COVID-19 Vaccine Boosters in People With Multiple Sclerosis: Improved SARS-CoV-2 Cross-Variant Antibody Response and Prediction of Protection

Abstract

Background and objectives: Although disease-modifying therapies (DMTs) may suppress coronavirus disease 2019 (COVID-19) vaccine responses in people with multiple sclerosis (pwMS), limited data are available on the cumulative effect of additional boosters. Maturation of Spike immunoglobulin G (IgG) to target a greater diversity of SARS-CoV-2 variants, especially past the BA.1 variant, has not been reported. In addition, the prediction of variant-specific protection, given that Spike antibody testing is not performed routinely, remains a challenge. We, therefore, evaluated whether additional vaccine doses improved the breadth of cross-variant recognition to target emerging SARS-CoV-2 variants. Machine learning-based models were designed to predict variant-specific protection status.

Methods: In a prospective observational cohort (n = 442), Spike IgG titers and live virus neutralization against D614, BA.1, BA.2, BA.5, XBB.1.1, XBB.1.5, and EG.5.1 variants were determined in 1,011 serum samples (0-12 months after 2-4 doses). Predictive protection models were developed by K-fold cross-validation on training and test data sets (random split 70:30).

Results: After primary vaccination, pwMS on immunosuppressive disease-modifying therapy (IMM-DMT) had 10-fold and 7.2-fold lower D614 Spike IgG titers than pwMS on low-efficacy (LE)-DMT and cladribine (p < 0.01). After 4 doses, pwMS on IMM-DMT had significantly lower Spike IgG titers, compared with pwMS on low-efficacy disease-modifying therapy, for D614 (p < 0.05), as well as BA.1, BA.2, BA.5, XBB.1, XBB.1.5, and EG.5.1(p < 0.01). The breadth of Spike IgG to recognize variants other than the cognate antigen increased after 4 doses of all DMTs. Although pwMS on IMM-DMT displayed reduced cross-variant recognition, a fourth dose resulted in a 2-4-fold increase in protection against newer variants and a reduction in two-thirds of pwMS without protective Spike IgG (p < 0.0001). Tixagevimab and cilgavimab did not induce additional cross-variant protection. Variant-specific predictive models of vaccine protection were influenced by treatment, time since primary vaccination, and age, with high sensitivity (99.4%, 95% CI 96.8-99.99) and specificity (72.0%, 95% CI 50.6-87.9) for XBB.1.5/EG.5.1 variants.

Discussion: Despite not eliciting adequate antibody response in pwMS on IMM-DMT, COVID-19 boosters improve the breadth of the humoral response against SARS-CoV-2 emerging variants. Vaccine protection can be predicted by statistical modeling.

Figure 1. Seroconversion in pwMS after Primary COVID-19 Vaccination

(A) Schematic showing time line of the sample collection in the study (red) and vaccination (orange) of the cohort alongside the different waves of infections in Australia (blue) from July 17, 2019, to December 12, 2023. A total of 1,011 serum samples from pwMS were collected at various time points after COVID-19 vaccination. Collection time lines for peak samples after any vaccine doses, i.e., 1MD2, 1MD3, 1MD4, and 1MD5, are shown in the blue rectangles. Additional samples, i.e., 2–12MD2, 3–9MD3, 3–6MD4, or 3MD5, are shown in brackets. In addition to the 903 serum samples shown on the schematic, n = 67 serum samples were collected after documented SARS-CoV-2 infection and n = 41 at uncategorized vaccination time points. (B) D614 Spike IgG titers in prevaccinated pwMS at baseline (n = 326). (C) Increase in D614 Spike IgG titers at 1MD2 compared with baseline in pwMS on LE-DMT, CL, and M/HE-DMT with paired BL and 1MD2 samples. (D) Comparison of D614 Spike IgG titers at 1MD2 in pwMS on LE-DMT, CL (median 291 days IQR 257 after last CL dose), M/HE-DMT, and IMM-DMT. LE-DMTs included dimethyl fumarate (n = 9), interferon β-1a (n = 2), teriflunomide (n = 2), interferon β-1b (n = 1), and glatiramer acetate (n = 1). M/HE-DMTs included ocrelizumab (n = 49), natalizumab (n = 19), fingolimod (n = 17), alemtuzumab (n = 5), ofatumumab (n = 3), and siponimod (n = 2). The NTZ/AL group is a subgroup of M/HE-DMT and included natalizumab (n = 19) and alemtuzumab (n = 2, >2 years after last infusion). The IMM-DMT group is a subgroup of M/HE-DMT and included ocrelizumab (n = 49), fingolimod (n = 17), alemtuzumab (n = 3, <2 years after last infusion), ofatumumab (n = 3), and siponimod (n = 2). The CD20-DMT group is a subgroup of IMM-DMT and included ocrelizumab (n = 49, median 154 days IQR 82 after last dose) and ofatumumab (n = 3, median 12 days IQR 1 after last dose). The S1P-DMT group is a subgroup of IMM-DMT and included fingolimod (n = 17) and siponimod (n = 2). Median and IQR are shown. *p < 0.05, **p < 0.01, ***p < 0.0001. 1MD2 = 1 month after second dose (primary vaccination); 1MD3 = 1 month after third dose; 1MD4 = 1 month after fourth dose; 1MD5 = 1 month after fifth dose; BL = baseline; CD20-DMT = CD20-depleting disease-modifying therapy; CL = cladribine; IMM-DMT = immunosuppressive disease-modifying therapy; LE-DMT = low-efficacy disease-modifying therapy; M/HE-DMT = moderate to high-efficacy disease-modifying therapy; NTZ/AL = natalizumab and alemtuzumab (2–4 years after last infusion); S1P-DMT = sphingosine-1-phosphate receptor modulator disease-modifying therapy; T = threshold of seropositivity; VAX = vaccine dose.

Figure 2. Longevity and Breadth of Immune Response in pwMS After Additional COVID-19 Vaccine Doses

(A) Loess curve depicting trajectory of immune response over a period of >2 years after primary vaccination in pwMS with median (solid line) and 95% CI (shaded area) fitted for D614 Spike IgG titers in each treatment group separately. (B) Median Spike IgG titers against D614, BA.1, BA.2, BA.5, XBB.1, XBB.1.5, and EG.5.1 at 1MD2 vs 1MD4 in pwMS on DMTs. LE-DMT 1MD2 (D614 n = 22, BA.1 n = 15, BA.2 n = 22, BA.5 n = 13, XBB.1 n = 15, XBB.1.5 n = 13, and EG.5.1 n = 13) and LE-DMT 1MD4 (n = 5 for all variants); CL- 1MD2 (D614 n = 32, BA.1 n = 24, BA.2 n = 32, BA.5 n = 19, XBB.1 n = 23, XBB.1.5 n = 19, and EG.5.1 n = 19) and CL- 1MD4 (n = 2 for all variants); M/HE-DMT 1MD2 (D614 n = 132, BA.1 n = 81, BA.2 n = 134, BA.5 n = 70, XBB.1 n = 80, XBB.1.5 n = 70, and EG.5.1 n = 70) and M/HE-DMT 1MD4 (n = 17 for all variants); IMM-DMT 1MD2 (D614 n = 104, BA.1 n = 70, BA.2 n = 106, BA.5 n = 61, XBB.1 n = 67, XBB.1.5 n = 61, and EG.5.1 n = 61) and IMM-DMT 1MD4 (D614 n = 15, n = 14 for all other variants). (C) Increase in Spike IgG titers against different variants from 1MD2 to 1MD4 in pwMS on LE-DMT, CL, M/HE-DMT, and IMM-DMT, represented as fold change (numbers as in B). (D) Change in breadth scores after 4 doses in pwMS on LE-DMT, M/HE-DMT, IMM-DMT, and CD20-DMT (numbers as in B). Median and IQR are shown in B and C. No statistical analyses were performed on the CL group because n = 2. *p < 0.05, **p < 0.01. 1MD2 = 1 month after second dose (primary vaccination); 1MD4 = 1 month after fourth dose; CD20-DMT = CD20-depleting disease-modifying therapy; CL = cladribine; IMM-DMT = immunosuppressive disease-modifying therapy; LE-DMT = low-efficacy disease-modifying therapy; M/HE-DMT = moderate to high-efficacy disease-modifying therapy.

Figure 3. Comparison of Spike IgG Titers and Protective Neutralizing Antibodies

Comparison of Spike IgG titers against D614, BA.2, XBB.1.5, and EG.5.1 in pwMS based on their neutralization status. Median and 25th and 75th percentiles shown by dotted lines within violin plots. Thresholds of binding/seropositivity (T, green) and protection (Tp, red) are shown by dotted lines. IgG = immunoglobulin G.

Figure 4. Effect of a Fourth Vaccine Dose on Cross-Variant Protection

Classification of pwMS on LE-DMT, CL, M/HE-DMT, IMM-DMT, CD20-DMT, and S1P-DMT into protection categories based on their serostatus and live virus neutralization capacity as determined for each variant according to binding/seropositivity and protection thresholds in Figure 3. **p < 0.001 1MD2 vs 1MD4, ****p < 0.0001 1MD2 vs 1MD4, §p < 0.01 LE-DMT vs CL, †p < 0.0001 M/HE-DMT vs IMM-DMT, #p < 0.0001 LE-DMT vs M/HE-DMT or IMM-DMT or CD20-DMT or S1P-DMT, ƒ p < 0.05 S1P-DMT vs CD20-DMT, ¤ p < 0.001 S1P-DMT vs CD20-DMT, ‡p < 0.05 S1P-DMT vs IMM-DMT. 1MD2 = 1 month after second dose (primary vaccination); CD20-DMT = CD20-depleting disease-modifying therapy; LE-DMT = low-efficacy disease-modifying therapy; CL = cladribine; IMM-DMT = immunosuppressive disease-modifying therapy; M/HE-DMT = moderate to high-efficacy disease-modifying therapy; S1P-DMT = sphingosine-1-phosphate receptor modulator disease-modifying therapy; 1–6MD4 = 1 to 6 months after fourth dose.

Figure 5. Effect of Evusheld in pwMS on CD20-DMT

(A) Comparison of D614 Spike IgG titers at 1MD2 vs 1–6MD4 in pwMS on CD20-DMT who received Evusheld (B) Classification of pwMS on CD20-DMT + Evusheld into protection categories based on their serostatus and live virus neutralization capacity as determined for D614, BA.2, XBB.1.5, EG.5.1. Median and 25th and 75th percentiles are shown by dotted lines within violin plots. Thresholds of binding (T, green) and protection (Tp, red) are shown by dotted lines. ****p < 0.0001 1MD2 vs 1MD4, ‡p < 0.001 CD20-DMT + Evusheld vs CD20-DMT only. 1–6MD4 = 1 to 6 months after fourth dose; CD20-DMT = CD20-depleting disease-modifying therapy; IgG = immunoglobulin G; pwMS = people with multiple sclerosis.

Figure 6. Machine Learning–Based Model for Prediction of Protection/Effectiveness of Vaccination

(A) Ranking of machine learning models for prediction of protection regarding ROC-AUC, PR-AUC, and balanced accuracy. All rankings were assessed on the test (unseen) data set. (B) Feature importance analysis using Histogram Gradient Boosting (HistoGBoost) applied on D614, BA.2, and combined XBB.1.5/EG.5.1 on the test data set. Mean and SD are shown. ADBC = Ada Boost; BagClass = Bagging; ExtraT = Extra Trees; DecisionT = Decision Tree; DMT = disease-modifying therapy; GaussianNB = Gaussian Naive Bayes; GradBoost = Gradient Boosting; HistoGBoost = Histogram Gradient Boosting; KNeighbors = K-nearest neighbors; MLPC = multilevel perceptron; QDA = quadratic discriminant analysis; RandomF = Random Forest; TS1MD2 = time since primary vaccination; TSLD = time since last vaccine dose.