACSAria II machine (Becton Dickinson) (5).

RNA extraction was completed using Qiagen’s RNeasy Plus Micro kit (Qiagen:74034), according to the manufacturer’s instructions with DNA elimination columns. Briefly, cells were pelleted in a centrifuge at 4°C for 8 minutes at 800xg, then lysed in RLT buffer with beta- mercaptoethanol. RNA was eluted with 22 µl of RNase/DNase free H₂0. RNA quality was assessed using an Agilent HS RNA ScreenTape (Agilent: 5067-5579) on an Agilent 2200 TapeStation. RNA concentration was quantified with a Qubit HS RNA assay kit (Invitrogen: Q32855) on a Qubit 4 (Invitrogen: Q33238).

Libraries were generated with the Takara SMARTer Mouse TCR a/b profiling kit (Takara: 634403) according to the manufacturer’s instructions. Twenty-one cycles were used for PCR1, after which 1 μl of product was used for indexing PCR2 with 19 cycles. Samples with RNA concentrations ≤ 1.0 ng/µl RNA used 22 cycles for PCR1, after which 2 μl of product was used for indexing PCR2 with 19 cycles. Library assessment and quantification was done using Qubit 1x HS DNA assay kit (Invitrogen: Q33231) on a Qubit 4 fluorometer, and a HS NGS Fragment kit (Agilent: DNF-474-1000) on an Agilent 5300 Fragment Analyzer. Libraries were normalized and pooled by calculating the nM concentration based off the fragment size (base pairs) and the concentration (ng/μl) of the libraries.

Sequencing was performed on an Illumina MiSeq using a MiSeq 600 v3 flow cell (Illumina: MS-102-3003). The pooled library was loaded at 13.5nM with 10% PhiX, generating 2x300 bp paired-end reads. Library generation and sequencing was performed by the University of Pittsburgh Health Sciences Sequencing Core (HSSC), Rangos Research Center, UPMC Children’s Hospital of Pittsburgh, Pittsburgh, Pennsylvania, United States of America.

MiXCR (platform 3.0.13; MiLaboratories Inc, Sunnyvale, CA, USA) was used to align and assemble the raw paired-end Fastq files sequencing reads and assemble identical and homologous reads into clonotypes, correcting for PCR and sequencing errors. The program provided detailed information for each clone, including fractions, counts, nucleotide sequence, and amino acid sequence, as well as alignments for multiple samples, repertoire overlap analysis, diversity estimation and segment usage (6). Mouse NKT TRA clonotypes, identified by the canonical CDR3 sequence, CVVGDRGSALGRLHF, were excluded from analysis (7, 8) as were mouse MAIT TRA clonotypes, identified by the canonical CDR3 sequence, CAVRDSNYQLIW (9). This methodology does not exclude TRB clonotypes associate with NKT or MAIT cells but the NKT TRA clonotype was never found at high frequency in peritoneal cells following tumor stimulation indicating that the high frequency response did not contain any NKT cells. Mouse MAIT TRA clonotypes were never found at high frequency in any sample suggesting that these cells were not a confounder in this work. Finally, results for TRA clonotypes, which have definitely excluded NKT and MAIT cells are provided for all analysis and do not change any of the conclusions.

GraphPad Prism version 7.0 (GraphPad Software, Inc., San Diego, CA, USA) was used for presentation and statistical analysis of the data.

Viral immuno-oncotherapy consisting of a replicating recombinant VSV (rrVSV) targeted to Her2, and single doses of anti-CTLA4 Mab and cyclophosphamide cures 3 day implanted Her2+ peritoneal tumors in Balb/c mice. Cure is immunologically dependent, requires both CD4 and CD8 T-cells and generates memory T-cells by 90 days after treatment (5, 10, 11). The anti-tumor memory response was assessed by re-challenging tumor-cured mice with IP administration of tumor cells. TCR clonal analysis, TCRα (TRA) and TCRβ (TRB), was performed on spleen cells harvested 5 days after challenge. A control group consisted of animals who received targeted rrVSV oncolytic immunotherapy, at least 100 days prior to challenge but were never implanted with tumor (virus control mice). All animals in the tumor- cured group showed an oligoclonal set of high-responding splenic TCR clones for CD4 and CD8 T-cells that was not observed in the control group ( Figures 1 – 3 ). The mean clone fractions of the top 3 clones was 5-fold higher for CD4 T-cells and 8-10 fold higher in CD8 T-cells in tumor-cured mice compared with virus control mice ( Figure 3 ). The high responder response to challenge varied among the cured animals but was uniformly low among the control animals ( Figure 2 ). At least one high frequency clone was higher than the highest control value in 5/5 mice (100%) for CD8 T-cells and 4/5 mice (80%) for CD4 T-cells ( Figure 2 ). This highest frequency oligoclonal response in spleen varied from 2-18 CD8 and 0-5 CD4 T-cell anti-tumor TRA or TRB clonotypes per animal (median CD8 = 2.5 and CD4 = 1.5).

Our previous work in this model system has shown that anti-tumor memory CD4 and CD8 T-cells accumulate in the peritoneum following IP tumor challenge in tumor-cured mice (5, 11). In this study, the peritoneal cell response to tumor challenge in experimental and control groups is displayed in Supplemental Figure S1 . The specific migration of anti-tumor memory T-cells to the peritoneum following stimulation was supported by the absence of high frequency peritoneal NKT cells, which could easily be identified by an invariant clonotype and were abundant in the spleens of experimental and control animals (12). Clonal overlap was determined between the spleen and peritoneal response in the same animals to show that the anti-tumor response seen in the peritoneum was reflected in the spleen, providing evidence that the spleen response was an anti-tumor response ( Table 1 ). In total, 70% of high-responding T-cells in the peritoneum were also high-responders in the spleen (62.5% of the CD4 and 82% of the CD8 T-cells). Each animal with a high-responding clonotype in the peritoneum had at least one identical high-responding clonotype in the spleen. The spleen response was therefore an excellent surrogate for the peritoneal response.

Clonal overlap was also determined between the peritoneal cell response among the different animals ( Figure 4 ). These results show that none of the highest-responder clonotypes in one animal were also high-responders in a different animal. As illustrated in Figure 4 , some of the high-responder clonotypes were found in other animals but only at low frequency. Basically, each animal had a private set of high-frequency tumor-responding CD4 and CD8 TCR clonotypes.

Tumor-cured mice (n=4) received peritoneal tumor challenge followed 5 days later by isolation of CD8 T-cells from peritoneal lavage. Each sample was divided in half. One half from each animal had flow cytometry tetramer analysis and TCR clonotyping performed individually for each animal. The other individual samples were combined and tetramer+ CD8 T-cells were identified and separated by flow cytometry. Flow cytometry tetramer analysis and TCR clonotyping was then performed on this mixed sample from 4 mice ( Figure 5 ). A combined sample of tetramer+ CD8 T-cells was required to assure an adequate concentration of mRNA for TCR clonotyping analysis. Purity of the tetramer+ mixture was 86.9% and the delineation between tetramer-positive and tetramer-negative cells with flow cytometry was not absolute ( Figure 5B ), leading to false positive clonotypes in the tetramer+ mixture. True positives were determined by using all high-frequency clonotypes (>1%) in the individual animals as index cases, determining the frequencies of the identical clonotypes in the tetramer+ mix and plotting the inverse ratio ( Supplemental Figure S2 ). High ratios were easily separated from low ratios and had similar patterns in TRA and TRB for each animal. Clonotypes with ratios 0.18 were considered true positives.

Analysis of the 3 highest clonotypes in each animal showed that 5/12 TRB clonotypes and 3/12 TRA clonotypes were tetramer+ ( Figure 6 ). None of these 8 highest frequency responding tetramer+ clonotypes had the same TCR peptide sequence and, in fact each of the 24 highest frequency responding clonotypes had a unique sequence (13). An analysis of all tetramer+ clonotypes found 11 TRB clonotypes and 12 TRA clonotypes with frequency > 1% ( Figure 7 ). The number of these tetramer+ clonotypes per mouse varied from 2 to 4 (median = 3) and each clonotype was unique ( Figure 7 ). The same clonotype was sometimes found in other animals (5 TRA and 5 TRB) but never at frequency >0.15% and usually much lower ( Figure 7 ). In summary, each animal had its own private high-frequency tetramer+ sequence and some had more than one. An unexpected finding was that tetramer+ frequency based on TRB clonotype frequency closely matched clone frequency results based on flow cytometry ( Figure 7 ). TRA frequency was not as closely matched probably because ~30% of T-cells express 2 functionally rearranged TRA mRNAs (14, 15).

The CD4 and CD8 TCR clonotype response in spleen and peritoneum to IP tumor challenge was compared in donor animals cured of implanted tumor by rrVSV oncolytic immunotherapy and host animals cured by T-cell transfer from the donor animals ( Figure 8 ). The critical finding was that the same high-responding TCR clonotypes were found in all 4 samples ( Figure 9 for TRB and Supplemental Figure S3 for TRA), strongly suggesting that they represented anti-tumor memory clones. The high responder peritoneal CD4 and CD8 T-cells in the host animals were derived from the donor mice because they had matching clonotypes. Independent clones arising from the hosts would have had their own private sequences as shown in the previous sections. These were the therapeutic T-cells because cure was obtained solely by transfer of donor spleen cells. These cells were first recognized by IP challenge in the donor mice demonstrating that a high clonotype response to challenge is a reliable method to identify anti-tumor CD4 and CD8 T-cell clones. Peritoneal clonotypes were classified as definite, active anti-tumor memory if they appeared in both donor and host peritoneal cells at a frequency greater than the mean of the top 10 clonotypes from all control animals ( Figure 9 ). The number of unique high-frequency anti-tumor memory TRB clonotypes in the peritoneum varied from 1 to 8 in CD4 and 4-8 in CD8 T-cells with a median of 4 CD4 T-cells and 6 CD8 T-cells. Donor spleen and host peritoneal clonotypes were also classified as definite, active anti-tumor memory if they appeared in both host PC and donor spleen at a frequency greater than the mean of the top 10 clonotypes from all control animals ( Figure 9 ). The number of high-frequency unique anti-tumor memory TRB clonotypes in the donor spleen varied from 1 to 8 for CD4 and 5-7 for CD8 T-cells with a median of 3 CD4 T-cells and 6 CD8 T-cells. The frequency of anti-tumor CD4 TRB clonotypes was 12.7-fold higher in host peritoneum than host spleen and 7.9-fold higher for CD8 TRB clonotypes, confirming our previous work, that anti-tumor memory CD4 and CD8 T-cells migrate to and expand at the site of challenge in tumor-cured mice (5, 11). TRB clonotypes were used in these analyses because their frequency had the best correspondence with clone frequency, as determined above. Importantly, the anti-tumor memory clonotypes were also found in blood, at similar frequencies to spleen, indicating that these analyses may be practical clinically ( Figure 9 and Supplemental Figure S3 ).

The resting frequency of anti-tumor memory T-cells was determined by harvesting spleen cells from cured, unchallenged, resting, donor mice and assaying TCR clonotypes from one aliquot of 1 x 10⁷ cells. A second aliquot of 5 x 10⁷ spleen cells were challenged by transfer to a host animal with a 3 day implanted peritoneal tumor. Host spleen and peritoneal TCR clonotypes were assayed 5 days later ( Figure 10 ). Clonotypes in the resting donor cured tumor mice that were identical to high frequency clonotypes in the peritoneum of challenged mice were anti-tumor memory clonotypes. Independent clones arising from the hosts would have had their own private sequences as shown in the previous sections. In addition, only memory cells from the donors and not naïve cells from the hosts, could be present at sufficient concentration and multiply rapidly enough in 5 days to produce the high frequency clones found in the peritoneum. Challenge of naïve animals with peritoneal tumor cells produced only low clonotype frequencies at 5 days (data not shown). Finally, our current and previous work has shown that spleens from donor mice cured of tumor by rrVSV oncolytic immunotherapy contain potent therapeutic anti-tumor memory T-cells (4, 5, 10, 11, 16, 17). Using the top 5 peritoneal clonotypes as index clonotypess, all CD4 and CD8 T-cell TRB clonotypes had identical clones in the donor unchallenged mice, although 2 CD4 clonotypes from one animal had notably lower resting frequencies than all the others ( Figure 11 for TRB and Supplemental Figure S4 for TRA). The mean frequency of these anti-tumor memory T-cells in resting T-cells averaged 0.028% for CD4 T-cells and 0.11% for CD8 T-cells ( Figure 12 ). The highest frequency clone for CD4 T-cells was 0.15% and for CD8 T-cells was 0.34%. These findings match well with tetramer-based studies of anti-viral human memory CD4 and CD8 T-cells and mouse CD4 T-cells but are 30-fold lower than several previous reports of mouse anti-viral memory CD8 T-cells. All studies agree that resting memory CD8 T-cells form a higher fraction of total CD8 T-cells than resting memory CD4 T-cells form of total CD4 T-cells (13, 18–24). TRA data was similar ( Supplemental Figure S4 ) but we focused on TRB clonotypes as the best surrogate for TCR clones for several reasons: T-cells almost invariable express a single TRB mRNA but ~1/3 express 2 TRA mRNA (14, 15); the main contribution to TCR-peptide binding comes from the TRB CD3 sequence (25); and data presented above showed a close correspondence between TRB clonotype frequency and cell clonal frequency.

Resting donor spleen TRB clonotypes had ~ 1/10 the frequency of challenged memory clonotypes in spleen and ~1/100 the frequency of challenged memory clonotypes in the peritoneum ( Figure 12 for TRB and Supplemental Figure S5 for TRA). CD4 T-cells increased from 0.028% to 0.26% to 2.44% and CD8 T-cells increased from 0.11% to 0.79% to 8.39%. The mean increase in TRB clonotype count from transferred donor T-cells to combined harvested spleen and peritoneal T-cells was 65-fold for both CD4 and CD8 memory T-cells ( Figure 13 for TRB and Supplemental Figure S6 for TRA). This indicated a minimum mean number of 6 doublings in 5 days because potential anti-tumor T-cells in other tissues such as lymph nodes, bone marrow and blood were not harvested. Doublings in individual clones varied from 2 to 8 ( Figure 13 ).

Frequency analysis of resting T-cells does not distinguish anti-tumor memory T-cells from background ( Figure 14 for TRB and Supplemental Figure S7 for TRA). The top 5 most frequent CD4 and CD8 TRA clonotypes from 3 animals are highlighted by large bold red, green or black diamonds within the frequency distribution of all resting spleen cells from the same animal. A stimulation is required to identify and quantify anti-tumor memory T-cells ( Figures 1 , 6 , 9 , 11 ). In this sample size of 30, potential resting anti-tumor memory T-cell clones, 28 were found within the top 625 clones. As expected, following initial expansion, these resting memory cells did not diminish to rare frequency but neither did they stand out from the mixed background of resting frequencies.