Clinical and Immunologic Biomarkers for Histologic Regression of High-Grade Cervical Dysplasia and Clearance of HPV16 and HPV18 after Immunotherapy

Disease caused by mucosatrophic human papillomavirus (HPV) remains common, despite preventive vaccines and screening strategies. Globally, persistent HPV infections cause one-third of infection-associated cancers (1). All squamous cancers of the cervix (SCC) are thought to arise from untreated intraepithelial disease, with particular emphasis on cervical intraepithelial neoplasia 2/3 (CIN2/3). However, while all squamous cancers of the cervix arise from untreated CIN2/3, not all CIN2/3 progress to cancer. We and others have reported rates of spontaneous regression in a relatively short, prospective observational protocol prior to planned therapeutic resection (2, 3), suggesting that in some people, endogenous immune responses can eliminate CIN2/3. Indeed, a previous stZudy of patients who spontaneously regressed, low-grade CIN1 lesions suggested higher granzyme B⁺ infiltrates into cervical tissue than those who did not regress (46), although this has not been reported for the regression of high-grade CIN2/3 disease. Patients who do not exhibit spontaneous lesion regression are indicated for ablative or surgical treatment interventions including cryotherapy, cold-knife conization, or loop electrical excision procedure (LEEP; refs. 4, 5). While effective, such interventions have a number of undesirable possible side effects including pain and cervical incompetence leading to preterm birth and low infant birthweight (6, 7). Moreover, even successful procedures of this nature may not clear infection by high-risk HPV (HR-HPV), thus leaving the patient at risk for recurrence of disease (8–10).

CIN2/3 is associated with expression of the viral early proteins E6 and E7. Expression of both E6 and E7 are functionally required to initiate and maintain transformation (11, 12). T-cell effector responses to these viral, nonself oncoproteins, which are constitutively expressed by transformed cells, are likely to play a role in mediating lesion regression. Thus, immune therapies that target advanced dysplasia or cancer for which the goal is to elicit HPV-specific effector responses frequently focus on these two antigens.

We have previously reported efficacy results from our phase IIb double blind, placebo-controlled trial of VGX-3100; a plasmid-based immunotherapy for high-grade dysplasia caused by infection with HPV16/18 (13). In that trial, patients with confirmed HPV16 and/or HPV18 infection and high-grade cervical dysplasia (CIN2/3) were randomized to receive either VGX-3100 (which encodes the E6 and E7 antigens of HPV16 and HPV18) or blinded placebo via intramuscular immunization followed by in vivo electroporation using the CELLECTRA device at weeks 0, 4, and 12 to. VGX-3100 was efficacious as defined by the primary endpoint of CIN2/3 lesion regression and the secondary endpoint of lesion regression with concomitant elimination of HPV16/18 infection (ref. 13; Supplementary Fig S1). This report focuses on identifying and understanding patterns that predict treatment success. Specifically, we explore whether or not cytologic, virologic, and immunologic assessments performed at study entry or in-study, as early as two weeks following the completion of VGX-3100 dosing (week 14) had predictive value for treatment success prior to a formal histologic efficacy assessment 22 weeks later (week 36). In addition, we also analyzed samples taken at the efficacy assessment to evaluate whether immunologic or virologic markers had correlative value with regards to histologic and/or virologic success or failure in the treatment of CIN2/3.

All patients designated as per protocol (i.e., those without protocol violations) with sufficient sample were used for analysis.

In situ hybridization was performed using formalin-fixed, paraffin-embedded tissue sections. Biotin-labeled HPV probe solutions (Dako Corporation) were applied to individual sections. These included separate type-specific probes for HPV16 and HPV18. Detection of hybridized probe was performed by tyramide-catalyzed signal amplification utilizing the Dako Genpoint Kit (Dako) as per the manufacturer's instructions. Chromogenic detection was performed with DAB/H₂O₂. Controls included tissue sections positive for the HeLa cell line for HPV18 and the SiHa cell line for HPV16. Biotin-labeled plasmid probes served as a negative control in each case. Cases with a discrete punctate reaction product specifically in tumor cell nuclei were interpreted as positive.

The lytic granule loading assay was performed as described previously (44) with the exception that CD137 was used in place of HLA-DR and CD38. Peripheral blood mononuclear cells (PBMC) were recovered from cryopreservation overnight in cell culture medium and spun, washed, and resuspended the following day. After counting, 1 × 10⁶ PBMCs were plated into a 96-well plate in R10 medium. For antigen-specific responses, cells were stimulated 5 days with a combination of 15-mer peptides overlapping by 8 amino acid residues corresponding to HPV16 E6 and E7 or HPV18 E6 and E7 that had been pooled at a concentration of 2 μg/mL, while an irrelevant peptide was used as a negative control (OVA) and concanavalin A was used as a positive control (Sigma-Aldrich). All peptides were resuspended using DMSO. No costimulatory antibodies or cytokines were added to cell cultures at any point. At the end of the 5-day incubation period, plates were spun to pellet cells and all samples were washed with PBS and subjected to staining for CD3-APCH7, CD4-PerCPCy5.5, CD14-Pacific Blue, CD-16 Pacific Blue, CD137-APC, Granulysin-FITC (BD Biosciences), CD8-BV605, Granzyme A-AF700 (BioLegend), CD-19 Pacific Blue, granzyme B-PETR (Invitrogen), and perforin-PE (Abcam). Staining for extracellular markers (CD4, CD8, CD137) occurred first, followed by permeabilization to stain for the remaining markers. CD3 was stained intracellularly to account for downregulation of the marker following cellular activation. Prepared cells were acquired using an LSRII flow cytometer equipped with BD FACSDiva software (BD Biosciences). Acquired data were analyzed using the FlowJo software version X.0.7 or later (Tree Star).

IHC was performed as described previously (44). In patients with visible residual disease, an excisional procedure was performed. In patients with no obvious residual disease, a biopsy was obtained at the site of the original lesion. All IHC assays employed a polymer/multimer–based secondary detection system. In the case of labvision assays, Vector ImmPRESS HRP appropriate for the species in which the primary antibody was raised was used for detection. For the Roche Ventana assays, Roche UltraView (a Universal kit using anti-mouse and anti-rabbit IgG/HRP conjugate cocktail) was used. For each IHC staining run, irrespective of which platform was used, positive control tissues were included that were treated with the primary antibody and a buffer negative where the primary antibody was omitted and acted as a negative control. Samples were initially allocated first for staining CD8 and Foxp3. Remaining tissue was then allocated for staining CD137, CD103, granulysin, PD-L1, and perforin. Differences in number of patients stained for each marker are due to sample availability. Whole-slide image capture was performed by Histologix (Biocity) at 20× magnification with a Hamamatsu Nanozoomer 1.0-HT digital slide scanner. Normal and dysplastic epithelium and subjacent stroma morphologic regions of interest (ROI) were digitally annotated, where present, onto each section image by the study pathologist. Quantitative image analysis of IHC staining within the annotated ROI was performed by OracleBio (Biocity) using Definiens Tissue Studio software. An analysis algorithm was developed to detect positive cellular staining across each tissue image. Within the algorithm, image colors were initially separated into respective stain components, for example, brown and blue. Cells were defined and generated on the basis of the presence of a blue (hematoxylin) stained nucleus. A threshold level based on identified positive and negative staining in control tissues was then applied to the positive brown colour intensity parameter within each cell, above which a cell was defined as positive. The number or area of positive and negative stained cells was then quantified within specific regions of interest across each tissue image.

All analyses are post hoc. Two-tailed Mann–Whitney U tests are used to compare data between groups and two-tailed Wilcoxon signed rank tests to compare data within a group, as indicated. Statistical analyses were carried out using Prism program 6.0 (GraphPad Software Inc.) and SPSS Stats 22 with the exact P value and bootstrapping modules (IBM Corp. Released 2013; IBM SPSS Statistics for Windows, Version 22.0).

We have previously reported that treatment with VGX-3100 resulted in histologic regression of high-grade cervical dysplasia (CIN2/3) and clearance of HPV16 and/or HPV18 in a phase IIb double blind, placebo-controlled trial (13). In this analysis, we evaluated cervical tissue obtained from these patients prior to dosing with VGX-3100 to determine whether any preexisting factors were predictive of the subsequent success or failure in treating CIN2/3. In patients with CIN2/3 observed over a 15-week window prior to planned resection, the intensity of intraepithelial CD8⁺ T-cell infiltrates in lesional mucosa has been previously reported to be associated with the likelihood of subsequent histologic regression (14). In this study, in patients treated with VGX-3100, we found no statistical association between the intensity of pretreated CD8⁺ immune infiltrates and subsequent treatment success in lesional epithelium and stroma (Fig. 1A, left and right graphs, respectively). Indeed, study entry CD8⁺ tissue infiltrates did not associate with lesion regression in either VGX-3100 or placebo recipients (Supplementary Fig. S2).

We next assessed biopsies from study entry for the presence of immunosuppressive factors that might blunt immune responses generated by VGX-3100 and thus possibly influence treatment outcomes. We performed IHC staining for the Foxp3 transcription factor, which may be indicative of regulatory T cells (Treg; Fig. 1B). Prior to treatment, the highest range of cells that were Foxp3⁺ in dysplastic tissue were found in treated patients who did not achieve the primary or secondary endpoint compared with those who did, although the median Foxp3⁺ infiltration was higher in patients who did meet those endpoints (Fig. 1B, left and right graphs, respectively). Statistical analysis revealed that there was no significant difference between the groups. As the ratio of CD8 to Foxp3 has been shown to change in response to immunotherapy and may be informative for clinical prognosis in some diseases (15, 16), we undertook this approach to analysis as well. Finally, the CD8⁺:Foxp3⁺ ratio at study entry did not associate with either the primary or secondary endpoints (Fig. 1C), suggesting this assessment did not have predictive power for clinical outcome.

The immunosuppressive ligand PD-L1 is expressed by a variety of cell types including dysplastic, neoplastic, and immune cells in various disease states including cervical dysplasia. The pattern of PD-L1 staining in dysplastic tissue at study entry was restricted to dysplastic epithelium, and was predominantly cytoplasmic (Fig. 1D), consistent with other reports of PD-L1 expression profiles in cervical dysplasia (17, 18). The highest range of cells that were PD-L1⁺ in dysplastic tissue were found in treated patients who did not achieve the primary or secondary endpoint compared with those who did, although there was no significant difference between the groups (Fig. 1D, left and right graphs, respectively), suggesting that similar to the intensity of CD8⁺ or Treg cellular infiltration, the expression of PD-L1 in pretreatment cervical lesions was not predictive of the success or failure of VGX-3100 to treat high-grade cervical dysplasia. These data and those presented above suggest baseline CD8, Foxp3, and PD-L1 assessments were not predictive for treatment response to this HPV-specific immunotherapy.

In addition to analyzing preexisting factors that might influence treatment outcomes, we assessed whether longitudinal cytologic and/or virologic testing performed during the study might predict the ultimate treatment outcomes. Specifically, we assessed the time period following the completion of VGX-3100 dosing but prior to the efficacy assessment. We assessed the utility of cytology and HPV typing data collected from cervical swabs for use as predictors of treatment success or failure. The diagnosis of “No intraepithelial lesion” at week 14 was compared with any abnormal diagnosis, at that timepoint, including “high-grade squamous intraepithelial lesion (HGSIL),” “atypical squamous cells, cannot rule out HGSIL (ASC-H),” and calculated sensitivity, specificity, and positive and negative predictive values. The predictive value of detecting HPV16 and/or 18 at the same timepoint was also assessed alone or in combination with the cytology data. Figure 2A presents the results of these analyses from cytology and virology performed at week 14, which is 2 weeks following the final dose of VGX-3100. Normal cytology (NIL) alone was not a strong predictor of ultimate histologic outcome, only 68% of patients with this cytology ultimately had histologic regression by week 36 (Fig. 2A). However, in patients with persistent high-grade cytology (HSIL/ASC-H) at week 14, only 17% of lesions would ultimately regress histologically in the remaining 22 weeks. Eighty-four percent (84%) of patients whose HPV16/18 was undetectable at week 14 were likely to regress by week 36. While the sensitivity and specificity of these individual measures were not uniformly high, the combination of week 14 cytology and virology was a much stronger predictor of ultimate histologic regression results. The combined findings of normal (NIL) cytology and HPV16/18 clearance at week 14 were much more likely to predict histologic regression at week 36 than a finding of HGSIL/ASC-H and persistence of HPV16/18 (negative predictive value 94%; sensitivity 96%). Taken together, these data suggest that the combination of clinical cytologic and virologic assessments from cervical swabs at week 14 have high predictive value for lesion regression status at week 36.

While only the combination cytologic and virologic assessments had strong negative and positive predictive values for lesion regression status, the strength of these assessments applied to approximately half of the patients in the trial receiving VGX-3100 (49 of 101 evaluable patients) who exhibited two positive indicators or two negative indicators. Cytology and virology results were not particularly predictive in the other half of patients, whose results were divergent, prompting us to pursue other indicators of treatment success. We assessed peripheral blood immune responses to VGX-3100 antigens at week 14 (two weeks following the third and final dose of VGX-3100). Postimmunization signatures at week 14 were analyzed for their ability to predict histologic regression status prior to the definitive determination at week 36.

Using multiparametric flow cytometry, we first assessed whether or not quantification of the frequency of CD8⁺ T cells that were specific for any of the VGX-3100 antigens would be sufficient to discriminate between patients who met primary or secondary endpoints and those who did not. We employed CD137 staining for this purpose, a marker that has been shown by our group and others to identify CD8⁺ T cells activated by their cognate antigen (Fig. 3A; Supplementary Fig. S3; refs. 19–21). While treatment with VGX-3100 significantly increased the frequency of HPV16/18–specific CD8⁺ T cells (Fig. 3B, top), there was no significant difference in the magnitude of response at this timepoint between treated patients who would go on to meet the primary or secondary endpoints from those who did not. In contrast, we observed a significant posttreatment increase in the frequency of CD8⁺/CD137⁺ cells in patients treated with VGX-3100 irrespective of histologic regression outcome (Fig. 3B, bottom), suggesting this measure spoke to general immune activation by VGX-3100 but not an immune signature specific for treatment success.

We reasoned that further characterizing the functionality of these CD8⁺ T cells might aid in identifying an immune phenotype that discriminates between treated patients whose lesions do or do not regress or clear HPV. We assessed antigen-activated CD8⁺ T cells (CD137⁺) for markers associated with lytic function: granzyme A, granzyme B, perforin, and granulysin (Fig. 3A; refs. 22–28). Expression of these lytic markers in the CD8⁺CD137⁺ population significantly increased in both treated patients who met primary and secondary endpoints as well as in treated patients who did not, in all but one analysis (Fig. 3C, bottom left; analysis of granulysin for regression endpoint), suggesting that the inclusion of lytic proteins with respect to responses to any VGX-3100 antigen was informative for immune activation induced in the treated arm of the study, but was not informative for lesion regression (Fig. 3C).

VGX-3100 targets four HPV antigens: E6 and E7 from HPV16 and E6 and E7 from HPV18. The majority of patients included in the efficacy analyses were infected with HPV16 alone (89%), while a minority were infected with either HPV18 alone (6%), or coinfected with both HPV genotypes (5%; Supplementary Fig. S4). We reasoned that immune responses directed against some antigens may have confounded our analysis to this point due to the fact that they were being weighed against clinical benefit without being relevant to the infecting HR-HPV type of an individual patient. Therefore, to perform the most focused and relevant evaluation, we undertook additional analyses which excluded immune responses against HPV16 antigens or HPV18 antigens if the patient did not have an active infection with that HR-HPV type. Immune assessment therefore focused on responses relevant only to the HR-HPV type causing the high-grade dysplasia in each patient. Results of these analyses showed that the frequency of lytic protein expression within this infecting HPV type-specific CD8⁺CD137⁺ T-cell population were significantly increased only in treated patients who met primary or secondary endpoints (Fig. 3D). The discriminatory potential of this analysis was noted for perforin (increase in mean of 5.17% and 5.54% after immunization for patients meeting primary and secondary endpoints) and granulysin expression (increase in mean of 4.80% and 4.75% after immunization for patients meeting primary and secondary endpoints) as well as coexpression of granzyme A and perforin (increase in mean after immunization of 5.27% and 5.60% for patients meeting primary and secondary endpoints), and coexpression of granzyme B with perforin (increase in mean after immunization of 4.79% for both primary and secondary endpoints, Fig. 3D). This observation confirms that lytic potential in peripheral CD8⁺ T cells specific to the infecting HPV type of the treated subjects is predictive of treatment success with VGX-3100.

After establishing that HPV type specificity was a key component of a peripheral correlate of successful treatment with VGX-3100, we performed direct in situ HPV typing of dysplastic lesions that remained present in treated patients at the time of the efficacy assessment (week 36) to determine whether unresolved lesions were HPV16 or HPV18 positive (Fig. 4A). In this trial, 56% of patients had mixed infections at study entry. However, assessment of HPV was performed on exfoliated cell samples. We reasoned that as VGX-3100 was designed to treat high-grade dysplasia driven by HPV16 or HPV18, but not other HR-HPV types, persistent lesions that were not positive for HPV16 or HPV18 might be due to non HPV16/18 types that were also present at diagnosis and may account for a portion of treatment failures noted in the trial. Forty-two of the 54 patients treated with VGX-3100 whose lesions had not regressed had sample evaluable for in situ analysis. Of these 42 patients, approximately one quarter (10) were found to have persistent high-grade lesions that were not HPV16 or HPV18 positive by in situ hybridization (Fig. 4B), and 6 of those patients were also negative by PCR from a cervical swab (Fig. 4B). Further analysis of the PCR data of the cervical swab samples revealed that 4 of these patients were positive for other HR-HPV types (Fig. 4B). Agreement of these two independent assays and the presence of other HR-HPV types by PCR suggests that there is a subset of patients who were formally classified as VGX-3100–treated nonregressors, but may not have been true treatment failures as HPV16/18 was undetectable, suggesting that in these patients, persistent dysplasia was likely due to infection with another high-risk HPV type. However, this observation accounts for only 11% of treatment failures in the VGX-3100–treated cohort, and thus other analyses of samples taken after the efficacy assessment were performed.

While our assessment of cervical tissue in this study demonstrates that pretreatment CD8⁺ infiltrates were not predictive of histologic response (Fig. 1A), we have previously reported that posttreatment assessment of cervical tissue revealed a statistically significant increase in CD8⁺ immune infiltration in VGX-3100–treated patients whose lesions regressed histologically and also cleared virus (13). We further assessed the quality of this response by performing quantitative digital image post hoc analyses of markers associated with antigen-induced activation, and differentiation into tissue-resident memory T cells. We assessed cervical epithelial tissue obtained prior to treatment with VGX-3100 and obtained at the week 36 efficacy assessment for the presence of cells expressing CD137. Virologic and cytologic data suggest that the majority of treatment successes may have exhibited lesion regression and viral clearance as early as week 14 and thus evidence of an active immune response might be difficult to detect in cervical tissue at week 36. However, we hypothesized that some residual activated cells might still be detected within the cervical epithelium. No significant elevation in CD137⁺ infiltrates from week 0 to week 36 was observed in patients with either regression alone or concurrent lesion regression and viral clearance (Fig. 5A). This observation may indicate that enough time had indeed passed after resolution of infection and associated pathology for an active immune response to wane and tissue homeostasis to be reestablished. However, VGX-3100-treated patients with persistent lesions had significant increases in the intensity of CD137⁺ infiltrates in residual dysplastic cervical tissue at week 36 (Fig. 5A), while patients in the placebo cohort whose lesions persisted to week 36 did not. This observation suggests that VGX-3100 induced an immune response characterized by CD137 expression in dysplastic epithelium that was still detectable in lesions that had not yet regressed by the week 36 endpoint evaluation.

As a detectable immune response was no longer present in subjects whose lesions had regressed, we turned to analysis of resident memory immune responses by evaluating CD103 expression, which is a hallmark of intraepithelial CD8⁺ T resident memory (Trm) cells (29–31). We compared the intensity of CD103 infiltration in cervical epithelium taken prior to VGX-3100 dosing and at week 36. VGX-3100–treated patients whose lesions regressed and HPV16/18 was cleared showed significant increases in the frequency of intraepithelial CD103⁺ cells in cervical epithelium (Fig. 5B). While the frequency of CD103⁺ cells was also elevated in treated patients whose lesions did not regress, the magnitude of these increases did not reach significance (P = 0.055). Patients whose lesions regressed, but failed to achieve the secondary endpoint of concomitant viral clearance also had statistically significant elevation of mucosal CD103⁺ infiltrates (P = 0.039). Patients who received placebo did not have significant increases in CD103⁺ tissue infiltrates (Fig. 5B). These data suggest that the active immune response in treated patients whose lesions regressed had converted to a resident memory phenotype by week 36 and that the ongoing immune response noted in the treated nonregressors may also have been converting to a memory response as well.

The increase in CD137⁺ and CD103⁺ immune infiltration observed in the cervical tissue of VGX-3100–treated nonregressors was paradoxical, because one might predict that immune infiltration of this nature would be a hallmark of patients who would be likely to regress their lesions, not a hallmark of lesion persistence (32–34). We reasoned that if an immune response was present in the cervical epithelium of patients with persistent disease, there could be an underlying immunoregulatory or immunosuppressive mechanism at work preventing efficient effector function. Our assessment of Treg and PD-L1 expression prior to treatment with VGX-3100 had indicated that neither were predictive of treatment failure (Fig. 1A–D). However, it has been reported that an infiltrating effector immune response can result in increases in either Treg infiltrates or PD-L1 expression (35–39). Thus, we compared Foxp3 infiltration and PD-L1 expression in dysplastic lesions before and after VGX-3100 treatment (week 36). In the residual lesions of VGX-3100–treated nonregressors, Foxp3⁺ infiltrates did not increase significantly between week 0 and week 36. This observation suggested that VGX-3100 did not elicit mucosal FoxP3⁺ infiltrates (Fig. 5C) and that FoxP3⁺ infiltration is likely not to be a mechanism of treatment failure. Indeed, because of increases in CD8⁺ infiltrates, the CD8:Foxp3 ratio increased in VGX-3100–treated regressors from week 0 to week 36 (Fig. 5D).

In contrast, in VGX-3100–treated nonregressors, PD-L1 expression in persistently dysplastic lesions was significantly increased from week 0 to week 36, a finding not replicated in patients receiving placebo (Fig. 5E). PD-L1 expression may be induced by a variety cytokines, including TNFα and IFNγ (16, 37, 40). Thus, the increase in PD-L1 expression in residual lesions suggested the ongoing presence of an active infiltrating effector immune response. In patients whose lesions did regress, PD-L1 expression did not change significantly (data not shown). These data suggest that upregulation of PD-L1 in persistently dysplastic lesions of VGX-3100–treated patients may have contributed to treatment failure, although the specific impact of PD-L1 in this context is speculative as most evaluable patients displayed less than 5% staining of this immunosuppressive ligand and when viewed in the context of variations of expression of PD-L1 in patients receiving placebo (Fig. 5E).

Finally, we evaluated cervical tissue for functional lytic markers for correlation with treatment success. As the peripheral immune analysis had revealed that changes in expression of granulysin or perforin were the strongest predictors of concomitant lesion regression and clearance of HPV16 or HPV18 (Fig. 3D), these cytolytic effector molecules were targeted for further analysis in cervical tissue. We reasoned that confirmation of the presence of a similar signature in cervical tissue would confirm that the responses noted in the periphery were mechanistically relevant for VGX-3100–driven efficacy. The intensity of granulysin⁺ immune infiltration in cervical epithelium did not change significantly between week 0 and week 36 in any group (data not shown). However, perforin⁺ immune infiltrates increased significantly at week 36 compared with study entry, only in the VGX-3100–treated patients who met primary or secondary endpoints (Fig. 5E). This finding correlated directly with the immune signature noted in the peripheral blood (Fig. 3D).

Taken together, these data confirm that treatment with VGX-3100 drove immune infiltration into cervical lesions and that increased infiltration of lymphocytes with a cytotoxic signature marked specifically by perforin expression was discriminatory in favor of histologic regression of cervical HSIL and clearance of HPV16/18 infection. Expression of perforin, but not granulysin, was discriminatory, suggesting a possible specificity for the activity of perforin in the context of lesion regression that is not applicable to granulysin.

A detailed understanding of the mechanisms contributing to successful treatment of chronic disease with an immunotherapy may serve as a roadmap for continued development of the immunotherapy platform as a whole. We have previously reported on the successful clearance of HPV16/18–associated high-grade cervical dysplasia (CIN2/3) in a phase IIb double-blind, placebo-controlled trial of VGX-3100 (13). VGX-3100 not only led to the regression of CIN2/3, but also eradicated the underlying HPV16 or HPV18 infection in about 80% of the responders, thereby potentially reducing the risk of recurrent disease (41). However, although VGX-3100 induced detectable peripheral blood cellular immune responses in many patients, only about half of treated patients exhibited complete lesion regression during the study window. More in-depth analyses presented here show that cellular correlates of VGX-3100–induced regression of disease was multifactorial (Fig. 6), requiring both antigen specificity for the HPV type the patient was infected with as well as the expression of effector molecules strongly associated with lytic function. While all four lytic proteins analyzed (granzyme A, granzyme B, granulysin, and perforin) showed statistical association with lesion regression and elimination of HPV infection, the expression of perforin in particular in HPV-infecting type-specific CD8⁺CD137⁺ cells had the strongest association with successful treatment with VGX-3100. Upregulation of PD-L1 in the lesions of VGX-3100–treated patients with persistent disease was noted, although the relatively low frequency of PD-L1 expression in most patients suggests that while this may be a contributing factor, it may not represent the main driver of treatment failure. The pattern of PD-L1 staining was predominantly cytoplasmic as opposed to membranous, and while this is in keeping with other published reports of PD-L1 staining in cervical dysplasia (17, 18), it raises the question of what type of impact this localization has in immunosuppression.

The antibody responses elicited by the preventive HPV vaccines are qualitatively and quantitatively different from those elicited by natural infection in that vaccine-induced memory B cells have greater avidity and breadth of specificity than naturally occurring responses (43, 45). Generation of virus-specific T-cell responses in the clinical setting of preinvasive HPV lesions may be analogous—in natural infection, virus-specific T cells are exceedingly rare in the circulation. Natural infection is intraepithelial, and noninflammatory, so the likelihood of generating a strong systemic response is relatively low. In contrast, peripheral administration of antigens through vaccination is inherently systemic. The peripheral administration of VGX-3100 induces T cells expressing perforin that are detectable in the peripheral blood as well as in cervical epithelium whose presence appears to play a role in the elimination of preinvasive HPV disease in the cervix. In addition, cervical cytology and HPV testing specific for the two genotypes targeted by VGX-3100 (HPV16, HPV18) are two simple, commercially available tests routinely used in clinical practice that provide prognostic information (Fig. 6).

The increase in tissue infiltrates expressing CD137, CD103, or perforin subsequent to treatment with VGX-3100 provide indirect evidence of activation and differentiation mediated by recognition of cognate antigen. The success of peripheral therapeutic vaccination is predicated on the concept that peripheral immunization can drive an immune response that homes to diseased tissues, such as the cervical mucosa in the context of HR-HPV infection (42). One previous study of immune therapy for HR-HPV infection has reported immune changes in cervical mucosa following treatment, but did not assess activation status, memory phenotype, or the presence of effector molecules (42). In the earlier study, it became apparent that the endpoint was effectively being censored only three weeks after the final vaccine dose, as robust immune responses were observed in the residual dysplastic mucosa. Alternatively, in the current clinical study, where endpoints were assessed 6 months after the final dose of VGX-3100 tissue analyses reflected a shift toward homeostasis in patients whose lesions had regressed. We did not see these changes in the resection specimens in patients who had received placebo. The finding of increased CD103-expressing cells is encouraging in that the differentiation into tissue-resident memory cells could result in lower rates of recurrence in treated patients compared with women whose lesions regressed spontaneously. The HPV type–specific expression and persistence of perforin by cells in the periphery as well as cervical mucosa identify a mechanism by which these cells may mediate this effect. To our knowledge, this is the first report of immunologic findings in target tissues that are congruent with vaccine-induced changes in peripheral blood and the first time that changes in a peripheral cytolytic CD8⁺ T-cell response have been statistically linked with successful treatment of a neoplastic process by an active immunotherapy (Fig. 6). These findings will be further investigated in phase III, and if confirmed, would provide impetus to carry out more granular analyses to determine what constitutes an effective immune response, as well as identify barriers that prevent immune-mediated regression.

M. Dallas and J.J. Kim hold ownership interest (including patents) in Inovio Pharmaceuticals. S. Plotkin reports receiving speakers bureau honoraria from and is a consultant/advisory board member for Dynavax and Inovio Pharmaceuticals. D.B. Weiner is an employee of, reports receiving commercial research grants from, holds ownership interest (including patents) in, and is a consultant/advisory board member for Inovio Pharmaceuticals; reports receiving other commercial research support from GeneOne; and reports receiving speakers bureau honoraria from BMGF, Medimmune, and Merck. C.L. Trimble is a consultant/advisory board member for Inovio Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.P. Morrow, J. Yan, N.Y. Sardesai, J.J. Kim, D.B. Weiner, C.L. Trimble, M.L. Bagarazzi

Development of methodology: A.S. Khan, D. Weiner, M.L. Bagarazzi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.P. Morrow, J.D. Boyer, R. Vang, C.L. Trimble, M.L. Bagarazzi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.P. Morrow, K.A. Kraynyak, A.J. Sylvester, M. Dallas, D. Knoblock, N.Y. Sardesai, J.J. Kim, C.L. Trimble, M.L. Bagarazzi

Writing, review, and/or revision of the manuscript: M.P. Morrow, K.A. Kraynyak, A.J. Sylvester, M. Dallas, D. Knoblock, J. Yan, L. Humeau, N.Y. Sardesai, S. Plotkin, D.B. Weiner, C.L. Trimble, M.L. Bagarazzi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.J. Sylvester, K.A. Kraynyak, L. Humeau

Study supervision: A.S. Khan, L. Humeau, N.Y. Sardesai, M.L. Bagarazzi

The authors thank the patients who participated in this study and the entire HPV-003 clinical team from the participating study sites and Histologix and OracleBio for assistance with IHC staining and digital image analysis and FlowMetric for aid in flow cytometry. This work was supported by Inovio Pharmaceuticals Inc.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.