We first extracted the genes described by the authors of these seven studies as showing a twofold or greater change in expression in response to infection. Two studies did not provide any data on decreases in gene expression (Paulus et al. 2006
; Clement et al. 2009
). The trend in all studies was the observation of more increases than decreases in gene expression. This trend may be due either to sensitivity of measurements or to differences in expression analysis methods. Using online database annotation tools, we cross-referenced the published gene identifiers to obtain the mouse Entrez Gene ID as a common identifier for each gene (Dennis et al. 2003
; Diehn et al. 2003
; Maglott et al. 2007
). We selected mouse for the common organism identifier because a large number of herpesvirus studies are conducted in this species. This mapping and compilation yielded a total of approximately 650 genes (sources listed in ).
We next searched for common themes among these host responses to herpesvirus infection. Since the number of genes provided by most of these studies is small, we did not limit our query to those genes that matched exactly across multiple studies. Instead, we focused on commonalities across gene families. For instance, we considered effects on multiple subunits of the same protein (e.g., AP complex subunits) or different members of the same protein family (e.g., multiple kinesins) to indicate a likely importance of these families in the host response to herpesvirus infection. We found 50 gene families that were affected in two or more studies (; individual gene names provided in Supplemental Table 1
). We found 15 genes that did match exactly across two or more studies (). For example, the chemokine ligand CXCL12 was noted in two studies (; Prehaud et al. 2005
; Clement et al. 2008
), whereas the CXCL gene family members CXCL12, CXCL14, and CXCL9 are found in a total of four studies (; Kramer et al. 2003
; Kent and Fraser 2005
; Prehaud et al. 2005
; Clement et al. 2008
Gene families noted in two or more studies of neuronal or nervous system responses to alphaherpesvirus infection
Exact gene matches found in more than one study of neuronal or nervous system responses to alphaherpesvirus infection
When assessing changes in gene expression observed by large-scale microarray studies, it is important to keep in mind several caveats. First, the diversity of platforms used to measure changes, and the different approaches used to analyze these data, mean that it is extremely difficult to find exact transcript matches across different platforms and experimental studies. Different microarray platforms may hybridize a slightly different segment of a given transcript, and thus may measure one or all splice variants, or be more or less sensitive than another platform. Second, when comparing lists of affected genes or pathways from multiple studies, any summaries of pathways not
affected in a given study must be considered carefully. The failure to observe an effect on a given pathway is not conclusive unless the study has observed an effect on this pathway in another situation, so that their ability to successfully measure this pathway can be confirmed. As an example, Prehaud et al. (2005)
used the human NT-2 neuronal cell line as a model system to study neuronal responses to HSV-1 and rabies virus infection. They observed that while approximately 25% of the neuronal gene expression responses to rabies virus were immune-related, only 5% of the responses to HSV-1 infection were immune-related. This parallel approach provides confidence in the limited immune response of these neurons to HSV-1, since an abundance of immune response transcripts was elicited by rabies virus infection.
Exact gene family members may not match across multiple studies because we have mapped these results across infection models in multiple species (human, rabbit, mouse, and rat), with five different herpesvirus strains, during different phases of infection (productive, latent, and reactivation), and in both in vivo and in vitro studies. However, we gained further insight by examining the overlap between subsets of these data more closely. For instance, in vitro studies offer the advantage of studying a pure population of neuronal cells. On the other hand, the single-cell-type isolation of in vitro systems is artificial, and the possibility must be considered that responses observed in vitro would not occur in the presence of normal intercellular interactions in vivo. In contrast, animal models more fully encompass the nervous system’s true state, in that infected neurons interact closely with glia and microglia, as well as infiltrates from the adaptive immune system. However, in vivo models have the confounding issues of asynchronous infection, variability in the degree of immune cell infiltration, and potential for decreased detection sensitivity because expression changes are blurred by multiple cell types in the tissue sample. Therefore, cases where a gene family’s expression level has been observed to change both in vivo and in vitro strongly imply a cell autonomous neuronal response.