Patient A was a 53-year-old patient with onco-hematological disease with prolonged infection lasting for 51 days (Fig. 1).

Evolution of prolonged SARS-CoV-2 infection in Patient A. The blue line represents the trend of viral load based on Ct values, while black points indicate sampling days. The dotted and dashed black lines mark the thresholds for which the swab was considered negative (Ct > 42); samples with 40 < Ct < 42 were considered to have low positivity. The dashed red lines indicate the time points selected for sequencing analysis

Analyses of consensus sequences at three longitudinally collected time points (day 1, day 19 and day 38) indicated that the virus strains belonged to BA.2 Omicron lineage. Day 1 was collected on May 28th, day 19 on June 15th and day 38 on July 4th, 2022.

The consensus sequences from the day 1 sample presented 55 mutations (29 of which were located within the S gene) compared to the Wuhan-1 reference strain, while in the day 19 and day 38 samples, 54 mutations (27 located in the S gene) were detected in comparison to the reference strain. Overall, the vast majority of mutations that occurred throughout day 1 (mean = 99.5; σ= 0.69), day 19 (mean= 99.2;σ  = 0.64) and day 38 (mean= 99.4;  σ= 0.80) did not present high variability over time, with some notable exceptions (Fig. 2). The comparison of consensus sequences over the three time points revealed that two de novo mutations in the RdRP gene, A685S and M794I, emerged at the last time point (day 38) with a frequency of 62.4% and 95.4%, respectively, suggesting a within-host viral evolution attempt. Both the mutations are not common in SARS-CoV-2 Omicron variants [18]. A slight increasing trend in the frequency of R408S in the S gene (25.2% on day 1, 30.6% on day 19 and 31.1% on day 38) and of Q19E in the M gene (80.6% on day 1, 92.0% on day 19 and 91.3% on day 38) was recorded. Both these mutations then emerged as high prevalent in successive Omicron lineages in the global population [18]. Conversely, S99del and F100del mutations in the M gene showed variability in their frequency, with a total absence on day 1 and day 38 and a low frequency on day 19 (11.2% and 10.9%, respectively). K417N and N440K mutations in the S gene, common in several Omicron variants [18], were detected only on day 1 (frequency > 99%) and then disappeared on following days 19 and 38 (Table 1).

Heat map showing mutation frequencies in longitudinally collected Patient A samples. Genes and mutations are reported on the x-axis with “gene: mutation” wording, while time points are on the y-axis. The frequency is displayed through a color gradient ranging from light blue (for low mutations) to dark blue (for high mutations). A Heat map showing that the general representation of the detected mutations allows the discrimination of sites with low variability from those with high variability in mutation frequency. B Mutations presenting greater intrahost variability (MAF > 5% between at least two different time points) throughout three longitudinally collected samples from the same individual

Patient B was 75 years old with onco-hematological disease experiencing prolonged SARS-CoV-2 infection lasting for 146 days. The viral load, very high in the first 20 days and then gradually decreasing, showed a swinging trend (Fig. 3). Two nasopharyngeal samples collected on May 22nd, 2022 (day 1) and on August 18th, 2022 (day 86) were sequenced and analyzed.

Evolution of prolonged SARS-CoV-2 infection in Patient B. The blue line represents the trend of infection based on the E gene Ct values reported in the diagnostic center, while the black points indicate sampling days. The dotted and dashed black lines mark the thresholds for which the swab was considered negative (Ct > 42); samples with 40 < Ct < 42 were considered to have low positivity. The dashed red lines indicate the time points selected for genomic analyses

The consensus sequences produced by NGS allowed the identification of two different Omicron sublineages at two different time points: the BA.2 lineage was assigned to sample from day 1, while the day 86 sample was recognized as the BA.2.1 lineage. These results may indicate the occurrence of an intrahost evolution of SARS-CoV-2. Compared to reference strain sequence, a total of 55 mutations were detected in samples collected on day 1, 31 of which were in the S gene. On the other hand, 61 mutations were detected in day 86 samples, 34 of which were inside the S gene. Overall, the vast majority of mutations that occurred throughout day 1 (mean = 99.6;  σ= 0.42) and day 86 (mean = 99.7;  σ= 0.33) did not present high variability over time, with some notable exceptions. To investigate the temporal evolution of viral populations between the two time points, the frequency of genomic variations during 146 days of infection was analyzed (Fig. 4).

Heatmap showing mutation frequencies in longitudinally collected Patient B samples. Genes and mutations are reported on the x-axis with “gene: mutation” wording, while time points are on the y-axis. The frequency is displayed through a color gradient from light blue for low-frequency mutations to dark blue for high-frequency mutations. A Heat map showing that the general representation of the detected mutations allows the discrimination of sites with low variability from those with high variability in mutation frequency. B Mutations presenting greater intrahost variability (MAF > 5% between at least two different time points) throughout two longitudinally collected samples from the same individual

Compared to the first collected sample (day 1), in the sample collected on day 86, 10 de novo mutations have arisen: T35I in nsp9; Y324C in nsp13; P217H, N440K, K444N, L455F, V642G, and D1153Y in the S gene; T30I in the E gene and A182V in the N gene. For each mutation, the frequency on day 1 was 0%, while the frequency on day 86 was > 80% (Table 2). The mutations Q19E in E protein and R408S in S protein have been shown to increase their frequency over time, from 43.9 to 90.6% and from 24.7 to 32.9, respectively (Table 2).

Four de novo deletions in S protein (F374del, S375del, T376del and Y144del) emerged in the sample collected at day 86 (Table 2).

Most of these mutations and deletions are considered very rare and have not been subsequently seen at high prevalence in the global population [18]. The only mutations that were found at high prevalence in global population were N440K and R408S in S protein together with Q19E in M protein (Table 2) [18].

Furthermore, by comparing the two time points, we also observed the disappearance of four mutations present on day 1 but not on day 86: F694Y in the RdRP gene (from 18.4% to 0%) and S256L, P463R, and S1147L in the S gene. (Table 2).

Both Patient A and Patient B underwent phylogenetic analysis to explore the magnitude of intrahost viral diversity at the reported time points (Fig. 5, Additional files 7, 8).

Phylogenetic tree obtained through the analysis of 201 SARS-CoV-2 sequences collected during 2022, including the considered cases at each time point. Patient A and B time points are highlighted with yellow and light green label background, respectively. Except for the analyzed cases, all samples are reported with an internal system ID, which consists of progressive numbering for each patient’s nasopharyngeal swab (NPS). Clades classification is provided by NextClade consensus sequence analysis report

Every analyzed sample was located in the corresponding Omicron BA.2 clade, as reported in previous results. Phylogenetic analysis revealed within-host viral variability in both Patient A and Patient B. Specifically, there was strong homology between the day 1 and day 19 time points for Patient A, while major diversity emerged only later on day 38. Similarly, day 1 of Patient B was similar to that of Patient A at the first time points (day 1 and day 19), as they were classified as BA.2 Omicron lineage but an important viral evolution was discovered on day 86, which is also considered a SARS-CoV-2 lineage switch from BA.2 to BA.2.1.