**6. Development of an alternative assay for the leukopenic toxicity test based on biomarker gene expression**

The biomarker genes for the safety assessment of an influenza vaccine are characterized by the biological activity that can be detected by the ATT and LTT. Specifically, biomarker gene expression levels and the WBC count with body weight changes show a negative correlation [31]. Momose *et al*. (2015) reported that a virosomal influenza vaccine caused only a body weight loss and did not cause leukopenia; however, some of the marker genes showed increased expression levels at that time point [31]. In other words, it seems that all the marker genes cannot respond uniformly via the same mechanism of action. Therefore, we considered whether the leukocytopenic activity could be evaluated with the expression of the marker genes responsible for the leukopenic activity, and we searched for biomarker genes associated with leukocytopenic activity. Furthermore, we devised a method for WBC count-predicting systems involving only the biomarker gene expression levels. If this method is established, it will be possible to set up the WBC number prediction using the biomarker gene expression and body weight loss evaluation by the ATT in one test system. This strategy will reduce the number of animals required and shorten the testing duration. We tried to identify the genes useful for the prediction of the WBC count from the biomarker gene set by multiple linear regression analysis and a stepwise method [32]. In the multiple regression analysis method, a linear equation expressed by the following formula was derived, and a predicted value was calculated. In particular, the leukocyte count of the animals and data on the expression levels of all the biomarker genes were acquired. The animals were inoculated with the WPV or HAV, and the expression levels of marker genes and numbers of WBCs were then determined. Multiple regression analysis was performed on the acquired data. A linear equation was then derived. The regression equation is shown below.

$$\text{(Predicted WBC)} = \text{(intercept)} + \boldsymbol{\beta}\_1 \mathbf{x}\_1 + \boldsymbol{\beta}\_2 \mathbf{x}\_2 + ... + \boldsymbol{\beta}\_u \mathbf{x}\_u \tag{1}$$

values of individuals are analyzed, the deviation between the predicted value and the measured value is small [32]. In addition, variations in the WBC count owing to individual differences are reproduced with high accuracy [32]. Therefore, it was shown that the leukopenic activity can be predicted by means of the identified marker gene set. With this method, it is expected that it will be possible to carry out WBC count reduction assays and abnormal toxicity negative tests by expression analysis of one biomarker gene. As mentioned in the previous section, the development of adjuvanted vaccines has advanced in recent years. Some adjuvants, like WPVs, exert a leukocytopenic activity. Leukopenic activity is also present in compounds with an excellent adjuvant activity such as Poly I:C and R848 [43]. Therefore, we analyzed the CpG K3 adjuvant, which manifested a slight leukopenic activity according to the newly developed multiple regression Equation [32]. As a result, a slight leukopenic activity was observed in animals that received the CpG K3 adjuvant in combination with the SV. Furthermore, when the expression levels of the marker genes were analyzed, and the predicted WBC count was calculated from their expression levels, the leukocyte count reduction by the CpG K3 adjuvant could be predicted with high accuracy [32]. At the same time, however, it became clear that leukocytosis is unpredictable in this system [32]. When the CpG K3 adjuvant was inoculated at low concentrations, the leukocyte count tended to

Genomic Approaches Enable Evaluation of the Safety and Quality of Influenza Vaccines…

http://dx.doi.org/10.5772/intechopen.76361

121

**Figure 1.** Prediction of a leukocyte count reduction by means of marker gene expression levels in mice treated with influenza vaccine. In the leukopenic toxicity test (LTT), mice were intraperitoneally injected with 0.5 mL of one of influenza vaccines. The dosing of whole-particle inactivated influenza vaccine (WPV) and hemagglutinin-split vaccine (HAV) was 15 μg hemagglutinin (HA)/0.5 mL. Poly I:C was reconstituted in an appropriate volume of the HAV solution to obtain concentrations of 1, 5, 10, and 20 μg poly I:C per dose. Saline (SA) served as the cssssontrol. Sixteen hours after vaccination, blood was collected to assess the numbers of leukocytes, and the lungs were immediately excised. The collected lung tissue was subjected to gene expression analyses. The predicted values were calculated via the multiple regression equation. The coefficient values and values serving as the intercept are indicated in **Table 2**. The error bar

indicates standard deviation.

In this equation, "x" is substituted by the marker gene expression that corresponds to its coefficient (β); "n" indicates the number of factors corresponding to the number of selected genes. The intercept was used for calculation of the WBC number in the multiple regression analysis.

Precision of the prediction differs depending on the combination of marker genes. Therefore, by the stepwise method, a linear equation was derived that contains the combinations of marker genes having the highest prediction accuracy. As a result, some gene combinations (models) were selected (**Table 2**). Predicted leukocyte numbers produced by these models can predict leukopenia caused by WPVs with high accuracy (**Figure 1**). Even if the predicted


**Table 2.** Multiple regression by the stepwise forward selection method for the leukopenic reaction prediction model (cells/μl blood).

values of individuals are analyzed, the deviation between the predicted value and the measured value is small [32]. In addition, variations in the WBC count owing to individual differences are reproduced with high accuracy [32]. Therefore, it was shown that the leukopenic activity can be predicted by means of the identified marker gene set. With this method, it is expected that it will be possible to carry out WBC count reduction assays and abnormal toxicity negative tests by expression analysis of one biomarker gene. As mentioned in the previous section, the development of adjuvanted vaccines has advanced in recent years. Some adjuvants, like WPVs, exert a leukocytopenic activity. Leukopenic activity is also present in compounds with an excellent adjuvant activity such as Poly I:C and R848 [43]. Therefore, we analyzed the CpG K3 adjuvant, which manifested a slight leukopenic activity according to the newly developed multiple regression Equation [32]. As a result, a slight leukopenic activity was observed in animals that received the CpG K3 adjuvant in combination with the SV. Furthermore, when the expression levels of the marker genes were analyzed, and the predicted WBC count was calculated from their expression levels, the leukocyte count reduction by the CpG K3 adjuvant could be predicted with high accuracy [32]. At the same time, however, it became clear that leukocytosis is unpredictable in this system [32]. When the CpG K3 adjuvant was inoculated at low concentrations, the leukocyte count tended to

**Predictor variable Model 1 Model 2 Model 3 Model 4**

a linear equation expressed by the following formula was derived, and a predicted value was calculated. In particular, the leukocyte count of the animals and data on the expression levels of all the biomarker genes were acquired. The animals were inoculated with the WPV or HAV, and the expression levels of marker genes and numbers of WBCs were then determined. Multiple regression analysis was performed on the acquired data. A linear equation was then

(Predicted WBC) = (intercept) + *β*<sup>1</sup> *x*<sup>1</sup> + *β*<sup>2</sup> *x*<sup>2</sup> + …+*β<sup>n</sup> xn* (1)

In this equation, "x" is substituted by the marker gene expression that corresponds to its coefficient (β); "n" indicates the number of factors corresponding to the number of selected genes. The intercept was used for calculation of the WBC number in the multiple regression analysis. Precision of the prediction differs depending on the combination of marker genes. Therefore, by the stepwise method, a linear equation was derived that contains the combinations of marker genes having the highest prediction accuracy. As a result, some gene combinations (models) were selected (**Table 2**). Predicted leukocyte numbers produced by these models can predict leukopenia caused by WPVs with high accuracy (**Figure 1**). Even if the predicted

derived. The regression equation is shown below.

120 Influenza - Therapeutics and Challenges

**Intercept** 2141.2 5390.6 4222.5 5293.7 *C2* — — — 502.8 *Trafd1* −3196.2 −2886.6 — −1131.1 *Irf7* — — — −94.7 *Csf1* — — — 1118.1 *Ngfr* −1344.8 — — −360.1 *Ifi47* — — — 472.3 *Ifrd1* — — — −1628.5 *Psme1* 4099.9 — — — *Tap2* 3084.6 1839.1 — — *Cxcl11* −0.3847 −0.1217 — — *Lgals9* −8.0607 — — — *Zbp1* −197.49 — — — *Cxcl9* — — 1.8226 — *Lgals3bp* — — −552.93 — *Tapbp* — — 349.20 —

**Table 2.** Multiple regression by the stepwise forward selection method for the leukopenic reaction prediction model

(cells/μl blood).

*β β β β*

**Figure 1.** Prediction of a leukocyte count reduction by means of marker gene expression levels in mice treated with influenza vaccine. In the leukopenic toxicity test (LTT), mice were intraperitoneally injected with 0.5 mL of one of influenza vaccines. The dosing of whole-particle inactivated influenza vaccine (WPV) and hemagglutinin-split vaccine (HAV) was 15 μg hemagglutinin (HA)/0.5 mL. Poly I:C was reconstituted in an appropriate volume of the HAV solution to obtain concentrations of 1, 5, 10, and 20 μg poly I:C per dose. Saline (SA) served as the cssssontrol. Sixteen hours after vaccination, blood was collected to assess the numbers of leukocytes, and the lungs were immediately excised. The collected lung tissue was subjected to gene expression analyses. The predicted values were calculated via the multiple regression equation. The coefficient values and values serving as the intercept are indicated in **Table 2**. The error bar indicates standard deviation.

increase. This increase in the WBC count could not be predicted from the biomarker gene expression levels. The possible reason is that multiple linear regression analysis was performed on the animals inoculated with vaccines with leukopenic activity. The physiological association of the extracted biomarker gene with leukopenia could be another reason for this phenomenon. We have identified biomarker genes highly correlating with leukocyte counts in mathematical terms, ignoring the functions of the genes. There is a report that apoptosis of leukocytes caused by type 1 interferons (IFNs) could be a mechanism underlying the WBC count reduction by WPVs [47]. The marker gene set contains many genes related to type 1 IFN. By contrast, in model 4 (**Table 2**), more than a half of the genes related to type 1 IFN were omitted because they lacked predictability. As a result, it is conceivable that a correlation cannot be obtained from only a simple expression level because of the time lag between gene expression and actual leukocyte depletion; furthermore, a gene itself forms a complicated network.

alternative assay, phenotypic changes in animals, such as body weight loss, cannot be assessed in cultured cells. On the contrary, marker genes linked to these bioactivities can be identified at the cultured-cell level. Biomarker genes can make it possible to link cellular data with biological reactions observed in animal phenotypes. We are currently working on demonstrating the usefulness of marker genes and their expression mechanisms. Most of the marker genes are involved in an immune response and are related to type 1 IFN signaling and innate immune responses [39]. According to these findings, it is possible that the usefulness of biomarker genes evaluated in animals can be extrapolated to cultured cells, if such cell lines as peripheral blood mononuclear cells, immune cells, and alveolar epithelial cell lines are employed in the assays. If an alternative (*in vitro* method) for the ATT and LTT is developed, it will be possible to secure the safety and quality of the current ATT and LTT by animal-free testing. This approach is expected to reduce the number

Genomic Approaches Enable Evaluation of the Safety and Quality of Influenza Vaccines…

http://dx.doi.org/10.5772/intechopen.76361

123

**8. Establishment of a new evaluation method for vaccine or adjuvant** 

Analysis by the genomics technology can be applied not only to the search for biomarkers but also to mechanistic analyses. Besides, it is possible to classify each biological reaction by hierarchical clustering analysis, according to microarray analysis results. Microarray analysis is the most information-rich assay; however, it is inefficient in terms of cost and labor. In the case of a clear-purpose test such as safety evaluation and quality control, it is expected that robust results will be obtained by using only selected highly important genes for evaluation. Therefore, if we consider the function of the genes identified as safety or quality evaluation markers for influenza vaccines, then the biological activity profile of the vaccine may be predicted. For example, genes such as *Irf7* are induced by type 1 IFN [55], genes such as *Psmb9* and *Tap2* are involved in antigen presentation [56], and *Csf1* participates in the activation of monocytes and macrophages [57]. Thus, expression levels of these genes could serve as indicators of the mode of action and help in the development of a biological activation assessment tool. These genes are thought to be involved in innate immunity, in which responses are observed at a relatively early time point after vaccination. Indeed, expression of these genes was assayed 16 h after vaccination. Therefore, long-term toxicity due to the vaccine (e.g., autoimmune and chronic inflammatory reactions) cannot be assessed. Safety evaluation by means of these biomarker genes is helpful for the development of adjuvant-containing vaccines. This is because most adjuvants are designed to activate the innate immune system. Adjuvants enhance innate immunity via cytokine production and activation of antigen-presenting cells; however, strong activation of innate immunity causes uncontrollable inflammatory reactions. This problem could lead to a fever, pain, and swelling, which appear as adverse reactions. Thus, adjuvants are required to have strong innate-immunity–activating effects, but at the same time, good safety. On the other hand, it is difficult to distinguish between the effectiveness and safety of vaccines. For example, interleukin (IL)-6 and type 1 IFN are important for the induction of adaptive immunity and are favorable for vaccine efficacy [58, 59]. Nevertheless, excess production of IL-6 or type 1 IFN causes a cytokine storm. Thus, safety can be evaluated with the same biological vector as that of effectiveness. In other words, if the factor of effectiveness becomes excessive, toxic effects

of animals tested and to shorten the testing period.

**bioactivity based on biomarker gene function**
