Participants and clinical data. Patients with biopsy-confirmed locally advanced cervical, vaginal, vulvar, and anal cancers were enrolled in an IRB-approved (2014-0543) multi-institutional prospective clinical trial at The University of Texas MD Anderson Cancer Center and the Harris Health System Lyndon B. Johnson Hospital Oncology Clinic from September 22, 2015, to January 11, 2019. Cases of cervical and anal cancer were staged according to the FIGO 2009 staging system and the AJCC 7th edition staging system, respectively, which were in place when the study began. Patients were required to have a visible tumor and planned standard-of-care treatment for intact cancer. For cervical cancer, this involved definitive radiotherapy, including external beam radiotherapy and brachytherapy, with concurrent cisplatin. Patients with anal cancer received external beam radiotherapy with concurrent cisplatin and 5-fluorouracil. Patients with vaginal cancer received external beam radiotherapy and concurrent cisplatin with or without brachytherapy. Patients with vulvar cancer received concurrent cisplatin without brachytherapy. Patients with any previous pelvic radiotherapy were excluded. Patients receiving external beam radiotherapy received a minimum radiation dose of 45 Gy in 25 fractions over 5 weeks. Clinical, demographic, and pathologic data were collected prospectively.
Sample collection, DNA extraction, and TCR sequencing. Radiation oncology or gynecologic oncology clinicians at either MD Anderson Cancer Center or Lyndon B. Johnson Hospital used Isohelix buccal swabs (Isohelix, cat. #DSK-50) to collect tumor DNA samples. Patients underwent swabbing 5 times: at baseline, at the end of week 1 (after 5 fractions), at the end of week 3 (after 10-15 fractions), within 1 week before the first brachytherapy treatment or at the time of brachytherapy (week 5), and at the first follow-up visit (typically 12 weeks after treatment). For each patient, all samples were obtained from the same general tumor region. Peripheral serum samples were also collected at baseline, week 1, week 3, week 5, and follow-up. DNA was extracted from the tumor swab samples with the Isohelix Xtreme DNA lysis kit per the manufacturer’s instructions (Isohelix, cat. #XME-50). DNA was extracted from blood samples after Ficoll gradient separation with the DNeasy Blood and Tissue Kit (Qiagen, cat. #69504) according to the manufacturer’s instructions and without modification for erythrocytes. At these same timepoints, cytobrush samples of the patients’ tumors were also collected as described previously12 and used for flow cytometry.
Blood samples were collected in 10-mL ethylenediaminetetraacetic acid (EDTA)-containing tubes (BD Biosciences, cat. #366643) and diluted with 1 ´ phosphate-buffered saline (PBS) at a 1:1 ratio by volume. The diluted blood was then distributed into 50-mL conical tubes containing 8-mL aliquots of Ficoll-Paque PLUS Media (GE Healthcare, cat. #17144002). Each sample was spun at 400 x g with the brake off for 40 minutes at room temperature. The peripheral blood mononuclear cell layer was then removed using a serological pipette and placed into a Falcon 50-ml conical centrifuge tube. Twenty-five milliliters of 1 x PBS was added, and the sample was centrifuged at 400 x g for 10 minutes. The supernatant was discarded, and the pellet was resuspended in 10 mL of RPMI 1640. The cells were washed once more by centrifuging the sample at 400 x g for 10 minutes, disposing the supernatant, and then resuspending the cells in 10 ml of RPMI 1640. The cells were then counted and separated into aliquots of approximately 5 x 106 cells each. The aliquots were centrifuged again at 400 x g for 10 minutes at room temperature, the supernatant was removed, and the dry pellets were frozen in cryovials.
To Isolate DNA from peripheral blood mononuclear cells for TCR sequencing, the DNeasy Blood and Tissue Kit (Qiagen, cat. #69504) was used following the manufacturer’s instructions without modification for erythrocytes. Proteinase K (20 µl) and 5-10 µl of the frozen pellet were added to a 2-ml microcentrifuge tube and PBS was added to bring the total volume to 220 µl. Ethanol-free Buffer AL (200 µl) was added to the tube, which was vortexed to mix and homogenize the sample, which was then incubated at 56°C for 10 min. Following incubation, 200 µl of ethanol was added to the sample and mixed thoroughly via vortex. The entire mixture was then placed into a spin column in a 2-ml collection tube. The tube was centrifuged at 6000 x g for 1 minute at room temperature, and the collected liquid was discarded. The spin column was transferred into a new 2-ml collection tube, 500 µl of Buffer AW1 was added, and the column was centrifuged for 1 minute at 6000 x g at room temperature; the collected liquid and tube were discarded. The spin column was transferred into a new collection tube, 500 µl of Buffer AW2 was added, and the tube was centrifuged for 3 minutes at 20,000 x g at room temperature to completely dry the column’s membrane. The collected liquid and 2-ml collection tube were discarded. The dried column was placed into a 2-ml microcentrifuge tube, and 200 µl of Buffer AE was added to the membrane, which was incubated at room temperature for 1 minute. The tube was centrifuged for 1 minute at 6000 x g at room temperature. The resultant DNA elution was then used for TCR sequencing.
Multiplex PCR-based deep sequencing of the CDR3 region of TCRβ was performed using the proprietary immunoSEQ immune profiling system (Adaptive Biotechnologies). This system uses a library of known forward primers, each specific to a TCR Vβ segment, and reverse primers specific to a TCR Jβ segment. Both productive templates and nonproductive templates (CDR3 regions predicting out-of-frame receptor genes or premature stops) were assessed, but only the productive templates were included in the final analysis.
HPV Genotyping. Tumor swab DNA isolates were applied to the Linear Array HPV Genotyping Test and Linear Array Detection Kit (Roche, cat. #04472209 190 and #03378012 190, respectively).
The Working Master Mix was prepared by adding 125 mL of HPV Mg2+ to one vial of HPV MMX and mixing by inversion 10-15 times. Then, 50 mL of Working Master Mix was combined with 50 mL of isolated DNA in each reaction tube. Amplification was performed in an Applied Biosystems Gold-plated 96-Well GeneAmp PCR System 9700 with the following program: HOLD 2 min at 50°C; HOLD 9 min at 95°C; CYCLE (40 cycles, ramp rate 50%) 30 sec at 95°C, 1 min at 55°C, 1 min at 72°C; HOLD 5 min at 72°C; HOLD 72°C Indefinitely. Less than 4 hours after amplification, 100 mL of Denaturation Solution (DN) was added to the amplification products and mixed by pipetting.
For the hybridization reaction, HPV Strips containing probes were placed in wells of a 24-well tray (Roche, cat. #03140725 001). Working Hybridization Buffer (100 mL SSPE, 12.5 mL SDS, and 388 mL deionized water) and 75 mL of denatured amplicon was added to each well. The tray was hybridized in a shaking water bath at 53°C for 30 min with a shaking speed of 60 RPM. For this step and for each following step, each buffer was removed from the strips by vacuum aspiration. The strips were washed first with 4 mL Working Ambient Wash Buffer (133 mL SSPE, 13.3 mL SDS, and 2520 mL deionized water) by rocking plate 3-4 times, and then with 4 mL Working Stringent Wash Buffer in the shaking water bath at 53°C for 15 min with a shaking speed of 60 RPM.
To begin the detection process 4 mL of the Working Conjugate was added to each well and incubated for 30 min at room temperature on an orbital shaker at 60 RPM. To wash the conjugate off the strips, three rinses were performed by adding 4 mL Working Ambient Wash Buffer. In the first of these rinses, the tray was rocked gently 3-4 times, but for the second and third rinses the tray was shaken at 60 RPM on an orbital shaker at room temperature for 10 min. Then, 4 mL of Working Citrate Buffer (25 mL CIT and 475 mL deionized water) was applied to each well and the tray was shaken at 60 RPM on an orbital shaker at room temperature for 5 min. After removing the final buffer by vacuum filtration, 4 mL of Working Substrate (4 mL SUB A and 1 mL SUB B) was added to the wells and the tray was shaken at 60 RPM on an orbital shaker at room temperature for 5 min. The substrate was aspirated from the wells and a final rinse of 4 mL deionized water was applied to each well containing a strip.
Each strip was removed from the tray by forceps and dried on a clean surface for an hour. Results were interpreted by aligning each strip with the Linear Array HPV Genotyping Reference Guide. HPV genotypes corresponding with positive bands on the strips were recorded for each sample. The results were validated by confirming that the negative control showed no bands, the positive control showed bands for HPV16, and the b-globin internal controls were present on each sample strip and the positive control strip.
Flow cytometry. Cells were dislodged from cytobrush samples using vortex agitation. Dithiothreitol solution (1X Hank’s balanced salt solution, 4% bovine serum albumin, 2 mM dithiothreitol; Invitrogen, cat. #P2325) was added for mucous breakdown if large amounts of mucous were present, and cells were passed through a 70-um cell strainer. After centrifugation, cell pellets were suspended in sterile complete RPMI media containing penicillin-streptomycin and gentamicin antibiotics (Fisher Scientific, cat. #SH30027FS, #SV30010, and #BW17-518Z, respectively).
Lymphocytes were isolated from peripheral serum samples by density gradient centrifugation with Lymphoprep medium (Fisher Scientific, cat. #NC0460539). The mononuclear cell layer was collected and resuspended in RPMI media.
Custom peptide sequences of 10 E6 and 4 E7 domains were commercially synthesized (Biosynthesis Inc). Cervical brush-derived cells were incubated with 10 ug/mL of pooled peptide at 37°C overnight (12-16 h). The next day, Golgiplug (BD Biosciences, cat. #555029) was added to the solution, and the solution was incubated at 37°C for 4-6 h to allow for IFNɣ to accumulate. As a positive control for immune cell activation and cytokine production, following overnight incubation in media alone, cells were incubated with cell activation cocktail containing PMA/ionomycin (BioLegend, cat. #423301/2) and Golgiplug at 37°C for 4-6 h prior to staining.
Lymphocyte immunostaining was performed according to standard protocols. Briefly, cells were fixed using the FOXP3/Transcription Factor Staining Buffer Set (eBioscience, cat. #00-5523-00) and stained with a 16-color panel of antibodies from BioLegend, BD Bioscience, eBioscience, and Life Technologies. Cells were stained with intracellular monoclonal antibody for 30 minutes at 4°C in the presence of anti-Cd16/Cd32 monoclonal antibody (BD Bioscience), fixed with FOXP3/Transcription Factor Staining Buffer Set (eBioscience, cat. #00-5523-00), and held in flow cytometry staining buffer (2 mM EDTA, 2% fetal bovine serum; Corning). Counting beads (Thermo Fisher) were used for single-color controls. The cells were analyzed using a 5-laser, 18-color LSRFortessa X-20 Flow Cytometer (BD Biosciences) and FlowJo 10.6.1.
Functional expansion of antigen-specific intratumoral T-cells. Cytobrush samples were collected as described previously26. Each sample was placed in a 15-ml conical tube and immediately transported to the lab, where 5 ml of cold RPMI-1640 with 10% fetal bovine serum and 1% penicillin/streptomycin were added to the tube. The tubes were vortex-agitated to dislodge cells into the media. For brushes with substantial mucus, 2 mm of dithiothreitol (Sigma) in Hanks’ balanced salt solution (Invitrogen) with 4% bovine serum albumin (Sigma) was added and allowed to rest for 10 min at room temperature, after which cells were passed through a 70-um cell strainer. The samples were centrifuged for 5 minutes at 400 x g at room temperature and then washed twice by centrifuging and resuspending them in 10 mL of room-temperature RPMI media. After the second wash, cells were counted and resuspended so a minimum of 1.0 x 106 cells per 100 μl of solution could be placed in each well. One hundred microliters of solution were distributed into each of 3 wells on a 96-well round bottom plate. Each well contained 100 μl of a custom mix of HPV13-SLP peptide (10 ug/mL) (Supplemental Table 1)11–16, 100 μl of PMA (50 ng/ml)/ionomycin (250 ng/ml) cocktail, or 100 μl of RPMI. The cells were cultured in an incubator at 37°C overnight (16-20 h). At that time 0.4 μL of Golgiplug at a 1:500 concentration was added to the wells and the plates incubated for another 4 to 6 h. The plate was centrifuged and the media aspirated, and the cells were reconstituted with 200 μl of room-temperature 1x PBS and transferred into two 1.5-ml microcentrifuge tubes for DNA isolation.
DNA isolation was performed using the QIAamp UCP DNA Micro Kit (Qiagen, cat. #56204) according to the manufacturer’s instructions.
All centrifugation steps were performed at room temperature, and samples and buffers were brought to room temperature prior to centrifugation. Proteinase K (10 μl) was added to the microcentrifuge tubes containing 100 μl of media. Then, 100 μl of Buffer AUL was added, and the samples were pulse-vortexed for 15 seconds and incubated at 56°C for 35 minutes. Samples were centrifuged at 17,800 x g for 3 minutes, and then 50 μl of ethanol was added. The samples were then pulse-vortexed and then incubated for 3 minutes at room temperature. The samples were centrifuged at 13,500 rpm 17,800 x g spin for 3 minutes and then transferred to the QIAamp UCP MinElute columns in 2-ml collection tubes and centrifuged at 13,500 rpm 17,800 x g for 5 minutes. Each column was transferred into a second collection tube and spun once more before being placed in another clean column. Buffer AUW1 (500 μl) was added to the column, which was centrifuged once more and then transferred to a clean collection tube. Buffer AUW2 (500 μl) was added to the column, which was again centrifuged and placed in a clean collection tube before being centrifuged once more at 17,800 x g for 5 minutes. The column was placed into a clean 1.5-ml centrifuge tube, and 100 μl of microbial DNA-free water was added to the column, which was incubated for 10 minutes at room temperature before being centrifuged at 17,800 x g for 5 minutes. Buffer AUE (100 μl) was added to the membrane, which was incubated for 10 minutes and centrifuged for 5 minutes. Next, 50 μl of Buffer AUE was added to the membrane, which was incubated for 10 minutes and then centrifuged for 5 minutes. Extracted DNA was then stored at 4°C until use. Isolated DNA was subjected to TCR sequencing (described above).
Analysis of flow and T-cell repertoire characteristics. The TCR metrics we studied were Total Templates, Productive Templates, Total Rearrangements, Productive Rearrangements, Productive Clonality, Sample Clonality, Productive Entropy, Max Productive Frequency, Max Frequency, and Out of Frame Rearrangements. To study changes in the flow characteristics over time, we compared the means for blood and tumor samples at each time point with baseline means by using a paired sample t-test. We compared median changes from baseline for TCR characteristics by using a Wilcoxon signed-rank test. We calculated the log2 fold change of any variable in blood or tumor samples that significantly changed from baseline. We then performed a Wilcoxon signed-rank test to assess if the degree of change differed between the blood and tumor samples. We also fit univariate Cox proportional hazards models for each of the flow and TCR variables at the static timepoints as well as for the fold changes in blood and tumor samples. Clustering of immune variables with and without clinical characteristics was performed using both machine learning algorithm “mclust” and using unsupervised hierarchical clustering. HPV remodeling was defined as a >1.5fold change in proportions of HPV-responsive clones.
Identification and analysis of HPV-specific clones. Public and exclusive repertoires were created for HPV peptide, PMA, and controls based on amino acid overlap in CDR3 sequences. Each sequence was annotated using the mcPAS21, VDJdb27, and TBAdb28 databases, downloaded on March 24, 2020. We then used the R package immunarch29 to monitor the frequency of these HPV-specific CDR3 amino acid sequences over time as well as the most abundant clones overall. We also calculated the numbers of unique samples and patients that had each HPV-specific clone. To study whether the HPV-specific repertoire changes over time, we grouped clones from 2 patients based on the clones’ presence in HPV, PMA, and control wells into 7 groups: HPV+PMA+CTRL+, HPV+PMA+CTRL-, HPV+PMA-CTRL+, HPV+PMA-CTRL-, HPV-PMA-CTRL+, HPV-PMA+CTRL-, and HPV-PMA+CTRL+. We calculated the number and proportion of clones present in each sample and determined the clones’ presence in each sample as a binary variable. We then compared the number and proportions of clones in blood and tumor samples for all 7 groups at each timepoint to those at baseline by using the Wilcoxon signed-rank test. We calculated the log2 fold change from baseline to determine if the degree of change is associated with survival. We also assessed whether the number and proportion of clones in these groups’ blood and tumor samples were associated with survival by using a univariate Cox proportional hazards model.
We also performed the full analysis using subsets of data stratified by cancer type (anal vs. cervical, vaginal, and vulvar) and histology (squamous cell carcinoma vs. adenocarcinoma and adenosquamous carcinoma) and by HPV+ and HPV16+ subsets.
TCR motif identification and analysis. Significant motifs in the CDR3β portion of patient T-cell sequences relative to expected frequencies in a reference set of unselected naïve TCRs were identified using the (GLIPH) algorithm20. A local convergence minimum probability score cutoff of 0.001 and local convergent minimum observed versus expected fold-change of 10 was used. A simulated resampling depth of 1,000 was used. A minimum motif length of 3 was set, and discontinuous motifs were not allowed.
We built univariate Cox proportional hazard models for the counts and proportions of each motif at static time points as well as for the dynamic changes in the motifs from baseline. We also tested for baseline differences in the counts and proportions of the motifs using an analysis of variance test and conducted post hoc comparisons with a Bonferroni adjustment to identify differences between motifs. We also compared the proportions of motif absence or presence at baseline for each patient and assessed whether each patient experienced an increase or decrease in the proportion of each motif by week 5 using individual Fisher exact tests. We calculated the overall P-value for comparing all the motifs. The significance was adjusted for the 10 comparisons by dividing the type I error of 0.05 by 10. We tested for associations between clinical characteristics and CD8+ (% Live Lymphocytes), CD4+ (% Live Lymphocytes), dendritic cells, and TCR characteristics in tumor and blood samples. Age and BMI were each fitted in a simple linear regression model. We used a Wilcoxon signed rank test to test for associations for nodal status, stage, and histology. Statistical significance was set at an α of 5% for a 2-sided P-value. All available samples were used for analyses. Analyses were conducted using RStudio 1.2.5033 Orange Blossom30.