Injection site vaccinology of a recombinant vaccinia-based vector reveals diverse innate immune signatures

Poxvirus systems have been extensively used as vaccine vectors. Herein a RNA-Seq analysis of intramuscular injection sites provided detailed insights into host innate immune responses, as well as expression of vector and recombinant immunogen genes, after vaccination with a new multiplication defective, vaccinia-based vector, Sementis Copenhagen Vector. Chikungunya and Zika virus immunogen mRNA and protein expression was associated with necrosing skeletal muscle cells surrounded by mixed cellular infiltrates. The multiple adjuvant signatures at 12 hours post-vaccination were dominated by TLR3, 4 and 9, STING, MAVS, PKR and the inflammasome. Th1 cytokine signatures were dominated by IFNγ, TNF and IL1β, and chemokine signatures by CCL5 and CXCL12. Multiple signatures associated with dendritic cell stimulation were evident. By day seven, vaccine transcripts were absent, and cell death, neutrophil, macrophage and inflammation annotations had abated. No compelling arthritis signatures were identified. Such injection site vaccinology approaches should inform refinements in poxvirus-based vector design.

Introduction A range of vaccine vector systems based on vaccinia virus (VACV) and other poxviruses have been developed, with several sold as products and many more in development and in human clinical trials [1]. These include Modified Vaccinia Ankara (MVA) [2,3] [9]. A large series of recombinant MVA (rMVA) vaccines have been evaluated in non-human primate (NHP) studies [10] and in human clinical trials [1,11], with MVA-BN-Filo recently licensed in Europe as part of a heterologous prime-boost Ebola vaccine [12]. MVA is also licensed as a smallpox vaccine (sold as Imvanex/Imvamune). Recombinant poxvirus vector systems have a number of attractive features for vaccine development including a large payload capacity (at least 25,000 base pairs), potential for cold chain-independent distribution, lack of vaccine DNA integration and induction of both cellular and humoral immunity [1]. Nevertheless, a range of strategies are being sought to improve immunogenicity and reduce reactogenicity [2,[13][14][15][16][17]. Both these key characteristics of vaccines are largely dictated by the early behavior of the vaccine at the injection site. However, a comprehensive RNA-Seq approach to characterize the post-inoculation injection site responses has not been undertaken for a recombinant poxvirus-based vaccine.
The Sementis Copenhagen Vector (SCV), derived from the Copenhagen strain of VACV, was recently described [18,19]. SCV can replicate its DNA but is rendered unable to generate viral progeny in vaccine recipients by virtue of a targeted deletion of the D13L gene that encodes the essential viral assembly protein, D13. Recombinant SCV vaccines are produced in Chinese Hamster Ovary (CHO) cells modified to express D13 and the host range protein, CP77 [19]. A single construct recombinant SCV vaccine encoding the structural gene cassettes of both chikungunya virus (CHIKV) and Zika virus (ZIKV) (SCV-ZIKA/CHIK) was constructed with each polyprotein immunogen driven by the same synthetic strong early late promoter [20], but from two distinct distant loci from within the SCV genome [18]. A dual ZIKV and CHIKV vaccine was deemed attractive as these virus co-circulate in overlapping geographic regions, and can co-infect both mosquitoes and humans [21][22][23]. SCV-ZIKA/CHIK was shown to protect against CHIKV and ZIKV in a series of mouse models [18]. In NHPs the vaccine also induced neutralizing antibodies against VACV, CHIKV and ZIKV and provided protection against ZIKV viremia [22].
Systems vaccinology uses mRNA expression profiling to gain a detailed molecular understanding of the behavior of vaccines in vivo, thereby informing design and development [24]. The approach has been used to understand and predict immunogenicity [25,26], reactogenicity/safety [27,28] and adjuvant activity [29,30]. Most systems vaccinology studies have analyzed peripheral blood post vaccination, as this is readily accessible in humans. However, herein we described RNA-Seq and bioinformatics analyses of injection sites after intramuscular (i.m.) vaccination. Adult wild-type mice were vaccinated with SCV-ZIKA/CHIK and muscles were harvested at 12 hours post vaccination to characterize early injections site innate responses and identify adjuvant signatures. As vector-induced cytopathic effects (CPE) are only just beginning (at least in vitro) at 12 hours post infection [19], this was also deemed a PLOS PATHOGENS suitable time to investigate expression in vivo of both viral vector genes and expression of the recombinant immunogens. Muscle tissue was similarly analyzed on day 7 post i.m. vaccination to determine the persistence of vaccine transcripts and characterize the evolution of injection site inflammatory responses at a time when vaccine-induced adaptive immune responses are being generated. This is also the time when inflammatory lesions develop after VACV vaccination [31][32][33][34][35][36][37], with acute transient injection site reactions also the most common adverse reactions observed in clinical trials of MVA [38]. Finally, feet were harvested on day 7 post vaccination to determine whether SCV-ZIKA/CHIK vaccination might be associated with an arthropathy signature. Several MVA vaccine trials reported transient acute arthralgia as an adverse event [38][39][40], with arthropathy associated with new CHIKV vaccines remaining a standing concern for regulators [41] after the experience with a live-attenuated CHIKV vaccine [42]. The characterization provided herein of vaccine gene expression and innate host immune responses at the injection site provide both a process and insights that may inform future endeavors to improve immunogenicity whilst limiting reactogenicity of poxvirus-based vaccine vectors. RNA-Seq reads from each of the five groups aligned to the three viral genomes (the vector, SCV, and the two recombinant immunogen inserts from ZIKV and CHIKV); MQ-quadriceps muscles from mock vaccinated, SCV12hQ mice quadriceps muscles from SCV-ZIKA/CHIK vaccinated mice 12 hours post vaccination, SCVd7Q -quadriceps muscles from SCV-ZIKA/CHIK vaccinated mice taken 7 days post vaccination, MF-feet from mock vaccinated mice 7 days post vaccination, and SCVd7F -feet from SCV-ZIKA/CHIK vaccinated mice 7 days post vaccination. The number of viral reads is expressed as a percentage of the number of reads mapping to the mouse genome, with 3 biological replicates providing the SD (S1 Fig). The bars plotting to �0% had values ranging from 0 to 3.5x10 -5 %. (B) RNA-Seq reads from each of the five groups aligned to the house-keeping gene, RPL13A, also expressed as a percentage of the number of reads mapping to the mouse genome. (C) IGV visualization of reads aligned to the recombinant structural polyprotein immunogens of ZIKV (prME) and CHIKV (C-E3-E2-6K-E1), which are encoded in the SCV-ZIKA/CHIK vaccine. All reads from all replicates are shown (for details see (S1A Table). As expected, no reads mapped to the non-structural genes of ZIKV or CHIKV (NS1-5 and nsP1-4, respectively), as these are not encoded in SCV-ZIKA/CHIK. (Vertical purple lines for ZIKV indicate base call errors after a string of Gs). (Reads mapping to the SCV genome are shown in S1B Table). (D) H&E staining of injection site 12 hours post vaccination. Dotted ovals indicate muscle cells in early stages of necrosis (pink staining). (E) Top; IHC with anti-Ly6G staining for neutrophils (parallel section to D focusing on area of effects (CPE) in infected cells within a few days and that SCV is unable to produce viral progeny [1,18,19]. The data is also consistent with studies on MVA, where luciferase expression by MVA was lost 48 hours post inoculation [43]. The paucity of vaccine reads in the feet 7 days post vaccination (Fig 1A, SCVd7Q) also illustrates that the vaccine does not disseminate to and/or persist in joint tissues (a potential safety concern; see below). The percentage of reads mapping to a murine house-keeping gene, RPL13A [44], was similar for the 3 samples from quadriceps muscles, and for the two samples from feet (Fig 1B), illustrating that the low vaccine read counts for mock and day 7 samples (Fig 1A) were not due to low read counts for those samples.
A criticism of virally vectored vaccines has been that viral vector transcripts can be markedly more abundant than recombinant immunogen transcripts, resulting in immune responses excessively directed towards vector proteins rather than the recombinant immunogen(s) [45][46][47]. However, �20% of all the SCV-ZIKA/CHIK vaccine reads mapped to the two recombinant immunogen genes, even though the ZIKV and CHIKV sequences were relatively small (2067 bp and 3747 bp, respectively), when compared to the large SCV genome (�190,000 bp). This perhaps attests to the strength of the poxvirus synthetic strong early late promoter [20] used for the CHIKV and ZIKV immunogens in the SCV-ZIKA/CHIK vaccine [18].
Expression of two immunogens in a single poxvirus vector construct carries the risk that one immunogen is expressed significantly better than the other, a problem encountered in a variety of settings [1]. A comparable number of reads mapped to the recombinant CHIK and ZIKA inserts (Fig 1A), with these two inserts distantly separate from each other in the SCV genome and driven from the same promoter [1]. This approach would seem largely to ensure (at least in SCV-ZIKA/CHIK) that comparable levels of mRNA are produced for each of the two immunogens.
Read alignments to genes encoded by SCV are described and annotated in detail in S1B Table. Immune and cell-death modulating proteins are highlighted, along with annotations regarding their activity in mice and their activity in the Copenhagen strain of VACV, from which SCV was derived. Many of these genes are referred to below.

Injection site histology and immunohistochemistry at 12 hours post vaccination
H&E staining of the intramuscular injection sites showed that some skeletal muscle cells displayed fragmented pale cytoplasm with loss of striation and small condensed pyknotic nuclei indicative of necrosis ( Fig 1D,

PLOS PATHOGENS
Injection site vaccinology of a vaccinia-based vector necrotic cells were partially surrounded by mixed inflammatory cells infiltrates (high densities of purple nuclei) and some cellular debris ( Fig 1D). Immunohistochemistry (IHC) with a neutrophil-specific marker, anti-Ly6G [51,52], illustrated that the infiltrates contained abundant neutrophils ( Fig 1E, top panel, Ly6G). Interestingly, neutrophils have been shown to contribute to adjuvant activity [53]. The infiltrates also contained areas staining with ApopTag indicating apoptosis (Fig 1E, (Fig 1F). Using parallel sections, IHC with 4G2 (S3B Fig) and a control antibody is shown in S3C Fig, with no significant staining observed in the latter. The spherical/oval cytoplasmic staining patterns (Fig 1F) likely reflect the well described cytoplasmic factories wherein the poxvirus coordinates protein expression and subjugates host functions [56][57][58]. No significant staining was observed in MQ. These results illustrated that the immunogen mRNA expression seen in Fig 1A and 1C translates into protein expression in muscle cells in vivo.

Injection site host cell death signatures at 12 hours post vaccination
The mode of cell death for a host cell expressing vaccine immunogens can have important implications for immunogenicity, with necrosis often favored over apoptosis [14,59]. RNA-Seq analysis of the mouse i.m. injection sites 12 hours post vaccination (MQ vs SCV12hQ; full gene list in S2A Table) provided a set of differentially expressed genes (DEGs) (S2B Table; FDR or q <0.01, fold change >2 and sum of all counts across the six samples >6). The up-regulated DEGs (n = 1390; S2B Table) were analyzed by Cytoscape (S2C Table), with cell death terms suggesting the presence of apoptosis, necroptosis and necrosis ( Fig 1G). Skeletal muscle cells are generally resistant to apoptosis [60,61] and VACV's apoptosis inhibitor, B13R [14], was also expressed at the injection site (S1B Table). Skeletal muscle cells have recently been shown to be able to undergo necroptosis [62]. Skeletal muscle necrosis is well described [63,64] and H&E staining was consistent with muscle cell necrosis (Figs 1D and S3A). As ApopTag staining was clearly present in the aforementioned infiltrates ( Fig 1E, ApopTag), the apoptosis signatures ( Fig 1G) may largely be associated with infiltrating leukocytes such as neutrophils, which are highly prone to apoptosis [65]. MVA can induce apoptosis in vitro and in certain settings in vivo [14,66] and SCV can induce apoptosis (at least in vitro, S4 Fig); however, the mode of cell death elucidated in vitro may not be recapitulated in primary skeletal muscle cells in vivo.

Large Toll-like receptor signatures at 12 hours post vaccination
The up-regulated DEGs (for MQ vs SCV12hQ; S2B Table) analyzed as above by Cytoscape returned multiple terms associated with innate immune responses (S2C Table). To provide insights into the early innate host immune responses and potential adjuvant signatures induced by SCV-ZIKA/CHIK vaccination, the full DEG list (1608 genes) for MQ vs SCV12hQ (S2B Table) was analyzed by Ingenuity Pathway Analysis (IPA) using the Up-Stream Regulator (USR) function and the direct and indirect interaction option. The list of USRs (S2D Table) illustrated a highly significant Toll-like receptor (TLR) signature, dominated by TLR3 and 4, followed by TLR9, 7 and 2 (Fig 2A). Although other TLRs (TLR1, 5, 6, 7/8, 8) were also identified, the number of unique DEGs responsible for these annotations was low (Fig 2A, numbers in brackets), arguing that these were less reliable USRs as they arose from subsets of DEGs already used in the annotations for TLR3, 4, 9, 7 and/or 2 (S2D Table, Target molecules in dataset). Given the common signaling pathways used by all TLRs, primarily involving MyD88 and/or TICAM1/TRIF, overlap in genes induced via the different TLRs would be expected.
The z-scores and p values for TLR signatures determined herein (Fig 2A) showed a remarkable concordance with previously published TLR-knockout mouse studies (summarized in S2E Table). The higher the z-score/p value, the more important the TLR was for infection and pathology in TLR-knockout mice infected with replication competent poxviruses. Specifically, the top TLR, TLR4, is stimulated by an unknown ligand present in/on VACV particles, with TLR4 required for effective antiviral activity and protection against mortality in mice after VACV infection [67]. TLR3 stimulation is likely mediated by dsRNA derived from the abundant complementary RNA transcripts produced late in the VACV infection cycle [68]. TLR3 stimulation in VACV-infected mice promotes inflammatory cytokine production, immunopathology, and affects mortality [69]. TLR7 (which detects ssRNA) is expressed on plasmacytoid  Table) of 1608 DEGs identified in quadriceps muscles 12 hours post SCV-ZIKA/CHIK vaccination (MQ vs SCV12hQ; S2B Table). Circle diameters reflect the number of DEGs associated with each IPA USR annotation. Numbers in brackets indicate the number of unique DEGs associated with each annotation over the total number of DEGs associated with the TLR annotation; circles with colored fills contain >3 DEGs uniquely associated with the indicated TLR annotation. A46R is expressed in the cytoplasm of infected cells. (B) Cytosolic sensor signatures identified by IPA USR analysis (S2D Table). Sensors divided into 3 categories associated with dsRNA (red circles), DNA (blue circles) and inflammasome activation (black circles). Circle diameters reflect the number of DEGs associated with each annotation. VACV genes encoding cytoplasmic inhibitors are shown in blue, with the black cross indicating that the gene/protein is not present or not functional in SCV (or in the Copenhagen strain of VACV), the red cross indicating that the gene was not detected by RNA-Seq of MQ vs SCV12hQ, the green cross indicate that the activity in mice is unknown (see S1B Table). . To the best of our knowledge, there is no literature suggesting a role for TLR1, 5 or 6 in poxvirus infections, consistent with the low number of unique DEGs for these annotations (Fig 2A). The role of TLR8 in VACV infections remains controversial [79], with TLR8 non-functional in mice [80].
VACV produces an inhibitor of TRIF, MYD88, TRAM and MAL, called A46 or VIPER (encoded by A46R), a protein reported to be active in murine systems [81]. However, A46 is expressed in the cytoplasm of SCV-infected cells and not in neighboring uninfected cells that may also express TLRs. Such cells might sense TLR agonists comprising viral pathogen-associated molecular patterns (PAMPs) and/or damage-associated molecular patterns (DAMPs) released by SCV infection-induced CPE [19,82,83].

Multiple cytoplasmic sensor signatures at 12 hours post vaccination
The IPA analysis of DEGs for MQ vs SCV12hQ (S2D Table,   . Activation of the proteases Caspase 1 (gene CASP1) (canonical) and Caspase 11 (gene CASP4) (non-canonical) represent the central outcomes of inflammasome activation, with VACV stimulation of the inflammasome well described [89]. ALVAC is also reported to stimulate the inflammasome via AIM2 in both human and mouse cells [88], with a minor AIM2 signature also seen after SCV-ZIKA/CHIK (Fig 2B). Viron assembly is arrested at the viroplasma stage in SCV-infected host cells due to the deletion of D13L [19], which may limit inflammasome activation.
Poxviruses encode a number of proteins that seek to limit the activity of host immune responses (S1B Table, yellow highlighting), with some of these inhibiting the activities of cytoplasmic sensors (Fig 2B, blue text). VACV's decapping enzymes (D9 and D10, encoded by D9R and D10R) are expressed at the vaccination site (S1B Table). Both proteins inhibit dsRNA accumulation, with D10 functional in mice [90]; whether D9 is functional in mice is unknown (Fig 2B, green cross). DHX9 is involved in both DNA and RNA sensing and is targeted by VACV's E3 protein (encoded by E3L), with PKR inhibition the best defined activity of E3 [91][92][93]. VACV DNA is usually shielded from cytoplasmic sensors during replication in viral factories via wrapping in ER membranes; however, this wrapping is lost during virion assembly [94]. The DNA sensor PRKDC/DNA-PK (a DNA-dependent protein kinase) showed a z-score of zero (Fig 2B), perhaps due to the inhibitory activity of C4 (a protein encoded by C4L) [95].
The expression of PRKDC/DNA-PK mRNA was not significantly altered (S2D Table), consistent with C4 protein-protein interactions [95] inhibiting the transcriptional modulation mediated by this upstream regulator. Transcripts for another PRKDC/DNA-PK inhibitor, C16L, were not detected by RNA-Seq in SCV12hQ (Fig 2B, red cross and S1B Table). CrmA (from cowpox) and VACV's homologue, B13R/SPI-2, inhibit caspase 1 (and other caspases), but are not functional in the Copenhagen strain of VACV [96] (Fig 2B, black cross). NLRP1 was not identified by the IPA USR analysis, potentially due to the expression of F1L (S1B Table) [97].

Interferon response factor signatures at 12 hours post vaccination
Interferon response factors (IRFs) are key transcription factors triggered by PAMPs, with IRF3, IRF7 and IRF1 dominating ( Fig 3B) in the IPA USR analysis (S2D Table, direct and indirect) of the DEGs from MQ vs SCV12hQ (S2B Table). These 3 IRFs are also in the top 5 USRs (sorted by p value) when the "direct" only option was used for the IPA USR analysis (S2D Table, [96]. IRF1 has a role in positive feedback maintenance of ISG expression [116,122] and has been shown to promote adaptive immunity in certain settings [123,124]. Other IRF USRs included IRF2, 4, 5, 6, 8 and 9 ( Fig 3B). IRF8 has a critical role in development and maturation of myeloid cells such as dendritic cells [125] and IRF5 is predominantly expressed by myeloid cells and regulates inflammatory responses, generally downstream of TLR-MyD88 pathways [126]. IRF4 has a negative z-score ( Fig 3B) and is a negative regulator of TLR signaling [127].

Canonical NF-κB family signatures at 12 hours post vaccination
The NF-κB family of transcription factors play key roles in immunity, with the IPA USR analysis (S2D Table,

PLOS PATHOGENS
Injection site vaccinology of a vaccinia-based vector injection site (Fig 3C, green circles), consistent with the TLR signaling USRs described above. The dominance of the NFKBIA, but not another NF-κB inhibitor NFkBIB, may reflect the activities of the K1L and A49R genes in vaccine-infected cells. Both K1 and A49 proteins prevent degradation of NFKBIA [96], with A49 binding the ubiquitin ligase B-TrCP [128]. B14 (encoded by B15R in the Copenhagen strain) binds and inhibits IKBKB [129], and intracellular M2 (encoded by M2L) inhibits RelA (p65) nuclear translocation, but is not active in mice [96] (Fig 3C, blue cross). A55 (encoded by A55R) dysregulates NF-κB signaling by disrupting p65-importin interaction, is active in mice [130] and is expressed by SCV at the injection site (S1B Table). As the SCV-encoded inhibitors of NF-κB signaling are expressed only in the SCV-infected cells, the dominant NF-κB signatures are likely largely associated with uninfected cells stimulated and/or recruited by the pro-inflammatory environment [131].

Th1 cytokine signatures at 12 hours post vaccination
The cytokine USR profile at 12 hours post vaccination is dominated by cytokine signatures generally associated with Th1 responses (Fig 4A, red circles), in particular TNF, IL-1β and IFNγ, with in vivo induction of these cytokines by VACV suggested by previous studies [36 ,132,133]. This Th1 dominance is consistent with studies on recombinant MVA vaccines [134,135]. TNF is required for optimal adaptive immune responses to VACV and other immunogens [36,136]. IL-1 is important for host immune responses to VACV [132], with many vaccine adjuvants also inducing the release of IL-1 [137]. Finally, IFNγ has anti-VACV activity   Table. https://doi.org/10.1371/journal.ppat.1009215.g004

PLOS PATHOGENS
Injection site vaccinology of a vaccinia-based vector [138] and has adjuvant properties in a range of settings [139][140][141]. Although IL-27 was initially associated with Th1 responses, it is now recognized as a promoter of T regulatory cells [142] ( Fig 4A, IL27).

Chemokine signatures at 12 hours post vaccination
The chemokine signatures ( Fig 4B) are dominated by (i) CXCL12, which is made by many cell types and is strongly chemotactic for lymphocytes, (ii) CCL5 (RANTES), which is inter alia chemotactic for T cells and (iii) CXCL2, a neutrophil chemoattractant (consistent with Fig  1E). CXCL12 and CCL5 are also involved in dendritic cell (DC) recruitment [151,152]. CCL2 ( Fig 4B) is also induced by MVA [153,154] and is involved in DC maturation and induction of T cell immunity [155].
VACV infected cells secrete a number of proteins that bind and inhibit certain chemokines, although only A41 (encoded A41L) is active in the Copenhagen strain of vaccinia ( Fig 4B, purple text). A41 inhibits CCL21 [96], perhaps consistent with its low z-score ( Fig 4B). B29, encoded by B29R (C23L) inhibits multiple CC chemokines, but is inactive in the Copenhagen strain [156]. CCI inhibits a series of chemokines ( Fig 4B), but is not expressed on the Copenhagen strain of VACV [157,158].

STAT signatures at 12 hours post vaccination
Cytokines and chemokines bind to their receptors and activate transcription via STATs. The dominant STAT signatures (from S2D Table, IPA direct and indirect) were STAT1, STAT4 and STAT3 (Fig 4C), with STAT1 and STAT3 representing the top USRs by p value and STAT1 also the top USR by z-score when analyzed by IPA using direct only interaction (S2D Table, direct only). STAT1 forms complexes primarily STAT1-STAT1 homodimers (stimulated by IFNγ signaling) and STAT1-STAT2-IRF9 (ISGF3) (stimulated by type I IFN signaling). Using Interferome to interrogate the "Target molecules in dataset" listed for the STAT1 signature (S2D Table, direct and indirect), nearly all the target molecules were deemed IFNγ inducible (with most also inducible by type I IFNs). The relatively low p values and z-scores for STAT1 dimers would thus appear to be an under-annotation within IPA. The dominant STAT1 signature ( Fig 4C) is consistent with the dominant IFNγ signature in Fig 4A. The cytoplasmic VACV-expressed H1 (encoded by H1L) inhibits STAT1 and STAT2 [159], but again only in cells infected with VACV. C6 (encoded by C6L), as well as the aforementioned binding of TBK1 adaptors, also binds the TAD domain of STAT2 [160]. STAT4 signaling is induced by a number of cytokines including IL-12 and IL-2 [161] and is critical for IFNγ production during generation of Th1 responses [162]. STAT3 signaling is induced by a number of cytokines including IL-6 and OSM (and growth factors such as GM-CSF), with BCG vaccination recently shown to cause STAT3 phosphorylation in antigen presenting cells [163]. STAT6 is involved in driving Th2 responses [164] and has a negative zscore (Fig 4C), consistent with the Th1 dominance illustrated in Fig 4A.

Dendritic cell associated signatures
A range of mediators affect dendritic cell activities, with many of these identified as USRs by IPA analysis of DEGs for MQ vs SCV12hQ (Fig 4D andS2D Table; for references see S2F Table). The VACV protein A35 inhibits a number of these mediators (see above), as well as inhibiting class II antigen presentation [150]. Multiple key mediators needed for induction of adaptive immune responses by dendritic cells would thus appear to be active at the injection site 12 hours post vaccination.
Extensive bioinformatics treatments of >30,000 peripheral blood transcriptomes from >500 human studies of 5 vaccines provided 334 publically available gene sets in the form of Blood Transcription Modules (BTMs). BTM gene sets are associated with specific subsets of cells and/or their activities [165]. BTM gene sets associated with dendritic cells and dendritic cell activities (S2F Table) and Gene Set Enrichment Analyses (GSEAs) were used to determine whether genes from dendritic cell BTMs were significantly represented in the MQ vs SCV12hQ gene list (S2A Table). The GSEAs provided highly significant results (Fig 4E), illustrating that signatures associated with dendritic cells and their activities can be readily identified at the injection site 12 hours post vaccination. Such signatures likely underpin the immunogenicity of the poxvector system.

Injection site signatures at day 7 post vaccination
The most common side effects reported for MVA (licensed as a small pox vaccine in Europe as IMVANEX) were at the site of subcutaneous injection; most of them were mild to moderate in nature and resolved without any treatment within seven days [166]. To gain insights into the injection site responses after SCV vaccination, RNA-Seq of muscles on day 7 post-vaccination was undertaken to provide a gene list (MQ vs SCVd7Q, S2G Table), from which a DEG list (n = 1413 genes) was generated (S2H Table) by applying the same filters as above (q <0.01, FC >2 and sum of all counts across the six samples >6). Of the 1413 DEGs, 1337 were up-regulated, with 633 (47%) of these also up-regulated DEGs for MQ vs SCV12hQ.
Cytoscape analyses of up-regulated DEGs from MQ vs SCV12hQ (S2C Table) were compared with MQ vs SCVd7Q (S2I Table). Multiple top signatures (by FDR) associated with T cells and B cells were substantially more significant on day 7 than at 12 hours (Fig 5A and S2J  Table). For instance, FDR values associated with the GO Process terms "positive regulation of T cell activation" and "T cell differentiation" were �9 logs more significant by day 7, when compared with 12 hours post vaccination (Fig 5A and S2J Table). T cell receptor associated KEGG Pathways and GO Component terms were also more significant on day 7 (Fig 5A and  S2J Table). "T cell receptor complex" was also the top "GO Cellular Component" term by p value for day 7 up-regulated DEGs (S2J Table, Enrichr). A similar pattern emerged for B cell terms (Fig 5A and S2J Table).
GSEAs (as in Fig 4E) using gene lists from BTMs [165] associated with T cell differentiation and division, and B cell differentiation into plasma cells, showed significance for SCVd7Q, but not SCV12hQ (Fig 5B). Thus remarkably, these BTMs were able to identify signatures at the

PLOS PATHOGENS
Injection site vaccinology of a vaccinia-based vector  Table; statistics by paired t test for the full lists (parametric data distribution). (B) GSEAs were used to interrogate MQ vs SCV12hQ and MQ vs SCVd7Q gene lists using T cell and B cell BTMs (left to right) M14 (n = 12); M4.5 (n = 35); M156.0 and M156.1 (gene sets combined, n = 56) (for details of BTM gene sets see S2J Table). (C) As for A illustrating the downward trend of cell death annotations. Anotations not identified by the IPA analysis were nominally given a -log 10 FDR value of zero (y axis). Color coding as for A, but also Green line-UniProt Keywords, Brown line-Reactome Pathways. For descriptions of annotations see S2K Table. (Statistics by Wilcoxon Signed Rank tests; non-parametric data distribution). (D) IPA Diseases and Functions analysis of DEGs (up and downregulated) from MQ vs SCV12hQ and MQ vs SCVd7Q, illustrating the downward trend in z-scores for macrophage and neutrophil annotations (for description of annotations see S2L Table and S2M Table). (Statistics by Wilcoxon Signed Rank tests). (E) Major IPA USR pro-inflammatory cytokine annotations identified at 12 hours ( Fig 4A and S2D

PLOS PATHOGENS
Injection site vaccinology of a vaccinia-based vector injection site on day 7 that were associated with the development of adaptive immune responses. (IgG responses are known to be induced after SCV-ZIKA/CHIK vaccination [18] Table) illustrated that the cell death pathway annotations identified at 12 hours ( Fig 1G) were considerably less significant or absent for day 7 (Fig 5C and S2K Table). Analysis of the 1413 DEGs from MQ vs SCVd7Q (S2H Table) with IPA Diseases and Functions feature (S2L Table), showed a significant reduction in the z-scores of neutrophil-associated annotations on day 7 when compared to 12 hours ( Fig 5D and S2M Table). (The Cytoscape analysis also showed a highly significant reduction in FDR values for neutrophil terms, S2I Table, graph on right). These analyses indicate that progression of cell death and neutrophil infiltration is not a feature of SCV vaccination, likely consistent with the inability of SCV to produce viral progeny [19]. SCV does not cause a spreading infection, with vaccine-derived mRNA lost by day 7 ( Fig  1A). A similar IPA Diseases and Functions analysis of macrophage-associated annotations also illustrated a significant reduction by z-scores (Fig 5D and S2M Table), further indicating that injection site inflammatory responses were abating by day 7 [65].
The dominant pro-inflammatory cytokine USRs identified at 12 hours post vaccination ( Fig 4A) were substantially lower by day 7 post vaccination with respect to both−log 10 p values and z-scores (Fig 5E and S2D Table vs S2N Table). Fold changes in cytokine mRNA expression levels relative to MQ were also substantially lower on day 7 (Fig 5E), with the exception of IFNγ, which had a fold change relative to MQ of 4.17 at 12 h and a fold change of 5.26 relative to MQ on day 7, perhaps due to the emerging Th1 T cell responses (see above). IPA Diseases and Functions also showed reduced significance and z scores on day 7 for Inflammatory response (-log 10 p value 110.7 to 59.8, z-score 7.7 to 5.5) and Chronic inflammatory disorder (-log 10 p value 70.1 to 38.1, z-score -0.23 to -2.2) (S2L Table). These analyses again argue that inflammation at the injection site is abating on day 7, with persistent inflammation at the injection site generally deemed undesirable in most vaccination settings [167][168][169].

Loss of neutrophils and presence of eosinophils on day 7 post vaccination
H&E staining of the injection sites day 7 post vaccination supports the bioinformatics results described in the previous section. When compared with 12 hours (Fig 1D), necrotic muscle lesions were largely absent, with the cellular infiltrates less disseminated and more focal ( Fig  5F). In addition, in contrast to 12 hours (Fig 1E), neutrophils (stained with anti-Ly6G) were not observed in the day 7 cellular infiltrates ( Fig 5G), although the occasional neutrophil could be seen in blood vessels, illustrating that the staining had worked (Fig 5G, insert, arrowhead). In contrast to Fig 1E, ApopTag staining was also largely negative on day 7 (not shown). Loss of neutrophils is consistent with inflammation resolution [65]. Table) had much lower z-scores and p values on day 7 post vaccination (S2N Table). Numbers in the circles represent the log 2 Table) and MQ vs SCVd7Q (S2G Table); bars with red outline indicate significant fold change (q<0.05). https://doi.org/10.1371/journal.ppat.1009215.g005

PLOS PATHOGENS
Injection site vaccinology of a vaccinia-based vector Another feature of the resolving infiltrates on day 7 post vaccination (clearly evident from H&E staining) was the presence of eosinophils (Fig 5H), despite the retention of a dominant Th1 signature (Fig 5E). Many of these cells showed the morphological features of immature band eosinophils, as distinct from segmented mature eosinophils (Fig 5H).
At 12 hours post vaccination, genes specific to skeletal muscle were generally slightly downregulated (Fig 5I and S2A Table), consistent with the SCV infection-associated necrosis or pyroptosis ( Fig 1D). On day 7 post vaccination, Myh3 and Myh8 were significantly up-regulated ( Fig 5I and S2G Table), consistent with these genes being transiently up-regulated after muscle injury [170]. Tnni1, a skeletal myogenesis marker [171], was also up-regulated (Fig 5I). However, stable expression of Tnni2, Myoz1, Myoz3 and Dmd by day 7 (Fig 5I), argues that muscle regeneration had been largely completed at this time [170]; consistent with the H&E staining ( Fig 5F).

Concordance with vaccine and virus infection gene sets
The Molecular Signatures Database (MSigDB) provides a collection of >31,000 gene sets for use in GSEAs. All these gene sets were used to interrogate the full pre-ranked gene lists for MQ vs SCV12hQ (S2A Table) and MQ vs SCVd7Q (S2G Table). This analysis thus uses all the genes, rather than just the DEGs. Several signatures (with q<0.05) associated with influenza and yellow fever vaccines were identified (S5 Fig

No detection of adventitious agents or microbial infections
The SCV-ZIKA/CHIK vaccine was tested for sterility and mycoplasma before inoculation. However, insights into unforeseen microbial contamination (adventitious agents) present in the vaccine preparation [172] or microbial infection(s) that might have been introduced during vaccination [173], can also be gleaned from injection-site vaccinology. The reads that were not mapped or assigned to the mouse (S1C Fig) or the vaccine genomes (Fig 1A) by the STAR aligner were thus analyzed by Kraken, a metagenomic sequence classification tool [174,175]. Contamination of RNA-Seq samples with RNA from various sources from the laboratory and the environment is a well-recognized phenomenon [176][177][178], so samples that did not receive the SCV vaccine (MQ) provided a baseline for such background contamination. Unmapped and unassigned reads from MQ primarily identified Homo sapiens, Pasteurella and murine retroviruses (S6 Fig). Almost identical metagenomic patterns emerged for unmapped/unassigned reads from SCV12hQ and SCVd7Q (S6 Fig), arguing against the presence of adventitious agents in the vaccine, or injection-site infections in SCV-vaccinated animals being responsible for the adjuvant signatures described above.  Table). A DEG list was generated after application of two filters q<0.05 and the sum of counts

PLOS PATHOGENS
Injection site vaccinology of a vaccinia-based vector across all 6 samples >6, resulting in only 22 DEGs, of which 8 were up-regulated (S2Q Table). Of the 8 up-regulated DEGs, Tnfaip6 and Crispld2 have anti-inflammatory activities [179,180] and Thbs4 and Daglb have pro-inflammatory activities [181,182], with Gbp3 and Spon2 associated with antiviral responses [183,184] and Mrgprf associated with the itch response [185]. These results argue there were minimal transcriptional changes in joints after SCV-ZIKA/ CHIK vaccination.
IPA Diseases and Functions analysis of the 22 DEGs returned a significant "Inflammatory response" signature with a negative z-score, and cellular infiltrate terms with low z-scores (Table S2Q). Comparison of the 22 DEGs with those reported for collagen induced arthritis (CIA) (GSE13071) [186] also identified no obvious concordance (S2Q Table). With only 22 DEGs such analyses are somewhat underpowered, so the entire gene list (S2P Table) was interrogated for the presence of arthritic signatures using pre-ranked GSEAs and the 16 available genes sets associated with arthritis/arthralgia available from MSigDB. No significant arthritic signatures were identified, with all q values >0.99 (Fig 6A). Thus, even when using the entire RNA-Seq derived data set, no indication of arthritic signatures were evident.
To the best of our knowledge there is no evidence to suggest that CHIKV antigens migrate to joints and cause arthropathy. Instead, CHIKV arthritis [187], and viral arthritides generally [188], are associated with replication of virus in the joints, with SCV-ZIKA/CHIK not detected in feet (Fig 1A). Nevertheless, after the experience with an attenuated CHIKV vaccine (TSI-GSD-218) that caused arthralgia in 5 of 58 vaccine recipients [42], arthropathy caused by  Table). (B) MF vs SCVd7F provided 22 DEGs (with 2 filters applied q<0.05 and count sum >6) of which 8 were up-regulated (Table S2Q). Three of these were also up-regulated DEGs for CHIKV arthritis day 7 post infection (Table S2O; Table) and compared with the same annotations identified by IPA analysis of DEGs for CHIKV arthritis (S2O Table). https://doi.org/10.1371/journal.ppat.1009215.g006

PLOS PATHOGENS
Injection site vaccinology of a vaccinia-based vector CHIKV vaccines remains a standing concern for regulators [41]. We thus also compared the DEGs and pathways identified for SCVd7F, with those identified for CHIKV arthritis [116]. We re-analyzed the FASTQ files generated in the latter publication (deposited in NCBI Bio-Project PRJNA431476) using STAR aligner and the more recent mouse MM10 genome build (GRCm38 Gencode vM23). The complete gene list for day 7 feet (peak CHIKV arthritis) is provided in S2O Table. Three of the 22 DEGs (Daglb, Tgtp2 and Gbp3) were also present in the up-regulated DEGs for day 7 feet of CHIKV infected mice (S2O Table); however, the fold change of the latter two were substantially higher after CHIKV infection than after SCV vaccination (Fig 6B). The four pathways identified by IPA Diseases and Functions analysis of the 22 DEGs (S2Q Table) were also present for CHIKV arthritis, but the z-scores and p-values for these annotations were very much lower for SCVd7F (Fig 6C).
Overall these results argue that SCV-ZIKA/CHIK vaccination was not associated with a compelling arthritic signature, even though the injection sites (quadriceps muscles) were in the same legs as the feet that were used to generate the MF vs SCVd7F gene set.

Discussion
We provide herein a detailed injection site vaccinology analysis of a recombinant SCV vaccine in mouse muscle, to provide insights into vaccine gene expression, and host adjuvant signatures and immune responses. Of all the reads mapping to the SCV-ZIKA/CHIK vaccine, �20% mapped to the recombinant immunogens. IHC illustrated immunogen protein expression in skeletal muscle cells, with these cells showing histological signs of necrosis [63,64] and bioinformatics analyses indicating the presence of necrosis and necroptosis [62]. Adjuvant signatures were driven by TLRs, cytoplasmic RNA and DNA sensors and the inflammasome, with neutrophils potentially also contributing [53]. By day 7 vaccine transcripts and neutrophils were largely absent, and inflammation was abating, with the presence of what appeared to be tissue repair-associated eosinophils [189][190][191][192][193]. Although a previous live-attenuated CHIKV vaccine was associated with some arthropathy [42], no compelling arthritic signature was evident after SCV vaccination.
There was a marked concordance between (i) the z-scores for specific TLR and cytoplasmic sensor signatures identified at 12 hours post-vaccination and (ii) the relative importance of these pathways during the full course of infection, as gleaned from infection of knock-out mice with replication-competent poxviruses (summarized in S2E Table). The early dominant pathways identified by this injection site vaccinology approach was remarkably consistent with pathways previously published to be important for protection and/or immunopathology over the course of poxvirus infections (S2E Table). The concordance suggests that the early innate signatures identified herein are not overly dependent on (i) the ability of the vector to produce viral progeny [19] or (ii) the recombinant immunogen inserts. The concordance also suggests that these signatures may often be shared amongst different poxviruses and poxvirus vectors.
The ability to identify T and B cell response signatures on day 7 post-vaccination using GSEAs and BTMs, suggests RNA-Seq analyses of the injection site may also provide insights into the ensuing systemic adaptive immune responses. Migration of primed T cells to the site of infection is well described [194][195][196], so T cell signatures might be expected. However, plasma cells are ordinarily thought not to undergo such migration, arguing that the plasma cell signature may be due to non-specific migration, perhaps consistent with the relatively lower significance in the GSEA (Fig 5B, q = 0.03).
The immunogenicity of SCV, and likely poxvirus systems in general, would appear to be underpinned by the ability of such vectors to stimulate a broad range of pathways that are known to be stimulated by adjuvants that are already licensed for use in humans or are being

PLOS PATHOGENS
Injection site vaccinology of a vaccinia-based vector tested in humans. For instance, multiple TLR signatures were identified at 12 hours post vaccination (Fig 2A). The TLR4 agonist monophosphoryl lipid A is a component of Fendrix (hepatitis B vaccine) and Cervarix (human papilloma virus vaccine) (which are formulated in ASO4 adjuvant), as well as Shingrix (herpes zoster vaccine) (formulated in AS01 adjuvant). The dsRNA TLR3 agonists, Ampligen (Rintatolimod) and Hiltonol, also have well described adjuvant properties [197], with Hiltonol being tested in therapeutic cancer vaccine trials (Clinical-Trials.gov Identifier: NCT04345705 and NCT02423863). The TLR9 agonist, CpG oligonucleotide (CpG 1018), was recently approved as an adjuvant in Heplisav-B (hepatitis B vaccine). A range of cytoplasmic sensor signatures with known adjuvant activity were also identified (Fig 2B), including multiple inflammasome signatures; the best known human adjuvant, alum, is believed to mediate its activity via activation of the inflammasome [198]. Stimulation of cytoplasmic dsRNA sensors represents a key adjuvant activity for replicon-based RNA vaccines [199,200] and the utility of STING-activating adjuvants is being actively explored [201,202]. Although SCV does not generate viral progeny, it does replicate its DNA [19], perhaps explaining the dominant STING signature (Fig 2B). Virulent poxviruses inhibit STING activation via unknown factors, an inhibitory activity not found for MVA [203]. This activity is perhaps similarly absent for SCV or is inactive in mice.
A desirable feature for any vaccine is the avoidance of reactogenicity, a term describing a series of post-vaccination adverse events often associated with excessive injection site inflammation and systemic reactions such as fever [166,204]. A potential goal of transcriptome-based vaccinology is the identification of reactogenic signatures; however, consensus regarding the composition of such signatures, and/or when and where best to sample to obtain such signatures, has yet to be established [27]. CCL2 and CXCL10 up-regulation in peripheral blood was identified as potential biomarkers of vaccine-elicited adverse inflammation in mice after a number of different vaccines given i.m. [28]. These chemokines featured prominently at the injection site 12 hours after SCV vaccination (Fig 4B), although fold-change had reduced substantially by day 7 (log 2 4.12 to -0.73, and 8.35 to 3.17, respectively) (S2A Table and S2G  Table). CCL2 is also induced by MVA [154], can be important for avoiding immunopathology [51], and is induced by the licensed adjuvants, Alum and MF59 [205]. CXCL10 is also induced by MVA [206] and by the licensed adjuvant, MF59 [207]. Given the extensive clinical safety record of MVA vaccination [208] and the lack of overt injection site reactogenicity or fever observed in NHPs after SCV-ZIKA/CHIK vaccination [22], CCL2 and CXCL10 up-regulation at the injection site would thus not appear to be compelling biomarkers for adverse events after MVA or SCV vaccination. Similarly, a whole-blood systemic adverse event signature for yellow fever 17D vaccination has been reported, with 32 up-regulated genes on day 1 (but not day 3) associated with a range of systemic adverse events (either within 24 hours or a median time post vaccination of 6 days) [209]. GSEAs illustrate that this signature is highly significantly present in the SCV12hQ gene list, and is also significantly present in the SCVd7Q gene list (S2R Table). All the core enriched genes (S2R Table) were type I IFN stimulated genes (by Interferome), with type I IFN stimulated gene induction clearly present at the SCV injection site (Fig 4A). However, MVA vaccination is also associated with short term (i) increases in local type I IFN responses [68,111] and (ii) elevated serum IFNα levels [210,211]. Of note, SCV encodes B19R/B18R, an inhibitor of type I IFN responses [148]. Unlike Yellow fever 17D, MVA and SCV vaccinations are not associated with significant viremias or viral dissemination, reducing the probability of excessive serum type I IFN responses and systemic adverse events; although pyrexia, headache, myalgia, nausea, fatigue and/or chills are seen in a small percentage of MVA vaccine recipients [166,208]. Clearly, adverse event signatures identified in peripheral blood, may not be overly informative for understanding adverse events at the injection site. In addition, not only the presence of specific gene transcripts but also the magnitude

PLOS PATHOGENS
Injection site vaccinology of a vaccinia-based vector of gene induction are likely to be important, with the latter not fully taken into account by GSEAs. In humans, sampling injection sites is difficult, although emerging micro-sampling techniques may provide new avenues [212]. Ultimately RNA-Seq of injection sites in animal models should be able to provide early warnings in the vaccine development process of potential reactogenicity issues. Herein we show that, although SCV retains the ability to replicate its DNA ( [19], the injection site reactogenicity (like IMVANEX) [166] has largely resolved by day 7 post vaccination.
How might the information provided herein find utility for poxvirus vaccine design? The ZIKA and CHIK immunogens are inserted into B7R/B8R and A39R, respectively [18]. The 12 vector genes that were not expressed in vivo post-vaccination (S1B Table) offer other potential insertion sites for recombinant immunogens that would ostensibly have minimal impact on vaccine behavior. However, expression of these genes in human muscle might be checked, perhaps via use of human skeletal muscle organoids [213]. The multiple adjuvant pathway stimulated by SCV (Fig 2) might argue for a certain level of redundancy [211], which might allow certain inhibitors to be reintroduced with the aim of reducing reactogenicity, without compromising immunogenicity. For instance, B13R (also known as SPI-2) is absent in the Copenhagen strain of VACV and in SCV, and inhibits caspase I, a key protease for generation of bioactive IL1-β [214]. Reintroducing VACV IL-1 decoy receptors, SPI-2 (B13R) or B16R [215], into the vector may affect CD8 T cell responses [216,217], but have minimal effects on antibody responses [218]. Such secreted proteins should reduce the bioavailability of IL-1, a potent pyrogen, and may thereby reduce the risk of adverse events such as fever [144,166]. Recent sequencing of ancient Variola viruses from Viking corpses perhaps supports such strategies, as active expression of immune modulating genes may be associated with reduced pathogenicity [219,220]. Co-formulation of SCV with adjuvants [221,222] or encoding genetic adjuvants within SCV [223] might appear superfluous, given the large number of adjuvant pathways already being activated. Introducing apoptosis inhibitors to improve immunogenicity [14] may have minimal impact for i.m. injections of SCV (or other pox vectors) as skeletal muscle does not appear readily to undergo apoptosis [60,61], with B13R also well expressed at the injection site (S1B Table). The absence of the chemokine inhibitors B29R (C23L) (not expressed) and CCI (not present/functional) may contribute inter alia to effective recruitment of dendritic cells. Deletion of A41L might be tested to determine whether this would increase immunogenicity, given CCL21 recruits T cells and enhances T-cell responses [224]. Complement control proteins VCP (encoded by B27R) and C3 (encoded by C3L) were expressed at the mRNA level, with VCP deletion from VACV increasing anti-VACV antibody responses [13], suggesting deletion of these genes might enhance immunogenicity. One might consider deletion of N2L (an inhibitor of IRF3) as this was shown to improve the immunogenicity of a recombinant MVA vaccine [225] and also reduced the virulence of VACV [226]. However, the IRF3 signature is already very dominant (Fig 3B), so additional IRF3 activation (if possible) may not translate to significant improvements in immunogenicity. Deletion of C6L increased the immunogenicity of a recombinant MVA vaccine [17], presumably by relieving IRF3, IRF7 [105,106] and/or STAT2 inhibition [160]. However, C6L deletion could risk excessive type I IFN responses and increased reactogenicity, with excessive type I IFN responses associated with adverse events after administration of the yellow fever vaccine [209].
The presence of eosinophils in the resolving lesion on day 7 was unexpected. Eosinophils have been reported in pruritic papulovesicular eruptions in a case of generalized vaccinia after smallpox vaccination [227] and are well described as drivers of allergic diseases such as eosinophilic asthma [228,229]. However, recently a role for eosinophils in tissue repair and wound healing has emerged [189,190], particularly for muscle tissues [191][192][193]. That the eosinophils in these resolving post-vaccination lesions are distinct from inflammatory eosinophils is supported by the absence of IL-5 mRNA expression (S2G Table), with anti-IL-5 therapy used for eosinophilic asthma [230]. Transcripts for eosinophil cationic protein (Ear1) and eosinophil peroxidase (Epx) (granule components of inflammatory eosinophils) were also not detected. Eotaxins (CCL11, CCL24 and CCL26) were not up-regulated, with IL4, IL13, IL3 and GM-CSF (CSF2) transcripts absent (S2G Table). Using GSEAs we were unable to find any eosinophil gene signatures in the full MQ vs SCVd7Q gene list, suggesting that signatures for inflammatory eosinophils are distinct from tissue repair-associated eosinophils, with the signatures for the latter yet to be defined.
A limitation of using mice to analyze VACV-based vaccines is that several VACV-encoded inhibitors are not active in mice (encoded by M2L, A38L) and the activity of others in mice is not known (D9R, H1L, C10L). The activity in mice of B28R and B29R is also not known, but these inhibitors are not active in the Copenhagen strain of VACV. Others were found not to be expressed in mice (C16L, C23L, C22L), although it is unclear whether they are poorly expressed by SCV, poorly expressed in muscle or poorly expressed in mice. How critical these genes are to the overall interpretations presented herein is difficult to assess, given the presence of multiple overlapping and potentially cross-compensating pathways. Another limitation of this study is that we have not established which signatures are associated with the SCV vector and those that are associated with the ZIKA and CHIK immunogens. The concordance with existing poxvirus literature (S2E Table) might argue that the dominant signatures are associated with the vector. This contention is supported by the general observation that recombinant proteins are poorly immunogenic and require adjuvant before they can induce significant immune responses [231,232]. CHIKV and ZIKV immunogens may assemble into virus-likeparticles (VLPs) [233,234], but even VLPs often require formulation with adjuvants [235,236]. Should SCV be developed as a smallpox vaccine, separate injection site vaccinology studies might be warranted for SCV, which comprises the Copenhagen strain of VACV with D13L deleted [19]. Some changes in signatures from those reported herein might be expected as SCV would retain (i) A39R which encodes a protein with proinflammatory properties [237] and (ii) B7R/B8R which encode a virulence factor [238] and a chemokine receptor that regulates leukocyte trafficking [158]. SCV's DNA replication and protein expression may also be altered as VLP formation may inter alia compete for lipids with SCV [239][240][241], influence SCV-induced cell death [242] and/or modulate SCV's cytoskeletal rearrangements [243,244]. Also absent for SCV vaccination would be any competition for transcription-associated factors [245] imposed by the two synthetic strong early late promoters (Fig 1A).
An extensive history of poxvirus vector development has led to the first licensed recombinant poxvirus based vaccine for human use (MVA-BN-Filo) [12], with several others in late stage clinical trials. Injection site vaccinology may facilitate rationale refinement of pox vector design and contribute to progressing more such technologies towards registration and licensure.

Ethics statement
All mouse work was conducted in accordance with the "Australian code for the care and use of animals for scientific purposes" as defined by the National Health and Medical Research Council of Australia. Mouse work was approved by the QIMR Berghofer Medical Research Institute animal ethics committee (P2235 A1606-618M). Mice were euthanized using CO 2 .

PLOS PATHOGENS
Injection site vaccinology of a vaccinia-based vector polyprotein cassettes were inserted into A39R and B8R/B7R gene loci, respectively. The vaccine was produced in a non-GMP SCS line (comprising CHO-S cells expressing D13L and CP77 [19]) using protein-free cell culture conditions. SCV-ZIKA/CHIK was released from infected cells by multiple freeze thaw cycles, cell debris removed by centrifugation, followed by sucrose cushion purification and resuspension in 10 mM Tris HCl pH 8, 150 mM NaCl and storage at -80˚C. The vaccine was tested for sterility (turbidity in broth medium) and for mycoplasma (by PCR).

RNA-Seq
At the indicated times post vaccination, mice were euthanized using CO 2 and quadriceps muscles or feet placed individually into RNAlater (Life Technologies) overnight at 4˚C and then homogenized in TRIzol (Invitrogen) using 4 x 2.8 mm ceramic beads (MO BIO Inc., Carlsbad, USA) and a Precellys24 Tissue Homogeniser (Bertin Technologies, Montigny-le-Bretonneux, France) (6000 rpm on ice, 3 times 12 sec for feet and 2 times for 10 seconds for muscle). Homogenates were centrifuged (14,000 g x 15 min) and RNA extracted from the supernatants as per manufacturer's instructions. Following DNase treatment (RNAseq-Free DNAse Set (Qiagen)) and RNA purification (RNeasy MinElute Kit), RNA concentration and purity was determined by Nanodrop ND 1000 (NanoDrop Technologies Inc.). RNA samples were pooled so that for each group of 6 mice, 12 sets of quadriceps muscles or feet (severed at the bottom of the tibias after euthanasia) were used to create 3 biological replicates which contained equal amounts of RNA from 4 different mice. All replicates were then sent to the Australian Genome Research Facility (AGRF, Melbourne, Australia) for library preparation and sequencing. RNA integrity was assessed using the Bioanalyzer RNA 6000 Nano assay (Agilent) and libraries were prepared from 200 ng of total RNA using TruSeq Stranded mRNA library preparation kit (Illumina). The resulting libraries were assessed by TapeStation D1K TapeScreen assay (Agilent) and quantified by qPCR using the KAPA library quantification kit (Roche). Libraries were normalized to 2 nM and pooled for clustering on an Illumina cBot system using HiSeq PE Cluster Kit v4 reagents followed by sequencing on an Illumina HiSeq 2500 system with HiSeq SBS Kit v4 reagents with 100 bp paired-end reads.

Mouse genome alignments and differential gene expression
Mapping to the mouse genome and differential expression analysis was conducted at AGRF under commercial contract using their in-house pipeline. The quality of the raw sequencing reads were assessed using FastQC and MultiQC. Adapters were trimmed using the TrimGalore (0.4.4) program and reads with a length <30 bp or quality <10 were removed. Filtered reads were aligned to the Mus musculus reference genome (mm10; GTF file GRCm38.6annotation release 105) using the STAR aligner (v2.5.3a) with default parameters plus a parameter to restrict multi-mapping reads ('-outFilterMultimapNmax 2'). Counts per gene were summarized using the featureCounts (v1.4.6-p5) utility in Subread. A counts matrix was generated from the collective samples using in-house scripts and input to R (3.5.0) for differential expression analysis. Differential expression analysis was undertaken using EdgeR (3.22.3) with default settings and no filters, given the importance of key genes with low transcript

PLOS PATHOGENS
Injection site vaccinology of a vaccinia-based vector abundance [116] and the small percentage of cells infected by SCV-ZIKA/CHIK in the quadriceps muscles (with whole quadriceps muscles harvested for RNA-Seq). Counts were converted to relative counts (CPM) and normalized using the TMM method and modelled using the likelihood ratio test, glmLRT().

Ingenuity Pathway Analysis (IPA)
Up-Stream Regulator (USR), and Diseases and Functions features of Ingenuity Pathway Analysis (IPA) (QIAGEN) were used to interrogate the DEG lists using 'direct and indirect interaction' or 'direct only' interaction options.

Kraken metagenomic sequence classification
RNA-Seq reads that were not assigned/mapped to any of the GRCm38, VACV, ZIKV or CHIKV reference genomes were analyzed using metagenomic sequence classification. Unmapped or unassigned reads were converted to fastq format using Bedtools version 2.26.0 [253]. Triplicates were concatenated to produce one read file per treatment. Exact-match database queries were performed using Kraken version 2.1.1 with the Minikraken2 version 2 database [254]. Data was visualized using Krona version 2.7.1 [255].

Cytoscape and GSEAs using BTMs
Protein interaction networks of differentially expressed gene lists were visualized in Cytoscape (v3.7.2) [256]. Enrichment for biological processes, molecular functions, KEGG pathways and other gene ontology categories in DEG lists was elucidated using the STRING database [257].

Statistics
Statistical analysis of experimental data was performed using IBM SPSS Statistics for Windows, Version 19.0 (IBM Corp., Armonk, New York, USA). The paired t-test was used when the difference in variances was <4, skewness was >2 and kurtosis was <2. Otherwise the nonparametric Wilcoxon Signed Rank tests was used. Reads that were not assigned by the STAR aligner to mouse or vaccine genomes for MQ, SCV12hQ and SCVd7Q were analysed by Kraken, a metagenomic sequence classification tool. The output for all classified reads (All) was dominated by human sequences (Homo sapiens), Pasteurella multocida (a commensal of dogs, cats and rabbits), and murine retroviruses (primarily Mus musculus mobilized endogenous polytropic provirus). Mice in our animal house facility routinely test negative for Pasteurella species, so this contamination is unlikely to have originated from the mice. About 10% of the mouse genome is made up of endogenous retroviruses, with multi-mapped reads (reads that align to multiple locations in the mouse genome) left unassigned by the STAR aligner. (TIF) S1