Highly Predictive Model for a Protective Immune Response to the A(H1N1)pdm2009 Influenza Strain after Seasonal Vaccination

Abstract

Understanding the immune response after vaccination against new influenza strains is highly important in case of an imminent influenza pandemic and for optimization of seasonal vaccination strategies in high risk population groups, especially the elderly. Models predicting the best sero-conversion response among the three strains in the seasonal vaccine were recently suggested. However, these models use a large number of variables and/or information post- vaccination. Here in an exploratory pilot study, we analyzed the baseline immune status in young (<31 years, N = 17) versus elderly (≥50 years, N = 20) donors sero-negative to the newly emerged A(H1N1)pdm09 influenza virus strain and correlated it with the serological response to that specific strain after seasonal influenza vaccination. Extensive multi-chromatic FACS analysis (36 lymphocyte sub-populations measured) was used to quantitatively assess the cellular immune status before vaccination. We identified CD4+ T cells, and amongst them particularly naive CD4+ T cells, as the best correlates for a successful A(H1N1)pdm09 immune response. Moreover, the number of influenza strains a donor was sero-negative to at baseline (NSSN) in addition to age, as expected, were important predictive factors. Age, NSSN and CD4+ T cell count at baseline together predicted sero-protection (HAI≥40) to A(H1N1)pdm09 with a high accuracy of 89% (p-value = 0.00002). An additional validation study (N = 43 vaccinees sero-negative to A(H1N1)pdm09) has confirmed the predictive value of age, NSSN and baseline CD4+ counts (accuracy = 85%, p-value = 0.0000004). Furthermore, the inclusion of donors at ages 31-50 had shown that the age predictive function is not linear with age but rather a sigmoid with a midpoint at about 50 years. Using these results we suggest a clinically relevant prediction model that gives the probability for non-protection to A(H1N1)pdm09 influenza strain after seasonal multi-valent vaccination as a continuous function of age, NSSN and baseline CD4 count.

Fig 1. Serological response to A(H1N1)/pdm09 as function of age and number of strains that are sero-negative at baseline (NSSN) in the pilot and validation studies.

A) Pilot study: Non-response (HAI<10) and non-protection (HAI<40) to A(H1N1)pdm09 at day 21 post-vaccination are higher (30% and 50%, respectively) in old (>57 years) as compared to young (<31 years) vaccinees (12% and 12%, p = NS and p = 0.04, respectively). Furthermore, HAI titers at day 21 among responders are significantly (P<0.03) lower in old donors as compared to young donors. B) Pilot study: Non-response and non-protection to A(H1N1)pdm09 are higher (21% and 29%, respectively) in donors which were sero-negative to all 3 vaccine strains at baseline (NSSN = 3) as compared to donors which were sero-negative to H1N1 but sero-positive to the other 2 strains in the vaccine (NSSN = 1, 0%, p = NS). HAI titers among responders are not related to NSSN. C) Validation study: Non-response and non-protection to A(H1N1)pdm09 at day 21 post-vaccination are validated to be higher (38% and 46%, respectively) in old (>50 years) as compared to young (<50 years) vaccinees (7% and 10%, p = 0.02 and p = 0.01, respectively). However, HAI titers among responders are not related to age in the validation study. D) Validation study: Non-response and non-protection to A(H1N1)pdm09 are higher (29% and 33%, respectively) in donors which were sero-negative to all 3 vaccine strains at baseline (NSSN = 3) as compared to donors which were sero-negative to H1N1 but sero-positive to the other 2 strains in the vaccine (NSSN = 1, 0% and 11%, p = 0.04 and p = 0.05, respectively).

Fig 2. Hierarchical network representation of immune cell-subset counts at baseline with respect to A(H1N1)pdm09 protection in the pilot study.

We monitored 36 immune cell subpopulations in A(H1N1)pdm09 sero-negative donors and compared donors who became either sero-protected or not at day 21 after vaccination. We observe a number of cell populations for which the counts are significantly different between protected and non-protected donors, specifically on the CD4⁺ T cell axis. The colors indicate the relative median counts of the groups. Significant differences were determined using the Wilcoxon-Test and indicated with * for p<0.05 and ** for p<0.01.

Fig 3. Prediction of serological response to A(H1N1)pdm09 as function of counts of major lymphocyte sub-populations in the pilot study.

A) H1N1 sero-negative individuals that were sero-protected (blue circles) to H1N1 at day 21 post-vaccination had higher counts of CD4⁺ T cells, CD8⁺ T cells or B cells (CD19) at baseline as compared to non-protected individuals (red triangles, P<0.05). However, there is no continuous correlation between the various cell counts and the HAI titer. B) Logistic regression shows that a model combining baseline CD4⁺ T cell counts with age and NSSN, is the best predictor of sero-protection, with a high ROC-AUC = 0.92 and significant p-value = 0.00002. Similar models for CD8⁺ T cells or B cells (CD19⁺) give reasonable albeit lower prediction values. C) The combination of CD4⁺ T cell counts, NSSN and age gives a highly accurate 89% prediction of non-protection (when selecting for a specificity = 100%, or positive predictive value ppv = 100%, in order to capture all non-responders), with a sensitivity of sen = 86% and negative predictive value npv = 69%. The other lymphocyte sub-populations counts give less accurate predictions.

Fig 4. Multi-factorial association of serological response to the A(H1N1)pdm09 influenza strain as function of age, NSSN and total, naïve and influenza specific activated CD4⁺ T cells in the pilot study.

A) High CD4⁺ T cell counts give rise to sero-protection (HAI>40) irrespective of age or NSSN. Age still plays a role in vaccinees with NSSN = 3 and low CD4⁺ T cell counts, where 80% of old (red circles) versus only 20% of young (blue circles) are non-protected (p = 0.01). The same trend (p = NS) is also seen for NSSN = 2 (triangles) with low CD4⁺ T cell counts, albeit with better response than NSSN = 3. The only 2 donors with NSSN = 1 are sero-protected even if they are old and have low CD4⁺ T cell counts. B) Naive CD4⁺ T cell counts show a trend (p = NS) for a positive association with serological response in all age groups (NSSN = 2–3). C) Influenza specific activated CD4⁺CD40L⁺ T cell counts are not associated with serological response in any of the age groups (NSSN = 2–3).

Fig 5. Prediction of non-protection to the A(H1N1)pdm09 influenza strain as function of the combination of age, NSSN and CD4⁺ T cells after the validation study.

The logistic regression model combining baseline CD4⁺ T cell counts with age and NSSN is validated with a high ROC-AUC = 0.85, significant p-value = 0.0056 and high accuracy of 85% for the same age groups (<31 and >49 years) as in the pilot study (left panels in A and B). However, the addition of the middle age group (31–49 years) in the validation study somewhat reduces the accuracy of the prediction when using age as a linear function (center panels in A and B), because donors with these ages respond rather like the younger donors. Transformation of age to a sigmoid based function (with a midpoint age of 50 years) gives the best prediction with accuracy 85% and a highly significant p-value = 0.0000004 when combining both studies (right panels in A and B). The multi-factorial risk profile for non-protection (HAI<40) to the A(H1N1)pdm09 influenza strain is clearly seen (C) when combining the sero-negative vaccinees from both studies (N = 80). Donors with high baseline CD4⁺ T cell counts (>860 cells/μL) are all protected (p = 0.02 for NSSN = 3), as well as young (<50 years) donors with low CD4⁺ counts but NSSN = 1–2. Non-protection is only observed for old donors with low CD4⁺ counts (20%, 50% and 64% for NSSN = 1, 2 and 3 respectively) and for young donors with low CD4⁺ counts and NSSN = 3 (24%). Lastly, a prediction model (D) for the probability of non-protection to the California H1N1 strain is obtained by simulating the continuous contribution of age (after logistic function transformation from 20 years young in blue to 80 years old in red), NSSN and baseline CD4⁺ T cell counts, where the combined effect of the 3 variables can be clearly seen.