Locked and loaded: engineering and arming oncolytic adenoviruses to enhance anti-tumor immune responses

ABSTRACT

Introduction

Oncolytic adenoviruses (Ads) are promising therapeutics to enhance anti-tumor immune responses and modulate the immune-suppressive tumor microenvironment (TME). Due to their potent ability to deliver genes in vivo, oncolytic Ads have been armed with a variety of immune stimulatory payloads to boost tumor immunotherapy.

Areas covered

We describe current knowledge about engineering oncolytic Ads to insert transgene payloads, including methods to increase its genome capacity for transgene insertion. We also review several categories of immune stimulatory payloads that have been used in oncolytic Ads to combat different barriers to effective immunotherapy.

Expert opinion

We anticipate that multi-armed oncolytic Ads alone or in combination with other types of immunotherapies will greatly improve the efficacy of oncolytic virotherapy in the future by targeting several immune barriers. However, the production and testing of multiple payload-armed Ads can be complex and time-consuming due to the limitations of current tools. Given this, we should develop new tools for rapid construction of armed Ads, improve animal models or new systems to compare the efficacy of multiple payloads, and carefully monitor the immunological toxicity induced by payloads or by systemic delivery. Most importantly, we should pursue payload-arming approaches with great care to ensure safety.

KEYWORDS:

• Oncolytic adenoviruses
• combination cancer immunotherapy
• immune payloads
• barriers to cancer immunotherapy
• immune checkpoints

1. Introduction

Oncolytic viruses are a novel class of self-amplifying therapeutic agents that have the natural ability to replicate in and kill cancer cells (reviewed in [1,2]). The original concept of oncolytic viruses is that their efficacy would come from killing primary and disseminated cancers by direct cell lysis. While this does occur, preclinical and clinical data suggest that much of the efficacy of virotherapy may actually arise from their ability to ‘make tumors hot’ or activate the immune system against cancer cells [3,4]. This ability to increase inflammation in tumors comes in part from viral replication of DNA, RNA, and proteins that can be recognized as pathogen-associated molecular patterns (PAMPs) to activate immune cells [3]. Direct viral lysis can also induce immunogenic cell death (ICD) through the release of cellular contents that be recognized as damage-associated molecular patterns (DAMPs) to activate immune cells [3]. Both PAMPs and DAMPs can trigger the secretion of pro-inflammatory cytokines and establish an anti-tumor immunological environment in the tumors. Direct viral lysis can also assist in exposing cancer antigens to antigen-presenting cells (APCs) like macrophages and dendritic cells (DCs) to trigger anti-tumor immunity [3].

A large number of viruses are being developed as oncolytics (reviewed in [1]). While each has its unique strengths and weaknesses, this review focuses on adenoviruses (Ads) as an exemplar for using viruses to amplify immune responses against cancer.

In this review, we will focus on (1) approaches to genetically engineering armed oncolytic Ads and (2) types of payloads to arm oncolytic Ads to overcome current barriers to effective cancer immunotherapy.

2. Adenoviruses (Ads)

Ads are non-enveloped double-stranded DNA viruses with icosahedrons that are about 90 nm in diameter and genomes of approximately 36 kilobase pairs (kbp) [5,6] (Figures 1 and 2). This relatively large genome size allows Ads to carry large payloads. Unlike certain pleiomorphic enveloped viruses, mature Ads form nearly identical particles in the nucleus of cells. This property makes them ideal to be produced at high concentrations from cell lysates, be purified on density gradients or chromatography columns, and ideal for scalable production and relatively simpler GMP production when compared to the purification of enveloped viruses [7]. The high stability of Ad viral proteins and DNA can be lyophilized and stored at relatively high temperatures for months or years so that adenoviral therapeutics can be shipped or stored without the need for low-temperature freezers [8–10]. This is important in remote areas that do not have −80°C freezers, but it is also important even in pharmacies in even well-developed countries [11].

Figure 1. Adenovirus virion. Human Ad26 EM image of the icosahedrons is shown overlaid with cartoons representing short and long fibers that may be present on different Ad serotypes.

Figure 2. Adenovirus genome and transcriptome. The E1A gene labeled yellow is an immediately early gene. Green labels are early and intermediate genes, while blue labels are late genes belonging to the major late transcript unit (MLTU). Fragments in the gray shadow show the protein products generated by alternative spliced transcripts. Scissors sign indicates the gene can partially or fully be deleted in oncolytic Ads. Brown diamonds point to the insertion sites for autonomous expression cassettes. Purple circles point to the insertion sites for exogenous splicing elements, or the viral ORFs can be replaced by a transgene. The red club sign indicated reported sites for IRES or 2A peptide insertion.

Ads are well documented as one of the most potent vectors for in vivo gene delivery of gene therapy or genetic vaccines. For example, a replication-defective human Ad5 (RD-Ad5) vector can generate higher anti-poxvirus antigen B5 antibody titers than modified vaccinia Ankara (MVA), Venezuelan equine encephalitis replicons (VEE-VRP), or even a replication-competent rVSV vector [12]. This makes Ads less attractive for gene therapy but makes Ads useful as genetic vaccines or in our case as oncolytic immunotherapies.

There are, in fact, many serotypes of Ads that infect humans, mammals, avian species, reptiles, and even fishes [13]. At the moment, there are more than 100 genetically distinct Ads that infect humans. The viruses in the human Ad virome can be up to 40% different at the DNA sequence level [13–16]. These differences translate into the ability for Ads to naturally target different receptors [14,15], to bind or avoid the binding of host factors in the blood [17–20]. The diverse pools of Ad also allow ‘serotype-switching,’ a shell game by switching Ad capsids with different serotypes to avoid neutralizing antibodies [21–23]. These features make Ads a ‘palette’ of different genetic oncolytic genetic platforms to carry immunomodulatory transgenes for cancer therapy.

3. The adenovirus genetic program

Following receptor binding, Ads enter cells by receptor-mediated endocytosis [5,24–26]. This is followed by endosome escape, trafficking to the nucleus, and delivery of viral DNA into the nucleus usually within 1 hour of virus binding to the cells [24]. Once viral DNA is delivered into the nucleus, host transcription factors bind rapidly to the promoters of immediate early genes to activate virus life cycle.

Ad gene activation and transcription are regulated in a stepwise manner. Viral genes are categorized into four groups based on the timing of expression: (1) immediate early gene: E1A, (2) early genes: E1B, E2A, E2B, E3, L1-52/55 K, (3) intermediate genes: IX, Iva2, and (4) late genes: L1-L5, or major late transcription unit (MLTU) (Figure 2). The early-late gene switch is tightly regulated by essential host factors and viral proteins expressed during each phase that activate viral gene promoters for the next stage as well as control RNA splicing site preferences [27,28].

E1A is the master regulator of the viral life cycle. E1A encodes the first viral proteins expressed after infection. Upon DNA entry into the nucleus, typically abundant and ubiquitous host transcription factors bind to the E1A enhancer and promoter to activate its expression without any input from viral proteins. E1A proteins then activate the expression of downstream early genes E1B, E2, E3, and E4 that are essential for viral replication, repression of host cell transcription and translation, and inhibition of host cell antiviral responses [29]. The transition from the intermediate phase to the late phase is dictated by the activation of the major late promoter and the expression of L1-L5 regions of the genome via the MLTU. The expression of these late proteins is mediated by L4-33 K, an RNA splicing factor whose expression is driven by the L4 promoter activated by E1A, E4-ORF3, and IVa2 [30,31]. This stepwise early to late gene expression pattern is evolutionarily developed in most large DNA viruses. It allows efficient virus production by first inhibiting host antiviral responses, amplifying viral DNA, and then producing viral structural proteins for the virion assembly [29].

In human cells with intact p300 and pRB pathways (e.g. A549 human lung cells), early genes are activated within 6 hours of infection and late genes are activated generally by 12 hours [32]. E1, E2, and E3 expression peaks early, but can persist throughout the life cycle. In contrast, E4 transcripts peak early and are largely quenched within 12 hours of infection [32,33]. DNA synthesis commences in A549 cells within 12 hours of infection and plateaus by 36 hours. Despite these early events in the viral life cycle, cells may not lose membrane integrity and die in culture until 2 to 3 days after infection.

This time course is similar when comparing human species C Ad6 and human species D Ad26 which are 34% genetically different when they infect A549 cells [32]. In other cells, like human bone marrow and B cell myeloma cells that have differential sensitivity to different Ads, some serotypes of Ads replicate their DNA over days (i.e. species C Ad5 and Ad6 and species D Ad26 and Ad48), and others fail to replicate and having their DNA degraded over that same period of time (i.e. species B Ad11 and Ad35) [34]. The observation from this report is contrary to other literature stating that the Ad life cycle completes within 24 hours. While short Ad life cycles can occur, some of these may be observed during infection of cells that are hyper-activated and that are already being driven into the cell cycle by other viral proteins in the cells like HeLa cells [35].

4. Complex transcription and splicing of mRNA during the adenovirus life cycle

Ads have evolved the ability to generate many mRNA transcripts and viral proteins from a relatively limited genome size. Ads use both strands of their double-stranded DNA genome to code proteins. They have also evolved a complicated mRNA transcription and splicing system to deploy many proteins at different stages of the life cycle off both strands of the genome. The E2B, E3, E4, and MLTU regions heavily use alternative splicing machinery to generate various viral proteins (Figure 2). With modern next-generation sequencing technologies [32,36] and long-read sequence technologies [33,37], the temporal distributions of different serotype Ad RNA transcripts are now better able to be dissected and used to help engineer Ads.

For example, there are over 900 mRNA patterns from species C Ad2 and more than 11,000 unique transcripts from species C Ad5 with different bioinformatics data processing pipelines [33,37]. In the late infection phase, L1-L5 genes and E3 genes (MLTU) are expressed under the control of the major late promoter. The viral transcripts from the MLTU are the most diverse and account for 50% of total spliced mRNA variants arising from the Ad genome [33]. Transcripts from the MLTU make up 80% of viral transcripts and are so abundant that they account for up to 40% of total RNA transcripts from cells during the late phase of the viral life cycle [37]. This detailed mRNA variant information from long-read next-generation sequencing is extremely valuable for determining payload insertion sites and designing transgene payload expression in the context of oncolytic Ads.

5. Strategies to increase genome capacity for transgenes

To arm an oncolytic Ad, foreign DNA sequences have to be inserted somewhere in the viral genome. Ad is an icosahedral virus with fixed genome packaging capacity (Figure 1). Wild-type human adenovirus type 5 (Ad5) has about 36 kbp in genome size and can package up to 105% of its normal genome size without destabilizing the viral genome [38]. This means that only about 2 kbp of foreign sequences can be added to an intact Ad genome without deleting any other DNA viral sequences. Because of this limitation, almost all recombinant Ads also have deletions somewhere else in their genomes to make more space for inserting transgenes.

5.1. E3 region deletions

Most researchers delete the E3 region of the virus to make space since this region encodes immune escape proteins that are dispensable for the replication of the virus in vitro [39,40] (Figure 2). Mastadenoviruses have genetically divergent E3 genes [39–41]. Some open reading frames (ORFs) are highly conserved in all Ads whereas other E3 ORFs may be unique to different serotypes or Ad species (Figure 3). For example, species C Ad6 and species D Ad26 are 34% genetically divergent with genomes of 35,760 and 35,152 bp, respectively [32]. Both express E3 12.5 K, 19 K, RIDa, RIDb, and 14.7k ORFs. However, Ad6 expresses 6.7 K and the adenovirus death protein (ADP) but does not express 23 K, 49 K, and 31 K. These differences in E3 ORF content translate into quite different sizes of E3 regions: 2,949 bp in Ad6 and 4,479 bp in Ad26 (Figure 3). The substantially larger size in Ad26 is compensated in part by having a shorter 8 repeat fiber protein than Ad6ʹs 18 repeat fiber. Therefore, depending on the Ad serotype, approximately 3 to 4.5 kbp of DNA capacity can be used for inserting transgenes after deleting E3 genes. By using a short-shafted fiber gene rather than a long species C flexible fiber gene, an extra 450 bp of DNA capacity can also be available (Figure 3), but the tropism and infectivity of Ads may also be changed by swapping the fibers from different serotypes.

Figure 3. Adenovirus genome, E3, and fiber regions. Cartoon comparing the sizes of Ad6 and Ad26 genome regions and ORFs.

In the E3 genes of species C Ads, the function of the E3-12.5 K ORF is still unknown, but its sequences are highly conserved between human species C Ads [42]. Deleting 12.5 K can be problematic and reduce virus potency in part because there is a putative poly adenylation sequence for MLP transcripts at the C-terminus of the ORF. Species C E3-CR1-α (6.7 K) is a membrane glycoprotein that localizes to the ER. 6.7 K has been shown to inhibit cell apoptosis by regulating Ca2⁺ homeostasis [43] and is involved in the degradation of TRAIL receptor 2 with E3-RID complex [44]. The E3-glycoprotein 19 K (gp19K) inhibits MHC-class-I and MIC molecules from being transported from the ER to the cell membrane to inhibit the presentation of viral peptides to cytotoxic T cells and detection by NK cells [45]; E3-ADP is known to enhance nuclear envelope rupture for efficient virus releases and cell death [46]. ADP over-expression accelerates cell death but does not necessarily increase the number of cells that are killed. Accelerating cell death may be useful if all efficacy is derived by direct oncolysis. In contrast, if the virus gains efficacy by expressing transgenes, accelerating death may reduce the duration of transgene expression and potency [47]. E3-RID-α/β suppresses cell apoptosis by internalizing Fas or TRAIL death receptors from the cell surface to protect the cell from death signals from immune cells [48–50]; E3-14.7 K is known to inhibit NF-κB signaling by blocking TNF receptor 1 (TNFR1) internalization and prevent TNF-induced cytotoxicity [51,52].

Previous studies have demonstrated that deletion of partial E3 (1.88 kb) gene or full E3 (2.69 kb) gene in Ad5, which deletes immune evasion ORFs, does not interfere with viral replication in vitro [38]. Up to 4.5 kbp of the E3 gene can also be deleted to make a total of 6.5 kbp capacity, but 3 kbp deletions are more common as bigger deletions generate less robust viruses [39,53]. Some studies also demonstrate that deletion of E3 genes may reduce Ad persistence in vivo [54,55].

For example, Wang et al. compared wildtype Ad5, Ad5(E3-RID-α/β + E3-14.7 K deletion), and Ad5(E3-gp19K deletion) in both immunodeficient and immunocompetent mouse tumor models [54]. They found that E3-RID-α/β + E3-14.7 K deleted Ad5 was cleared more rapidly in mice and had less anti-tumor responses. The tumors injected with Ad5(E3-RID-α/β + E3-14.7 K deletion) had elevated macrophage infiltration/low CD8⁺ T cell infiltration, and higher TNF and interferon-γ expression compared to wild-type Ad5 or E3-gp19K. While E3-gp19K deletion reduced MHC I display, this deleted virus generated similar immune profiles and potentially better tumor control in the CMT-93 immunocompetent mouse model. A caveat here is that Ad does not replicate well in mouse cells, so amplification of E3 genes via genome replication is weak in these models. In another study, Bortolanza et al. used an immunocompetent and human Ad permissive Syrian hamster tumor model to compare the persistence of wild-type Ad5 and Ad5(E3-CR1-α /gp19K deletion) [55]. They found E3-CR1-α and E3-gp19K deleted Ad5 was cleared much faster than wild-type Ad. However, the anti-tumor effect and overall survival were not reported in this study.

Hence, the deletion of E3 genes can increase insertion capacity, but some E3 gene deletions may cause faster Ad clearance in vivo. Some proteins from E3 genes may also be involved in immunogenic cell death for oncolytic efficacy by inhibiting normal cellular apoptosis and death to release DAMPs [3]. However, whether the deletion of immune evasion E3 genes is beneficial or detrimental for overall anti-tumor efficacy still requires more investigation since the removal of immune evasion viral genes could potentially enhance anti-tumor immune responses.

5.2. E1B deletion

E1B gene encodes two protein products: (1) E1B-19 K and (2) E1B-55 K. The E1B-22S mRNA is a polycistronic mRNA that encodes both E1B-19 K and E1B-55 K [56]. E1B-19 K is a potent inhibitor for cell apoptosis by mimicking anti-apoptotic Bcl-2 protein and binds to pro-apoptotic protein Bax [57]. Interestingly, E1B appears to be better able to protect cells from TNF-mediated apoptosis than E3 proteins, since E3 deletion has only mild effects on protection against TRAIL [32]. E1B-55 K inhibits p53 function by direct binding, regulates viral mRNA biogenesis, and inhibits cellular mRNA accumulation [58]. A dead cell will not support viral replication, so protecting the cell from apoptosis is a critical function of E1B proteins.

Both E1B-19 K and E1B-55 K can be deleted without overtly disturbing the viral life cycle [59]. Full deletion of E1B yields 1.8 kbp of extra payload capacity. The deletion of E1B-55 K is known to make Ad into a conditional-replicative adenovirus (CRAd) that decreases viral replications in several cell types based on three proposed mechanisms: (1) p53 deficiency-dependent, (2) late viral RNA export-dependent, and (3) Cyclin E expression-dependent [60], while the deletion of E1B-19 K is known to increase viral oncolytic property and viral release [61,62].

5.3. pIIIa cement protein gene deletion: an oncolytic Ad that does not generate progeny viruses

1.5 kbp more insertion space can also be gained by deleting the pIIIa gene that encodes a viral cement protein [63]. This generates a single-cycle adenovirus (SC-Ad) that replicates its DNA and completes the life cycle but does not generate infectious virions [63]. SC-Ad has advanced to human testing as a COVID-19 vaccine (NCT04839042). While SC-Ads do not generate progeny to spread infectious viruses to the second wave of cells, they do kill the cells that they infect [64]. Although SC-Ads do not generate infectious progeny that sets aside the spreading feature of oncolytic Ads to kill cancer cells, it trades with another feature: safety.

6. Types of transgene insertions and their locations

Five types of transgene insertion approaches and their insertion locations are summarized below (Figure 2):

6.1. Locations for transgene insertions

Theoretically, transgene cassettes can be inserted anywhere in the Ad genome. However, the complex expression, transcription, and mRNA splicing of Ads mean that not all locations will tolerate insertions. Transgenes are generally inserted in the place of E1 (only for RD-Ad vectors), in between E1A and B, in E3, between fiber and E4 at the end of both transcripts for the MLPU and E4, and occasionally in other sites (Figure 2). Farrera-Sal et al. provided an excellent discussion on different locations for transgene insertions into Ads [65].

Terry Hermistons group also provided interesting guidance on insertion sites in Ad5 by randomly inserting autonomous expression cassettes or exogenous splicing elements in the viral genome with a transposon system [66,67]. In E3-deleted Ad5, they found that the autonomous expression cassettes could be tolerated in E4 untranslated regions, between E1A and E1B, and within E1B [66]. For the exogenous splicing elements, they found different versions could be inserted in front of or into E3 ORF regions, into the polypeptide V ORF, at the ends of the MLPU or E4, between the ITR and E4 ORF, or within E4 ORFs (Figure 2) [67].

Data from long-read next-generation sequencing show that the E3 region is heavily spliced. These transcripts are derived not just from the E3 promoter, but also from the MLPU [33]. Although autonomous expression cassettes are commonly used in the E3 region in an E3-deleted Ad backbone, insertion in the E3 region can generate aberrant mRNA splicing and accidentally decrease transgene expression and viral titers [33]. Suzuki et al. also demonstrated that transgene insertion in E1 in the opposite transcriptional orientation or before E4 in the opposite transcriptional orientation gives the best viral titer and transgene expression level.

As a rule of thumb, it is better to avoid heavily spliced regions for inserting autonomous expression cassettes. It is also critical to check transgene sequences to avoid cryptic splicing donor (AGGT/AAGT), acceptor (CAG/G) sites, and exonic splicing enhancers (ESEs) within transgenes to prevent unexpected splicing. Several online software tools can help predict cis-splicing elements [68,69].

6.2. Types of transgene insertions

6.2.1. Replacing existing viral ORFs

Replacing a viral ORF with a transgene ORF utilizes the intrinsic viral expression program to express transgenes. It can make transgenes express as if they were early, intermediate, or late genes for programmed expression. It also entrains transgene expression into the viral program such that if the viral genes are not activated, the transgenes might not be expressed. This ORF substitution strategy has advantages in minimizing transgene cassette space over other approaches that require external regulatory sequence elements. As we reviewed, E1B and E3 genes are non-essential for viral replication and virion production, thus making those genes potential candidates for replacement with therapeutic transgenes. Terry Hermistons group has shown that E3-CR1-α(6.7 K)/gp19K, E3-ADP, and E3-RID-α/β/E3-14.7 K regions can be replaced by a transgene or multiple transgenes (Figure 2) [70–73]. Reporter and IFN genes also have been inserted in E3B regions [74]. Notably, the choice of exact insertion sites in the complex E3 region can affect neighboring E3 ORFs expression and the overall viral fitness [72]. This approach can save some space; however, human Ads do not replicate well in most mouse cells and other animal cells. Therefore, transgene cassettes that piggyback their expression on Ads’ transcripts will probably not be expressed well in some animal models of cancer.

6.2.2. Autonomous expression cassettes

An autonomous expression cassette is composed of an enhancer/promoter region to activate transcription, a transgene (usually a cDNA, sometimes a gene), and a polyadenylation (polyA) signal. Many times, these enhancers/promoters are promiscuous strong viral or cellular elements that will activate in nearly any cell (e.g. from cytomegaloviruses (CMV), Rous sarcoma virus (RSV), EF-1a, etc.). In other cases, these may be cell-specific or ‘cancer-specific’ enhancers/promoters (telomerase, surviving, probasin, etc.). Natural or artificial introns may be inserted between the transcriptional start site and the start methionine to increase mRNA trafficking out of the nucleus to increase the expression [75]. If introns are used, it should be careful not to disrupt normal Ad splicing in the insertion region as described above.

Notably, using an autonomous expression cassette allows immunostimulatory transgenes to be expressed in non-permissive, immunocompetent mouse cancer models. The primary negative effect of using expression cassettes is that they take up more space in the viral genome. A second negative effect is that they may always be ‘on’ in all cells, both cancer and non-target tissues. This can decrease specificity and run the risk of increasing side effects to the oncolytic immunotherapy particularly when they are given systemically.

6.2.3. Internal ribosome entry sequences (IRESs)

In mammalian cells, ribosomes engage the 5’ cap on mRNAs and slide until they find a good start codon to start protein translation. If that ribosome hits a stop codon or falls off, it will generally not re-engage downstream on the same mRNA. Because of this, if two cDNAs are located under one enhancer/promoter, generally only the first cDNA will have robust expression.

To circumvent this problem and conserve space, internal ribosome entry sites (IRESs) from RNA viruses can be utilized. IRESs are 300–600 bp sequences derived from RNA viruses that will allow ribosomes to be recruited and initiate protein translation without 5ʹcap recognition. Placing IRES upstream of the second cDNA will allow polycistronic gene expressions [76]. IRESs have been used to co-express transgenes on the same transcripts as E1A, E2B, and fiber [77]. Head-to-head comparisons of IRESs in these locations showed that fiber-linked IRES transcripts had the strongest expression, although this expression was delayed [77]. Intermediate levels of expression were observed with E2B-linked IRES transcripts, while E1A-linked IRES transcripts have the lowest luciferase expression level [77]. Although IRES elements eliminate the need for a second set of enhancer/promoters and polyA sequences, they are still relatively long and consume precious viral genome space. In addition, typically weaker expression of the second cDNA in an IRES cassette is observed. This can vary from gene to gene, so the expression of genes in IRES cassettes has to be measured empirically.

6.2.4. Translational truncation sequences

Translational truncation sequences have been co-opted from viruses like Foot and Mouth Disease Virus (FMDV) (reviewed in [78]). These so-called ‘2A’ peptide sequences are usually 18–22 amino acids (50–70 bp). When they are inserted in-frame between two ORFs, they induce ribosomal stalling and skipping during translation to yield two separate proteins from one mRNA [79]. 2A elements have the primary advantage of drastically reducing the size of two cDNA cassettes when compared to autonomous cassettes and IRES elements since 2A elements are only 50–70 bp in length. While 2A elements can work, there will always be a 2A sequence ‘scar’ on the c-terminus of the first protein and a proline added to the second protein. 2A elements have been used to link many transgene proteins together (i.e. IL-12 alpha and beta subunits), but also to fuse transgenes to E1A, pIX, or fiber genes [80–82]. However, the fusion of 2A proteins to the pIX capsid cement protein can cause instability of capsids [80]. Fiber-linked-2A peptide can cause repression of the E1A expression [81]. Thus, the use of 2A peptides in Ads should be empirically evaluated.

6.2.5. Insertion of exogenous splicing elements

Insertion of splice acceptor (SA) sites hijack intrinsic viral RNA splicing machinery to stitch foreign cDNAs into viral transcripts to allow transgene expression. This approach is similar to replacing a viral ORF but avoids deleting what may be an important viral protein. This approach has the great advantage of minimizing insert size because SA sequences can be as small as 30–50 bp. This strategy has been reported by using SA sites derived from Ad5 pIIIa(IIIaSA) [83], Ad40 long fiber (40SA) [84], Ad41 long fiber (41SA) [85], SV40 large T antigen [86] and beta globulin gene [87]. It is critical to check and optimize the transgene sequence to avoid potential inhibitory splicing cis-elements within transgene as they can interfere with viral protein splicing and transgene expression [81]. These elements usually need to be placed after stop codons but before polyA elements. These approaches are best made at the end of transcriptional units to avoid deranging normal viral splicing.

Terry Hermistons transposon-based approach identified several potential SA-transgene insertion sites within pV ORF, between E3-14.7 K and fiber, between fiber and E4, between E4-ORF2 and E4-ORF4, within E4-ORF6/7 and between E4 and ITR [67]. Robinson et al. also reported an insertion position before or after the hexon gene (Figure 2) [88]. The most popular insertion site is between the L5 gene (fiber) and the E4 region, expanding a new splicing transgene region, the so-called ‘L6’ region [85]. Robinson et al. compared insertion sites for both within E3 and after L3 genes and showed that L3 insertion has stronger transgene expression than SA insertion in E3. Farrera-Sal et al. compared the insertion location in the L6 region and between E4 and ITR regions, as well as using two different SA sites: IIIASA and 40SA [89]. They found insertion in the L6 region with 40SA gives the highest transgene expression. It should be noted that the types of SA sites and each transgene sequence may affect virus fitness.

7. Barriers to generating anti-tumor immune responses

Tumorigenesis is a slow process during which normal cells accumulate oncogenic mutations. Immunogenic cancer cells are more likely to be recognized and cleared by the immune system, so there is a natural selection to accumulate mutations that avoid immune surveillance [90]. These mutations can manifest into generating at least four major barriers to effective immune control of cancer (Figure 4):

Figure 4. Barriers to immunotherapy. (1) Insufficient immune signaling and priming, (2) tumor cell-intrinsic immune evasion, (3) barriers to immune cell infiltration, and (4) immunosuppressive tumor microenvironment.

1. Insufficient immune cell activation against self or tumor neoantigens

2. Tumor cell-intrinsic immune evasion

3. Physical barriers to immune cell infiltration into tumor

4. Development of an immunosuppressive tumor microenvironment (TME)

In this review, we will briefly focus on selected barriers to anti-cancer immunity and how Ad oncolytics might combat these challenges. We refer readers to several excellent reviews on arming oncolytic viruses to combat tumor immune evasion [91–93] for more detailed discussions on these topics.

7.1. Insufficient immune signaling and priming

Immune tolerance can be imposed during B and T cell ontogeny by central tolerance. However, the balance between immune activation and immune tolerance is also modulated later by innate immune cells, antigen-presenting cells (APCs), and other types of immune cells in the peripheral [94]. Innate immune cells can be activated by receptors of PAMPs and DAMPs to recognize pathogens [3]. These pattern recognition systems are good at detecting strange external invaders but less useful in detecting neoplastic cells as they lack these patterns. In the absence of innate immune triggers, long-term exposure of tumor antigens to T cells can gradually convert T cells into an anergic or dysfunctional state (Figure 4) [90]. Thus, slow growth in the absence of innate cell danger signals can allow tumor cells to evade T cell recognition. In other cases, inflammation itself can assist in the tumorigenesis [95], but this is a different discussion.

7.2. Tumor cell-intrinsic immune evasion

Immune pressure can drive cancer cells to be selected to evade cytotoxic T cells and other immune effects. For example, tumor cells that down-regulate MHC class I molecules (MHC-I) on their surface avoid being killed by cytotoxic CD8 + T cells [96]. While down-regulation of MHC-I would normally set up the cancer cell to be recognized by natural killer (NK) cells, tumor cells that can also up-regulate non-classical MHC class I molecules can also be selected to evade NK cell-mediated killing [96].

Cancer cells can also evade immune surveillance by constantly mutating their genomes as they grow. There may be an assumption that a higher mutation burden gives higher neo-antigen loads to be recognized by the immune system. However, these mutations can also increase tumor heterogeneity to generate escape clones and dilute potent tumor antigens to immunologically ‘cool down’ the tumor [97]. Cancer cells can also be selected for their ability to up-regulate immune checkpoint molecules like CD47, CD200, PD-L1, PD-L2, CTLA-4, LAG-3, TIGIT, and TIM-3 that inhibit dendritic cells (DCs), T cells, and natural killer (NK) cells [98]. Interestingly, overexpression of surface glycans by tumor cells, like sialic acids, can also be selected to suppress NK and DC cell functions by binding to inhibitory receptors Siglec-7 and Siglec-9, respectively [98,99].

7.3. Barrier for immune cell infiltration

The aberrant vasculature in some solid tumors causes high tumor interstitial fluid pressure (IFP) that can compress lymphatics and vessels reducing the ability of immune cells to enter tumors [100,101]. Tumor cells can also recruit non-cancerous stroma cells to form solid physical barriers that prevent immune cells from penetrating beyond blood vessels and lymphatics [102]. Cancer cells recruit cancer-associated fibroblasts (CAFs). These are a particular problem as they secrete extracellular matrix (ECM) materials, including collagen and fibronectin to form a tight extracellular physical barrier that prevents immune cells from migrating within solid tumors [102]. The added ECM also induces mechanical stress in the tumor to induce the production of transforming growth factor-beta (TGF-β) from these stroma cells. TGF-β secretion further increases collagen production by CAFs and inhibits T cell infiltration [101]. Beyond these physical barriers to immune cell infiltration, tumors frequently lack the expression of chemokines and proper adhesion molecules for T cells to home to the tumor and infiltrate into its core [103].

7.4. Immunosuppressive tumor microenvironment (TME)

TME formation is mediated by communications between tumor cells, stroma cells (endothelial cells and fibroblasts), and immune cells [104]. As mentioned above, CAFs can secrete ECM materials to confine the tumor space physically. In this confined space, uncontrolled growth of tumor cells not only depletes nutrients and oxygen that are essential for survival but also produces floods of toxic metabolites that suppress the ability of immune cells to activate and indeed even survive [105,106]. These metabolic signals affect immune cells in TME and can transform them into immune suppressor cells like regulatory T cells (Tregs), M2 tumor‐associated macrophages (TAM), and myeloid-derived suppressor cells (MDSCs) [98,104]. These immune suppressive cells can secrete immune suppressive cytokines, like TGF-β and IL-10, and express immune inhibitory ligands to inhibit new immune responses or to render infiltrating T cells or NK cells dysfunctional [98,104].

8. Arming oncolytic Ads to break down barriers to the immune system

Oncolytic Ads themselves can be potent immune adjuvants to make tumors immunologically hot to allow immune checkpoint inhibitors to be effective [3]. While Ads intrinsic immune-stimulatory functions are useful, their utility can be improved by arming them with immune-modulating payloads. A survey of the literature for immune payloads in Ads gives an overwhelming array of choices that are beyond the scope of this review to capture. Instead, we will touch upon a subset that can act as examples. We summarize several published immune modulating payloads expressed by oncolytic Ads for either single-armed oncolytic Ads (Table 1) or multi-armed oncolytic Ads (Table 2) for targeting each barrier of immunotherapy (Figure 5).

Figure 5. Strategic targeting by arming oncolytic Ads with immune modulation payloads. (1) Enhancing T cell priming, polarization, and memory T cell generation. (2) Inhibiting cancer immune evasion by cytokines and blocking immune inhibitory ligands. (3) Disrupting physical barriers and increasing immune cell infiltration. (4) Inhibiting immune suppressive cells.

Table 1. Payloads used in single-armed oncolytic adenoviruses.

Transgene	Targets	Function	Category	References
CD40L	DCs	Promote DC maturation	Activation	[117–120]
4–1BBL	T cells	Co-stimulation for T cell activation	Activation	[136–138]
OX40L	T cells	Co-stimulation for T cell activation	Activation	[139,140]
GITRL	T cells	Co-stimulation for T cell activation	Activation	[141]
SLAMF7-Fc/Eat-2	DCs	Promote DC maturation	Activation	[107,108]
LIGHT	T cells	Co-stimulation for T cell activation	Activation	[142]
IL-2	T cells	T cell activation	Activation	[146,147]
IL-7	T cells	Promote long-term survival of tumor-infiltrating T memory cells	Activation	[148]
IL-12	T cells	Polarize Th1 phenotype	Activation	[149–153]
IL-15	T cell and NK cells	T cell and NK cell activation, promote memory T cell generation	Activation	[154,155]
HP-NAP	DCs and T cells	Promote DC maturation and T cell activation	Activation	[109,134]
TNF-α	DCs and T cells	Induce immunogenic cell death and secretion of cytokines and chemokines	Activation/promoting infiltration/modulation of TME	[121]
CCL5 (RANTES)	Broad immune cell types	Increase immune cell infiltration in the tumor	Promoting infiltration	[181]
IFN-α	Broad cell types	Promote DC maturation, activate NK cells, upregulate MHC on tumors, modulate TME	Activation /modulation of TME	[122,123]
GM-CSF	Broad cell types	Promote DC maturation, activate NK cells, modulate TME	Activation /modulation of TME	[110,124–132]
Decorin	ECM	Disrupt tumor physical structure	Promoting infiltration	[167–169]
Hyaluronidase	ECM	Disrupt tumor physical structure	Promoting infiltration	[174–178]
Junction Opener	ECM	Disrupt tumor physical structure	Promoting infiltration	[179]
MMP-8	ECM	Disrupt tumor physical structure	Promoting infiltration	[172]
Relaxin	ECM	Disrupt tumor physical structure	Promoting infiltration	[170]
DNAse I	extracellular DNA	Disrupt tumor physical structure	Promote oncolytic Ad spread	[180]
anti-CTLA4	T cells	Block immune checkpoint CTLA-4 for T cell activation	Blocking inhibitory signals	[126]
Anti-PD-1/PD-L1 molecules	T cells	Block immune checkpoint PD-1 for T cell activation	Blocking inhibitory signals	[150,157–163]
TGF-β trap	TME	Block immune suppressive cytokine TGF-β	Blocking inhibitory signals	[185,186]
SIRPalpha-Fc/anti-CD47	CD47	Block immune checkpoint for DCs and macrophage activation	Blocking inhibitory signals	[164,165]
sCD200R1-Ig	Macrophages	Promote M2 to M1 polarization in macrophages	Blocking inhibitory signals	[191]
FAP-BiTE	CAFs	Targeting stroma cell CAFs to modulate TME	Blocking inhibitory signals	[189]
BiTEs/TriTEs	Macrophage	Kill M2 macrophage to modulate TME	Blocking inhibitory signals	[190]

Table 2. Payloads used in multi-armed oncolytic adenoviruses.

Transgene	Targets	Function	Category	References
4–1BBL-PD-1 fusion protein	T cells	Co-stimulation for T cell activation (4–1BBL) Block immune checkpoint PD-1 for T cell activation (PD-1-Fc)	Activation/Blocking inhibitory signals	[200]
CD40L + 4–1BBL	DCs and T cells	Promote DC maturation (CD40L) Co-stimulation for T cell activation (4–1BBL)	Activation	[195,196]
CD40L + ICOSL	DCs and T cells	Promote DC maturation (CD40L) Co-stimulation for T cell activation (ICOSL)	Activation	[143]
CD40L + OX40L	DCs and T cells	Promote DC maturation (CD40L) Co-stimulation for T cell activation (OX40L)	Activation	[197]
GM-CSF + decorin	TME	Promote DC maturation, activate NK cells, modulate TME (GM-CSF) Disrupt tumor physical structure (decorin)	Activation/Modulation of TME/Promote infiltration	[187]
GM-CSF + shRNA TGF-beta2	TME	Promote DC maturation, activate NK cells, modulate TME (GM-CSF) Block immune suppressive cytokine TGF-β (shRNA TGF-beta2)	Activation/Modulation of TME/Blocking inhibitory signals	[188]
IL12 + anti-PD-L1	T cells	Polarize Th1 phenotype (IL-12) Block immune checkpoint PD-1 for T cell activation (anti-PD-L1)	Activation/Blocking inhibitory signals	[201]
IL-2 + TNF-α	T cells and TME	T cell activation (IL-2) Cytokine and chemokine secretion (TNF-α)	Activation/Modulation of TME	[205–207]
CD40L + IL-6 blockage	DCs and TME	Promote DC maturation (CD40L) Block immune suppressive cytokine TGF-β (IL-6 blockage)	Activation/Blocking inhibitory signals/ Modulation of TME	[202]
IL-12 + GM-CSF	T cells and TME	Polarize Th1 phenotype (IL-12) Promote DC maturation, activate NK cells, modulate TME (GM-CSF)	Activation/Modulation of TME	[208]
IL-12 + GM-CSF + relaxin	T cells and TME	Polarize Th1 phenotype (IL-12) Promote DC maturation, activate NK cells, modulate TME (GM-CSF) Disrupt tumor physical structure (relaxin)	Activation/Modulation of TME/Promote infiltration	[171]
IL-12 + IL-18	T cells and NK cells	Polarize Th1 phenotype (IL-12 + IL18)	Activation	[156]
IL-12 + IL-23	T cells and NK cells	Polarize Th1 phenotype (IL-12 + IL23)	Activation	[198]
IL-12 + shVEGF	T cells, NK cells and TME	Polarize Th1 phenotype (IL-12) Anti-angiogenesis (shVEGF)	Activation/ Modulation of TME	[183]
IL-12 + decorin	T, NK cells and TME	Polarize Th1 phenotype (IL-12) Disrupt tumor physical structure (decorin)	Activation/Promote infiltration	[203]
IL-12 + IFN-γ + CCL5	T cells, NK cells and TME	Polarize Th1 phenotype (IL-12) Activation and modulation of TME (IFN-γ) Increase immune cell infiltration in the tumor (CCL5)	Activation/Promote infiltration	[133]
IL-12 + IL-15	T and NK cells	Th1 cell and NK cell activation (IL-12 + IL-15)	Activation	[199]
IL-15 + CCL5	Broad immune cell types	T cell and NK cell activation (IL-15) Increase immune cell infiltration in the tumor (CCL5)	Activation/Promote infiltration	[204]

Cytokines play key immunoregulatory roles. However, not surprisingly, some cytokines can provoke either anti-tumor or pro-tumor immune responses, depending on the doses and existing cytokines in the environment. For example, granulocyte-macrophage colony-stimulating factor (GM-CSF) is one of the most frequently used cytokine transgenes for arming oncolytic viruses. Indeed, GM-CSF is the payload in T-VEC, the first FDA-approved oncolytic virus in the U.S. [111]. GM-CSF stimulates the differentiation and maturation of APCs and DCs and therefore could be useful in improving immune responses against tumors. GM-CSF has been shown to enhance anti-tumor effects [112]. However, GM-CSF can also recruit immunosuppressive TAMs into tumors. Depending on the balance of other immune responses, GM-CSF can control the tumor or make it worse [112].

On the other hand, overexpression of a potent pro-inflammatory cytokine can also lead to the upregulation of immune inhibitory ligands and immune cell exhaustion as it may induce intrinsic negative feedback loops to prevent autoimmunity. For example, a high concentration of IFNγ expression may induce overexpression of PD-L1 on tumor cells to suppress anti-tumor immune responses [113]. It may be possible to balance these effects if an oncolytic virus can be carefully crafted to express multiple immune-modulating payloads to activate anti-tumor immune responses while also combating negative feedback immune-suppressive responses.

It should be noted that secreted cytokines can spread throughout the body. Therefore, they may risk broader side effects than immunostimulatory proteins that are embedded in the cell membrane. Conversely, an embedded protein trapped only in tumor cells may never be able to reach out and touch immune cells in tumor-draining lymph nodes to amplify their activity [114].

8.1. Enhancing T cell priming, polarization, and memory T cell generation

Oncolytic Ad itself can trigger strong innate immune activation. However, promoting DC maturation can more specifically activate T cells against tumor antigens with low immunogenicity [115]. DC maturation can be enhanced through ligation of CD40 ligand (CD40L) with CD40 that is expressed on DCs [116]. Alternately, DCs can be activated by stimulation with cytokines like TNF-α, IFN-α, IFN-γ, and GM-CSF (with the caveats noted earlier). Several oncolytic Ads have been armed with these factors that potentially target DCs. These have been reported to enhance anti-tumor immune responses by promoting Th1 cytokine secretions and expansion of anti-tumor T cells [117–133]. Interestingly, an Ad expressing the virulence factor neutrophil-activating protein from Helicobacter pylori (HP-NAP) can also enhance DC maturation and amplify priming of anti-cancer T cell responses [134].

T cell activation is another target to enhance anti-tumor responses. It has been shown that immune co-stimulatory factors 4–1BBL, OX40L, GITRL, LIGHT, and ICOSL can provide potent activation of strong T cell responses [135]. Several Ads have been armed with these and other co-stimulatory factors to enhance anti-tumor T cell responses [136–143].

Cytokine stimulation for T cell activation is another important target for armed oncolytic Ads. Different types of cytokines dictate the expansion lifetime of T cells, and cytokines also determine the differentiation of T cells into either pro-tumor or anti-tumor responses. IL-2, IL-15, and IL-7 allow T cell expansion and memory T cell generation [144]. IL-12, IL-18, and IFN-γ polarize T helper cells into the Th1 anti-tumor phenotype [145]. Those immune payloads have been shown to have positive effects on anti-tumor responses for oncolytic Ad therapy in animal models or patients [146–156].

8.2. Inhibiting cancer immune evasion

IFN-γ stimulation can induce MHC-I molecule expression on tumor cells to be recognized by CD8⁺ T cells [96]. For targeting immune checkpoints, adenoviral expression of anti-CTLA-4 [126], anti-PD-1/anti-PD-L1 or PD-1 decoy [150,157–163], and anti-CD47 or SIRPa decoy [164,165] has been shown to enhance anti-tumor immune responses. Anti-CTLA-4 and anti-PD1/PD-L1/PD-1 decoy target inhibitory signal to T cell responses, while anti-CD47/SIRPa decoy blocks ‘don’t eat me’ signal from tumor cells to macrophage or DCs [166].

8.3. Disrupting physical barriers and increasing immune cell infiltration

Several components of ECM have to be broken down to enhance immune cell infiltration into tumors: collagen matrix, hyaluronan, tight junction, and extracellular DNA.

Decorin is a small proteoglycan that can interact with collagen to interrupt fibrogenesis [167]. Oncolytic Ads expressing decorin demonstrated the ability to remodel tumor ECM and inhibit angiogenesis and tumor metastasis [167–169]. Another payload, relaxin, is a peptide hormone that can degrade collagen matrix by up-regulating matrix metalloproteinases (MMPs) [170]. Ad-relaxin has been shown to remodel tumor ECM to allow efficient viral spread, immune cell infiltration, and potentiate immune checkpoint inhibitor treatment [170,171]. Ad-MMP8 also has been used to break down ECM [172].

Hyaluronan is an abundant glycosaminoglycan that supports the structure of ECM [173]. Oncolytic Ad-hyaluronidase was shown to degrade hyaluronan in tumor ECM, promote viral spread and immune cell infiltration, and increase anti-tumor efficacy [174–178].

Tight junctions of epithelial cells also serve as a barrier to effective infiltration of immune cells and viral infection [179]. Yumul et al. demonstrated that oncolytic Ad5 expressing junction opener derived from Ad3 significantly improve anti-tumor efficacy compared to control Ad5 [179].

Extracellular DNA in the tumor is released from the dead cells which can be physical barriers to the infiltration of immune cells. Tedcastle et al. compared CRAd encoding DNase I or hyaluronidase. They found that CRAd-DNase has better viral spread in the tumor compared to the empty CRAd control [180].

Last, to actively increase immune cell infiltration, Ad expressing chemokines, such as CCL5, can be used to increase the T cell infiltration [181].

It should be noted that one step toward the development of metastatic cancer is gaining the ability to break down ECM to spread. Along these lines, relaxin has been reported to have both pro-tumor and anti-tumor effects because MMPs can facilitate angiogenesis and tumor metastasis [182]. To counter the effect of angiogenesis from remodeling ECM, Ad-encoded shVEGF can be used to down-regulated VEGF expression [183].

Great care should be used in arming replication-competent viruses with potent proteases and other enzymes. If such viruses were not stringently restricted to cancer cells, they could produce considerable damage to other tissues as they could begin degrading structures that are used naturally to hold tissues together.

8.4. Inhibiting immune suppressive cells

Oncolytic Ads can also be used to combat immune suppressive cells or cytokines. TGF-β is a major suppressive cytokine in the TME to induce pro-tumor immune cells (MDSCs, Tregs, M2 TAMs, and CAFs). TGF-β traps are recombinant TGF-β binding proteins that can neutralize secreted TGF-β [184]. Ads expressing TGF-β trap, decorin, or shRNATGF-β have been used to neutralize or inhibit TGF-β expression in TME [185–188]. Ad encoding Bi-specific T-cell engager (BiTE) to deplete CAFs and M2 TAMs has been demonstrated to change tumor suppressive TME [189,190], while sCD200R1-IgG is an immune checkpoint inhibitor to block CD200 inhibitory function of CD200R1 on macrophages [191].

Oncolytic viruses can also be used to combat metabolic pathways that help remodel the TME [192]. Although we did not find a specific payload-armed oncolytic Ad that targets metabolic suppression in TME, Rivadeneira et al. have demonstrated that an oncolytic vaccinia virus expressing leptin, a metabolic hormone, can enhance the oxidative activities of tumor-infiltrating T cells in TME and thus retain the anti-tumor activity of T cells in the metabolic insufficient environments [193]. Ads encoding leptin have been evaluated in the context of gene therapy against leptin deficiency-induced obesity [194], but oncolytic Ads armed with leptin have not been studied.

9. Multiple-armed oncolytic Ads for targeting several immunotherapy barriers

Arming oncolytic Ads with one immune-stimulatory payload to activate immune cells can greatly boost anti-tumor immune responses. However, targeting only one set of immune receptors/signaling pathways may induce negative feedbacks that squelch robust immunotherapy [113]. Given this, several groups have begun arming oncolytic Ads with multiple immune stimulatory payloads (Table 2).

9.1. The categories of immune payloads in multiple-armed onclytic Ads

By a literature survey (Table 2), the most common arming approach is always started with one payload from the ‘immune activation’ category and then in combination with payloads from either another factor of the ‘immune activation’ category that targets a different signaling pathway [143,156,195–199], ‘blocking inhibitory signals’ [188,200–202], ‘promoting cell infiltration’ [133,171,187,203,204] or ‘modulation of TME’ [171,183,187,188,202,205–208]. Although some multiple-armed oncolytic Ads are only armed with the payloads from the ‘immune activation’ category, the downstream effects of the activation of immune cells can actually increase the overall secretion of anti-tumor cytokines and chemokines and thus boost immune cell infiltration and change TME.

9.2. Approaches to increase transgene capacity for multiple-armed oncolytic adenoviruses

There is generally enough space in the Ad genome to allow one to express two transgenes in an E1-intact Ad. However, there is a limit to how many cDNAs can be squeezed into the fixed genome capacity of Ad vectors and still keep them replication competent. The construction of multi-armed oncolytic Ads can be challenging. Below are a few novel strategies to overcome Ad capacity limits.

9.2.1. Co-infection of replication-competent oncolytic Ad with helper-dependent adenovirus (HD-Ad) carrying multiple payloads

Helper-dependent Ads have all viral ORFs deleted [194,209,210]. This makes HD-Ads less likely to provoke anti-Ad T cells allowing them to be used for gene therapy lasting at least 7 years in non-human primates [211]. HD-Ads have also been used by many groups as gene-based vaccine platforms [212–215] demonstrating their ability to not only hide themselves from the immune system but to also stimulate the immune system. While HD-Ads can be stealthy, ablation of all their genes means that they themselves cannot be oncolytic viruses. While this is true, the genomes of HD-Ads (and RD-Ads) can replicate in cells if they are partnered with another replicating helper Ad. For example, Masataka Suzukis group uses HD-Ad to carry multiple immunomodulatory molecules for targeting different aspects of immune pathways: chemokine, cytokine, immune checkpoint inhibitors, and T cell or NK cell engager molecules [216]. When they co-infect HD-Ad with oncolytic Ad, HD-Ad can replicate and amplify transgenes with the viral protein from oncolytic Ads [216].

9.2.2. Using microRNA (miRNA) or short hairpin RNA (shRNA) expression cassettes instead of protein-encoded transgenes to target inhibitory ligands

The application of RNA interference (RNAi) approaches for arming oncolytic Ads for immune modulation is still underdeveloped. miRNA or shRNA can also be either delivered directly or from infected cells to other non-infected cells through extracellular vesicles [217]. miRNA or shRNA expression cassette also minimizes the insert size (100–200 bp) compared to a few thousand base pairs of genes encoding protein antagonists. For example, the shRNA TGF-beta2 expression cassette carried by oncolytic Ad is much smaller than the 1.2kb TGF-β trap protein [188], though their functional efficacy was not directly compared.

9.2.3. Fusion of two immune payloads to reduce transgene size

Another approach to minimize transgene size is to fuse two functional transgenes into a smaller, bifunctional gene. For example, there are many immune checkpoints decoy and co-stimulatory gene fusion proteins, such as PD-1-4-1BBL, PD1-Fc-OX40L, SIRPa-Fc-CD40L fusion, or bispecific antibody that stimulates multiple co-stimulatory factors [200,218–220].

10. Expert opinion

Oncolytic Ads armed with immune stimulatory payloads are advancing the field of immunotherapy by overcoming immune suppressive TME with the combination of immune checkpoint inhibitors or CAR-T therapy [143,201,205,221–226]. Several promising oncolytic Ad candidates are currently evaluated in clinical trials (reviewed in [227,228]). CG0070, a human serotype 5 oncolytic Ad armed with GM-CSF, has been advanced to Phase III clinical trials for treating bladder cancer (anonymized). We expect more and more interesting combinations of immune modulation payloads to appear in the future along with the advances in cancer immunotherapy. We will see more strategic targeting of oncolytic Ad on multiple immunotherapy barriers: immune signaling priming and activation, tumor cell immune evasion mechanisms, immune cell infiltration barriers, and TME. Until now, the field still has great potential since the payload choices and their combination can be unlimited.

However, it could be technically challenging to evaluate the syngeneic effects of multiple payloads when more and more payloads (variables) are used in oncolytic Ads. It is because those immunological effects have to be tested in immunocompetent animal models. For example, to properly control the effect of each payload and their interactions, for 2 payloads armed Ad, it takes 5 groups of animals to compare (no treatment, Ad vector only, payload A, payload B, and payload A + B). For 4 payloads, the total number of animal groups is 17. Moreover, although local administration of oncolytic Ad and engineered transgene greatly reduce the systemic toxicity of payloads, such as IL-12 [151,153], with more and more immune stimulatory factors arming, we would expect a higher degree of toxicity caused by enhanced inflammation in the tumors. Therefore, an effective screening approach or model system would be required to properly evaluate both the syngeneic effect of payloads and immunological toxicities.

Not surprisingly, the choice of animal model can also affect the efficacy prediction of oncolytic Ad for human patients. The most available syngeneic tumor models are murine tumor models, but mouse models are usually not permissive for Ad [229]. Thus, immunodeficient mouse xenograft models are usually used to evaluate oncolytic properties, while immunocompetent mouse tumor models are used to measure immune responses. That creates a problem that we cannot measure the interaction of oncolysis and immune stimulation at the same time in mouse models. Now, there are other immunocompetent animal models, such as the Syrian hamster tumor models that have some degree of permissiveness for Ad [230] or humanized mouse models for oncolytic Ad evaluation [231]. There were also several attempts to identify restriction factors for Ads in murine cells [229,232,233]. Young et al. investigated and excluded the viral infection step, gene transcriptions, mRNA splicing, and viral replication as the restricted points for Ad to complete their viral life cycle. Instead, they identified that significant reduction of late viral mRNA translation in murine cell lines compared to human cell lines. And the expression of Ad L4-100 K protein in trans can partially restore the Ad late protein translation, although not fully recovered when compared to the level in human cell lines [229]. If an Ad fully permissive, immunocompetent mouse model can be developed, it will greatly advance the field with better prediction of oncolytic Ad efficacy and with more available immunological tools on hand.

To discover novel immune payloads, the current method to construct recombinant Ads is also not practical for high-throughput screening. The traditional Ad cloning methods require a few months of work, including tedious recombination steps of plasmid in bacteria, a long time rescuing period, and a long viral amplification period [234]. Gibson DNA assembly technology offers a faster cloning process for Ad DNA backbone in bacteria [235,236], but 1–2 week rescuing period after transfection of viral DNA is still needed probably because of the lack of terminal protein on linearized Ad backbone [237]. A faster and simple Ad construction method for high-throughput screening will allow us to test variants of payloads quickly and facilitate new drug development.

Although not covered in this review, in order to systemically deliver oncolytic Ads into the majority of metastatic tumors in the body, there are several approaches for modifying Ad surface to reduce innate immune responses and to evade neutralizing antibodies [238–240], as well as de-targeting and re-targeting approaches to enhance tumor-specificity [241,242]. We expect to see this angle of research further advancing the field of oncolytic adenoviral therapy.

Finally, it is important that these concepts be pursued in a safe and responsible manner. Arming replicating viruses with potent immunostimulatory proteins or ECM-degrading enzymes can help patients. However, it is very important to ensure that these viruses are tightly restricted to cancer cells and are leashed to prevent them from spreading to off-target tissues or even other people during the treatment. DNA-based oncolytics have extra control features that are unavailable to RNA viruses, including controlling transcriptional activation by promoter selections and the ability of the virus to push the cell into the S phase. Ads’ DNA polymerases also have higher fidelity than RNA polymerases, making them less likely to mutate the safety control features. Safety switches that are unique in Ads include making them conditionally-replication competent in cancer cells. This can be done by removing promiscuous viral E1 promoters or using cancer-specific promoters or by mutating early proteins like E1 protein (reviewed in [1]). These restraints can be somewhat leaky if viruses are present at high multiplicities of infection. For example, even E1-deleted RD-Ads can replicate at high MOIs [243]. However, not all cells can host Ad infection, and most cannot enter the S phase, so this leakiness may be manifested most strongly in cancer cells. A different type of safety switch for Ads is making them single-cycle adenoviruses (SC-Ad) [63].

The immune system is another safety control and containment feature. Deleting E3 immune evasion genes makes the virus much less able to hide from the immune system. It may also be possible to use Ad with high seroprevalence in patients against certain Ad serotypes as a safety feature rather than a burden as demonstrated by William Wolds labs use of anti-Ad5 immunity to limit oncolytic Ad5 [244]. In conclusion, great care is needed in arming Ad and other oncolytics to maximize efficacy while also maintaining safety.

Article highlights

• Oncolytic adenoviruses (Ads) are highly immunogenic and potent vectors to deliver transgenes.

• Adenoviral mRNA splicing is highly complex. Intrinsic viral splicing events should be considered when determining the location of transgene insertion.

• Ad E3 and E1B genes can be partially or fully deleted to increase genome capacity, with consequences that need to be carefully evaluated.

• Transgenes can be expressed by replacing existing viral ORFs, inserting autonomous expression cassettes, using internal ribosome entry sites (IRESs), translational truncation sequences (2A peptides), and/or by the insertion of exogenous splicing elements.

• At least four tumor immunotherapy barriers need to be addressed:

  1. Insufficient immune signaling and priming

  2. Tumor cell-intrinsic immune evasion mechanisms

  3. Physical barriers to immune cell infiltration

  4. Immunosuppressive tumor microenvironments (TMEs).

• Multiple immune stimulatory payloads can be loaded into oncolytic Ads to overcome each barrier.

This box summarizes key points contained in the article.

Declaration interest

M Barry is the Chief Scientific Officer for Adze Biotechnology, he does not own stock in this company. M Barry is co-inventor of single-cycle adenovirus which has been licensed to Tetherex Pharmaceuticals, and he does not own stock in this company. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This project was supported by the Congressionally Directed Medical Research Program (W81XWH-19-1-0756), the Mayo Graduate School of Biomedical Sciences, and the Walter & Lucille Rubin Fund in Infectious Diseases Honoring Michael Camilleri, M.D. at Mayo Clinic.