The study was approved by the Italian Medicines Agency (AIFA) and by ethics committee (code 304, 2021) and written informed consent was obtained from all the participants. The inclusion criteria were age ≥18 years, SARS-CoV2 mRNA-based vaccination (fourth dose) and a life expectancy of at least 12 months at the time of vaccine administration. The main exclusion criterion was the presence of a previous laboratory-confirmed SARS-CoV-2 infection (documented by serology and/or molecular test). Moreover, patients experiencing a molecularly confirmed SARS-CoV-2 breakthrough infection (RT-PCR assay) or seroconversion to anti-Nucleocapsid antibody during follow-up were also excluded (n=93).

Anti-spike SARS-CoV-2 humoral (binding and neutralizing antibodies) and T-cell response were monitored before 4^th dose vaccination (pre-4) and 10-20 days after vaccination (post-4) in patients with hematological malignancies (HM, n=71), solid tumors (ST, n=39) and immune-rheumatological (IR, n=25) diseases. Moreover, the immune response was also compared in treatment-specific subgroups.

The response to vaccination was assessed by quantifying the anti-Nucleoprotein-IgG (anti-N) and the anti-RBD-IgG by commercial available diagnostic kit, according to the manufacturer’s protocol (Architect^® i2000sr Abbott Diagnostics, Chicago, IL). As reported in the leaflet, anti-N IgG were expressed as a ratio (S/CO) and values were considered positive when ≥ 1.4. Anti-RBD-IgG were expressed as binding arbitrary units (BAU)/mL and values were considered positive when ≥ 7.1. The anti-N quantification was aimed to identify asymptomatic SARS-CoV-2 infections during the follow-up and to exclude these patients from the analysis. The median level of anti-RBD was calculated using values from all patients (both non-responder and responder patients). To evaluate the functional activity of vaccination-induced antibodies against Omicron, a neutralization assay against BA.2 variant was performed on anti-RBD positive samples according to our previous publication (21). Briefly, heat-inactivated and two-fold serial diluted sera were incubated at 37 °C 5% CO2 for 30 min with equal volumes of 100 TCID50 SARS-CoV-2. Next, 96-well tissue culture plates with sub-confluent Vero E6 cell monolayers were infected with 100 µL/well of virus-serum mixture and incubated at 37 °C and 5% CO2. After 48 h, microplates were observed using a light microscope for the presence of the cytopathic effect (CPE). The highest serum dilution inhibiting at least 90% of the CPE was indicated as the neutralization titer and was expressed as the reciprocal of serum dilution (MNA₉₀). In a subgroup of patients (n=71) whose sample was available after the third dose, the neutralization titer against the Wild Type SARS-CoV-2 strain was also evaluated (22 HM, 10 IR and 39 ST).

The cell-mediated immune response to vaccination was assessed through a standardized whole blood assay (21), shared between clinical centers. Peripheral blood was collected in heparin tubes and stimulated with a pool of peptides spanning the Spike protein (Miltenyi Biotech, Germany) at 37°C (5% CO2). A pool of S protein peptides was generated by combining three peptide pools (all by Miltenyi Biotech): i) PepTivator^® SARS-CoV-2 Prot_S (sequence domains aa 304-338, 421-475, 492-519, 683-707, 741-770, 785-802, and 885–1273 of the Spike protein); ii) the PepTivator^® SARS-CoV-2 Prot_S1 (sequence domain aa 1–692 of the Spike protein) and iii) the PepTivator^® SARS-CoV-2 Prot_S+ (sequence domain aa 689–895 of the Spike protein). The peptide pool was used for cell stimulation at the final concentration of 0.1 ug/ml. The spontaneous cytokines release was calculated in unstimulated culture (background) and a superantigen (SEB) was used as positive control. Plasma was harvested after 16-20 h of stimulation and stored at -80°C. IFN-γ (detection limit 0.17 pg/ml) was quantified in the plasma samples using an automatic ELISA (ELLA, Protein Simple). The results are expressed as the amount of each cytokines after subtracting the background. A concentration equal to or greater than 12 pg/mL of IFN-γ was considered as positive (the cut-off was defined as the mean +/- 2 SEM of 50 anti-S and anti-N negative HCW before vaccination as reported in 4).

Quantitative variables were summarized as median and interquartile range (IQR), while categorical variables were reported as absolute count and percentage. Differences in seroconversion rates across subgroups were analyzed using the chi-square test, and from a multivariable logistic regression model we obtained the odds ratios with their 95% confidence intervals (CI). The model outcome was the seroconversion status after the fourth dose (yes vs no) and independent variables were identified on the basis of availability as required by the study protocol (disease, age, gender, comorbidities and therapy) and used to adjust the vaccination effect on outcome. Current therapy was stratified as follows: for HM patients: Rituximab (RTX), hematopoietic stem cell transplantation (HSCT), chemotherapy (CT), JAK inhibitors (JAK); IR patients: anti-CD20 (Rituximab, RTX), and others (methotrexate, mycophenolate mofetil, and azathioprine); ST patients: chemotherapy (CT), Immunotherapy (ImmTx) and targeted therapy (TarTx).

The Wilcoxon test was used to assess differences between pre-4 and post-4, while Mann-Whitney test was used to assess differences in antibody titers across groups.

The SPSS v.20.0 (IBM) statistical software was used for the analysis.

Between October 2021 and June 2022, 228 patients participating to the VAX4FRAIL study (NCT04848493); received the fourth dose of vaccine. 179 patients received the BNT162b2, 46 the mRNA-1273 vaccine, while for 3 the type of vaccine used was not reported. After having excluded patients resulting positive for antibodies directed to the SARS-CoV-2 nucleocapsid (N) antigen and that presumably had a breakthrough infection after the third or fourth dose, we analyzed the immune response before and after the 4^th dose of mRNA vaccine in the remaining 135 patients.

The breakdown of patients receiving the fourth dose of mRNA COVID-19 vaccine was: hematological malignancies (HM, n=71 – 52.6%), solid tumors (ST, n=39 – 28.9%) and immunorheumatological (IR, n=25 – 18.5%) diseases ( Table 1 ). The median age was 63 years (interquartile range 56-71) and 74 patients (54.8%) were women ( Table 1 ).

We first evaluated whether the fourth dose of vaccine was well tolerated. We found that the fourth dose was in general well tolerated and the safety profile in line with previous doses (22) ( Table 2 ). The most common moderate adverse events (AE) were pain at the site of injection (55%), fatigue (42.2%), pain in the bones (36.7%) and headache (26.6%). Very few patients had severe AE which were mostly pain in the bones (9.2%), fatigue (7.3%) and pain at the site of injection (6.4%) as shown in Table 2 .

All patient groups displayed a statistically significant improvement of the humoral response after the fourth dose of vaccine ( Figure 1A , P<0.0001 for all the groups). However, the antibody response to the spike RBD protein differed in the disease groups with patients with solid tumors responding better (p<0.001 vs IR and p<0.05 vs HM) followed by hematologic malignancies (p<0.005 vs IR and p<0.05 vs ST) and immunorheumatological diseases (p<0.001 vs ST and p<0.005 vs HM) without major discrepancies when analyzing the patients as a group ( Figure 1A ) or individually ( Supplementary Figure 1A ). In the HM group only 2 out of 21 (9%) negative patients and in the IR group 3 out of 12 (25%) patients who had not responded to the third dose responded to the fourth dose.”. Altogether, 17.6% of the patients not responding to the third dose became positive after the fourth dose.

Similarly, the spike-specific T-cell response (defined as IFN-γ levels ≥12 pg/mL) was statistically significantly improved in all of the disease groups after the fourth dose ( Figure 1B P<0.0001 for all the groups). Notably, differently from the B-cell response, the T-cell response was very similar across disease groups (no statistically significant differences among groups, Figure 1B ), suggesting that the T-cell response is less dependent on disease or treatment characteristics.

Patients with fragility have been the first to receive a fourth dose of vaccine. The new version of the vaccine targeting omicron VOC was not available at the time of vaccination. Thus, an important question that arises is whether these patients should undergo a booster dose with the adapted versions of the vaccines to the omicron VOC. We therefore evaluated the neutralizing activity of anti-RBD positive sera against SARS-CoV-2 omicron (BA.2) infectivity in a BSL-3 facility. Sera from all anti-RBD positive patients demonstrated a significantly increased neutralizing activity after the fourth dose of vaccine towards the omicron VOC BA.2 ( Figure 2A , P<0.0001 for HM and ST, P= 0.008 for IR). Of note, the neutralization titer after the fourth dose reached higher level in HM and ST than in IR patients ( Figure 2A , HM vs IR, P<0.05; ST vs IR, P<0.001). Similarly, to the antibody response, the percentage of anti-RBD positive patients showing a detectable neutralizing activity after the fourth dose of vaccine was higher within ST (97.4%) and HM (80.7%) and lower within IR (66.6%) patients. Interestingly, regarding IR patients, only a minority of them displayed a level of neutralization above baseline for the omicron BA.2 VOC. In order to evaluate whether the low level of neutralization titer was evident only towards the omicron VOC or also against the WT SARS-CoV-2, for a reduced number of patients (67 HM, 24 IR and 39 ST), we compared the difference in neutralizing abilities towards the WT and Omicron variant before and after the fourth dose. As shown in Figure 2B mirroring the increased titer of antibodies induced after vaccination in ST patients, there was a very high level of neutralizing antibodies towards both Omicron BA.2 and WT, with a statistically significantly higher level towards the WT virus (p<0.001). By contrast in HM and even more so in IR patients, the level of neutralization towards both the Omicron BA.2 and WT virus was very low ( Figure 2B ). 9% of patients with anti-RBD antibodies did not have neutralizing antibodies to either forms of the virus, while an additional 5.7% did not have neutralizing antibodies only to Omicron BA.2 variant ( Supplementary Figure 2A ), making these patients particularly vulnerable to SARS-CoV-2 infection. We supposed that the probability of finding neutralizing antibodies to both forms of the virus depended on the amount of anti-RBD antibodies elicited during the vaccination. Thus, we analyzed the correlation between the level of anti-RBD antibody response and neutralizing antibodies to WT or Omicron BA.2. As shown in Supplementary Figure 2B there was a positive correlation between anti-RBD antibodies and neutralizing antibodies to WT (p<0.0001, R^{2 =} 0.85) or Omicron (p<0.0001, R^{2 =} 0.69). In particular, anti-RBD levels below 100 BAU/mL correlated with no neutralizing antibodies to WT, while the level of anti-RBD correlating with the presence of neutralizing antibodies towards Omicron BA.2 was more difficult to identify. Nevertheless, a level of antibodies above 350 BAU/mL increased the chances of finding neutralizing anti Omicron BA.2 antibodies (86/89, 96.6%), while at levels above 1000 BAU/mL we found only one patient with still no neutralizing antibodies to Omicron BA.2 ( Supplementary Figure 2B ). There was also a clear correlation between neutralizing antibodies to WT and Omicron BA.2, suggesting that they may broadly recognize common moieties of the viruses ( Supplementary Figure 2B ). Thus, even though patients may have detectable anti-RBD antibodies the probability of finding neutralizing antibodies to Omicron BA.2 is low and correlates on the amount of anti-RBD antibodies.

We then asked if the fourth dose may equally boost the response to the WT and Omicron BA.2 VOC and evaluated the increment of neutralizing antibodies after the 3^rd and 4^th dose of vaccine towards the two viruses. As shown in Figure 2C , we found that the fourth dose was inducing a statistically significant (p<0.005) similar increment of neutralizing antibodies towards both viruses (fold increase: 3.29 ± 3.776 for WT and 3.053 ± 3.61 for BA.2). Thus, patients with ID and HM have a reduced neutralizing antibody response to SARS-CoV-2 independent on the variant. Whenever present, this neutralizing response is similarly boosted towards both forms of the virus across different booster doses.

HM and IR patients often undergo anti-B cell treatment (Rituximab) which can have a major impact on the humoral B cell response to COVID-19 vaccines, as we have shown in this (4) and other cohorts of patients (23). Thus, humoral and T-cell responses were evaluated in relation to the received or ongoing treatment and its presumed immunosuppression ( Table 3 ). It is clear that B-cell directed treatment has an impact on the anti-RBD response ( Supplementary Figure 3A ) but not on the T-cell response ( Supplementary Figure 3B ). More in details, in Figure 3 , we analyzed the B- and T-cell response to the fourth dose in relation to underlying disease and treatment. It is clear that both in HM ( Figure 3A , left panel) and IR ( Figure 3B , left panel) the only treatment which strongly affected the B-cell humoral response was anti-CD20 treatment (Rituximab). Patients with HM undergoing anti-CD20 treatment had a reduced anti-RBD response as compared with HSCT (p<0.001), CT (p<0.05) and JAK inhibitors (p<0.01) ( Figure 3A ). Patients with IR had a reduced anti-RBD response as compared with the other treatments (p<0.001) ( Figure 3B ). By contrast, the different treatments used in solid tumors (ST) had no effect on the induction of the humoral response. As expected, the T cell response to the Spike protein evaluated as IFN-γ production was not affected by anti-CD20 or any other treatment, independent on the pathology of the patients ( Figure 3 right panels) Hence, the antibody immune response to the vaccine is dependent mostly on the patients treatment rather than their pathology.

During the course of the study, the policy of vaccination has changed in Italy and only some of the centers of the VAX4FRAIL study protocol could administer the fourth dose. Thus, we could not follow-up with the population of patients with neurological disorders and some of the patients with the other pathologies. Still we had enough patients to draw conclusions from the fourth dose. We have tested the neutralization to BA.2 and not BA.4 or BA.5 Omicron VOC. Hence, the conclusions are not generalizable to the most recent VOC which seem to escape the immune response elicited by the WT vaccine (33, 34).