Patients.
During the study period between 2019 and 2022, seventeen patients (median age 6.0 years, range 0.7–24.7 yrs) underwent allogeneic HSCT for not malignant diseases (n = 8: 5 acute lymphoblastic leukemia and three acute myeloid leukemia) and malignant diseases (n = 9: 3 bone marrow failure, five congenital immunodeficiencies, and one autoinflammatory disease) (Table 1). Eight patients received stem cells from a haploidentical donor, 5 from an alternative donor, and four from a related donor. A myeloablative (MAC) CR was used in 16 patients, including total body irradiation (TBI) in five patients, busulfan (Bus) in 3 patients, and treosulfan (Treo) in the remaining eight patients. One patient received a reduced-intensity conditioning regimen (RIC) for congenital dyskeratosis. In consideration of the different ages of the patients and then of dental development (primary and secondary dentition), only patients older than two years of age are included in the statistical analysis of this study. In addition, the same cut-off was valid for estimating microbiome composition [16].
Table 1
Characteristics of the transplanted patients
Code | Age (yrs) | Underlying Disease | Type of donor | Conditioning regiemn | O-M | GI-M | Infections | Acute GVHD | Chronic GVHD |
1 | 2.9 | CGD | UD | Treosulfan-FludarabineCampath | 2 | 0 | 0 | 2 | 0 |
2 | 18.8 | CGD | RD | Treosulfan-Fludarabine- ATG | 1 | 0 | 0 | 0 | 0 |
3 | 2.4 | CGD | UD | Treosulfan-FludarabineCampath | 0 | 0 | 0 | 2 | 0 |
4 | 2.6 | TACI-CARD11 deficiency | Haploidentical α/β/CD19 depletion | Treosulfan-Fludarabine-Thiotepa | 0 | 2 | 0 | 0 | 0 |
5 | 10.9 | Hyporegenerative anemia | RD | Treosulfan-Fludarabine- ATG | 2 | 2 | 0 | 3 | 0 |
6 | 14.1 | GATA-2insufficiency | Haploidentical α/β/CD19 depletion | Treosulfan-Fludarabine-hiotepa | 1 | 0 | 0 | 1-Gastric | 0 |
7 | 24.7 | Bone Marrow aplasia | UD | TreosulfanFludarabine-Thiotepa | 0 | 0 | 0 | 0 | 0 |
8 | 2 | Dyskeratosis congenita 2nd HSCT | Haploidentical α/β/CD19 depletion | Fludarabine-Cyclophosphamide Campath | 1 | 0 | HCMV | 0 | 0 |
9 | 4.2 | ALL | Haploidentical CY- post | TBI-Fludarabine-CY -post | 1 | 0 | 0 | 0 | 0 |
10 | 17.3 | AML | UD | TBI-Fludarabine-L-PAM | 4 | 0 | 0 | 3 | 1 |
11 | 13.9 | AML | RD | Busulfan Fludarabine-L-PAM | 4 | 0 | HCMV | 0 | 0 |
12 | 11.2 | ALL | UD | TBI-Etoposide | 3 | 0 | pneumonia | 2 | 0 |
13 | 6.0 | ALL | Haploidentical CY-post | TBI-FludarabineEX-post | 0 | 0 | 0 | 0 | 0 |
14 | 13.9 | ALL | Haploidentical CY-post | TBIFludarabine-EX-post | 1 | 0 | Stenotrophomonas | 2 | 0 |
15 | 2.0 | AML | Haploidentical CY-post | Busulfan –Thiotepa-Fludarabine Cy-post | 0 | 0 | 0 | 3-Gut | 0 |
16 | 0.7 | 1 Mevalonic aciduria | Haploidentical α/β/CD19 depletion | Treosulfan –Thiotepa-Fludarabine | 0 | 0 | 0 | 0 | 0 |
17 | 0.7 | ALL | RD | Busulfan -Thiotepa Fludarabine | 2 | 0 | 0 | 3-Gut | 0 |
Legend: CGD: Chronic granulomatosis disease; ALL: Acute lymphoblastic leukemia: AML: Acute myeloid leukemia) O-M (Oral mucositis), GI-M (Gastro-Intestinal Mucositis), Campath (alemtuzumab), ATG (anti-thymocyteglobulin), RD (Related Donor), UD (Unrelated Donor), HCMV (Human CytoMegaloVirus), TBI (Total Body Irradiation), L-PAM (Melphalan), CY post (cyclophosphamide post tranplant); GvHD = Graft versus Host Disease. Acute or Chronic GVHD (grade-localization, if any) |
In detail, eleven patients developed oral mucositis subdivided in grade 1 (n = 5), grade 2 (n = 3), grade 3 (n = 1), and grade 4 (n = 2). Seven patients who received HSCT for not malignant developed mild oral mucositis (grade 0 in 4, grade 1 in 3, and grade 2 in 2), while patients transplanted for the malignant disease had more severe oral mucositis (grade 0 in 2, grade 1 in 2, grade 2 in 1, grade 3 in 1, grade 4 in 2). Gastro-intestinal (GI) mucositis was observed only in 2 patients (grade 2) that underwent transplants for not malignant disease. Acute GvHD (aGvHD) was observed in 8 patients four of them had aGvHD above grade 2, and 3 with a GI aGvHD (grade 3). Limited chronic GvHD was present in one patient (Table 1).
In terms of nutrition, it is known that there were different recommendations for the use of enteral or parenteral feeding [17]. In our case, fifteen patients received total parenteral nutrition for a median of 26.8 days (12–55 days).
Exploratory analysis of sequencing data.
Ninety-eight oral and fecal microbiomes were analyzed. The mean of the 16S mapped reads per sample obtained from the 98 samples was 194521+/-57929. These reads identified 1524+/-551 operational taxonomic units (OTUs) in the 49 oral swab samples and 1545+/-389 OTUs in the fecal samples analyzed.
The longitudinal analysis of the five principal phyla in the microbial community of oral samples in patients older than 2 years showed that the Firmicutes increase their abundance along the four-time point analyzed (pre-HSCT, at engraftment, + 30 days, and + 100 days after HSCT) from 46–72%. On the contrary, the Proteobacteria and the Bacteroidetes phyla decreased their abundance along the same time points from 20 to 10% and from 17 to 9%, respectively. Accordingly, the analysis of Firmicutes/Bacteroidetes (F/B) or Firmicutes/Proteobacteria (F/P) ratios showed that both increased their value along the four observations, with F/B ranging from 2.7 to 8.0 and F/P from 2.3 to 7.2 (Table 2a). Fusobacteria abundance was stable over the first three time points and decreased to half at 100 days from transplant (Table 2a). On the contrary, fecal microbiomes behaved differently since Firmicutes showed a stable abundance rate, while Bacteroidetes increased and Proteobacteria decreased their fullness along longitudinal points (Table 2b).
Table 2
Longitudinal analysis of relative abundance (%) of major phyla, and analysis of F/B or F/P ratios indicative of dysbiosis.
| a: oral all samples | b: fecal all samples |
Phyla | pre-HSCT [14] | Engraftment [15] | HSCT + 30d [9] | HSCT + 100d [5] | pre-HSCT [15] | Engraftment [14] | HSCT + 30d [9] | HSCT + 100d [5] |
Actinobacteria | 11 | 9 | 7 | 7 | 3 | 4 | 1 | 3 |
Bacteroidetes | 17 | 14 | 11 | 9 | 22 | 19 | 19 | 33 |
Firmicutes | 46 | 55 | 60 | 72 | 43 | 44 | 50 | 44 |
Fusobacteria | 6 | 6 | 5 | 3 | 1 | 2 | 0 | 0 |
Proteobacteria | 20 | 16 | 17 | 10 | 31 | 31 | 29 | 20 |
F/B | 2,7 | 3,9 | 5,5 | 8,0 | 2,0 | 2,3 | 2,6 | 1,3 |
F/P | 2,3 | 3,4 | 3,5 | 7,2 | 1,4 | 1,4 | 1,7 | 2,2 |
Analysis was performed in all samples from oral or fecal microbiome, analyzing samples from patients with malignant or non-malignant pathologies. F/B Firmicutes/Bacteroidetes ratio, F/P Firmicutes/Proteobacteria ratio, In squared parenthesis were indicated the number of samples for each time point. |
Alpha and beta diversity indexes.
The alpha diversity profiling indexes used to analyze the microbial community in oral samples were the Chao1, Shannon, or Simpson indexes. The first index demonstrated a correlation between microbial flora richness, while the other indexes gave information about community richness and evenness. All the indexes showed in the oral swabs the lowest values at the HSCT engraftment, while they increased gradually with the highest values after 100 days. All alpha indexes analyzed reached statistical significance in oral swabs samples (Fig. 1a), but only the Chao1 reached significance in fecal samples (Fig. 1c).
All three indexes in samples before transplant showed a lower value in patients affected by malignant disease compared to the specimens from patients with non-malignant diseases. The fecal samples behaved similarly but always showed lower values when compared to the oral microbiome (data not shown). The analysis performed at engraftment showed a similar trend with a single exception. The value of the Chao1 index of oral microbiome resulted higher for samples from malignant diseases (data not shown).
Beta diversity was analyzed using the Bray-Curtis index and PERMANOVA statistical evaluation without reaching significance either in oral or in fecal samples (Fig. 1b, 1d).
Longitudinal analysis of microbiome taxonomic composition.
The analysis of microbiome at engraftment compared to the one before HSCT, showed that the families of Corynebacteriaceae and Peptostreptococcaceae, the genus Abiotrophia, and the species Streptococcus mitis were more significantly abundant at engraftment in both oral and fecal samples (Supplementary material Table S1). Different Prevotella and Veillonella species, were found to be significantly more abundant at engraftment in oral samples only (Supplementary material Table S1). In addition, some species of Prevotella and the Campylobacter gracilis were among the taxa that increased abundance over time passed from HSCT in oral microbiomes. (Supplementary material Table S2, Supplementary material Table S3).
Different species belonging to the Actinobacteria phylum, like Atopobium parvulum, Actinomyces odontolyticus, Actinomyces sp., and Rothia mucilaginosa, as well as Gemella haemolysans, Streptococcus infantis, Streptococcus pneumoniae all three belonging to the Firmicutes, Fusobacterium nucleatum (member of Fusobacteria) and the genus Escherichia belonging to Proteobacteria were instead more abundant in fecal samples at engraftment in the same comparison as above (Supplementary material Table S1).
Streptococcus genus increased abundance after transplant (+ 30d), both in oral and fecal samples. In addition, different Streptococcus species were more abundant 30 days from transplant in fecal samples, except for Streptococcus oralis, showing a preference for oral specimens (Supplementary material Table S2). In oral samples, Gemella morbillorum, Veillonella alcalescens, Peptostreptococcaceae, and Serratia were described to be more abundant both at the engraftment and 30 days after HSCT compared to specimens before transplant (Supplementary material Table S1 and Supplementary material Table S2). Gemella haemolysans and, to a lesser extent, Rothia mucilaginosa showed increased abundance in fecal samples at the engraftment and a further increase 30 days after HSCT (Supplementary material Table S1 and Supplementary material Table S2). Thirty days after HSCT, different Streptococcus species showed higher abundance compared to specimens collected before HSCT, but none resulted in an increasing amount compared to the engraftment.
One hundred days after transplant, oral samples showed a higher statistical relevance and higher abundances of Prevotella aurantiaca, Campylobacter gracilis and Lachnoanaerobaculum, (Supplementary material Table S3). At the same time point, the fecal samples showed no increase in specific taxa (Supplementary material Table S3). It is to be stressed that in the comparison between samples after 100 days after HSCT and before it, oral samples showed a higher abundance of Fusobacterium periodonticum and Enterobacteriaceae in the samples before HSCT by three different algorithms (Supplementary material Table S3).
The same was found in fecal specimens for Abiotrophia defectiva, Enterococcus faecalis, and the microbial genus Gemella that were typically found again on pre-HSCT isolates (Supplementary material Table S3).
Microbial flora associated with oral mucositis.
An equal number of patients with malignant and non-malignant diseases suffered complications like oral mucositis (OM), but the grade of mucositis severity was slightly higher in subjects with the malignant disease compared to non-malignant. The analysis of phyla-abundances at engraftment showed the progressively increasing expansion of the Fusobacteria was related to more severe oral mucositis. Indeed, samples from patients that did not show oral mucositis had only 2% of Fusobacteria. Their abundance increased in OM-positive specimens from 8–18%, with the highest percentage associated with more severe forms (grade > = 3) of mucositis (Fig. 2). Longitudinal analysis of differential abundances in oral samples with mucositis (all grades) showed that Mycoplasmatales descendants and the Prevotella oris were among the taxa with the most statistically significant increase in specimens at engraftment or after 30 days from HSCT, respectively (Supplementary material Table S4). Moreover, in the oral microbiome, Mycoplasma (at engraftment) and Prevotella oris (after 30 days from transplant) were associated with the best statistical correlation with grade ≥ 2 oral mucositis in oral swabs samples (Supplementary material Table S5). On the contrary, Ruminococcaceae, Rothia, and the Streptococcus genus in oral swab samples were found relevant, from a statistical point of view, in patients that never reported oral mucositis, so to be considered as protective taxa (Supplementary material Table S4, Supplementary material Table S5).
On the contrary, the analysis of fecal samples showed Egghertella lenta, Bacteroides ovatus, Ruminococcus gnavus, Lachnoclostridium lavalense, Pseudoflavonifractor, and different species of Clostridium always associated with specimens of reporting oral mucositis (Supplementary material Table S4, Supplementary material Table S5).
Networks of interaction among microbial taxa in oral mucositis.
The network correlation between the taxa abundance and oral mucositis was also studied in the oral microbiome using sparse correlation for compositional data (SparCC) analysis. The MDI for each informative comparison was also colculated. Thus, in patients undergoing the myeloablative (MAC) regimen, the MD-Index at engraftment of cases of patients developing mucositis of any complexity, compared to samples from patients that never developed mucositis analyzed at the level of genera was 0.2469 (Fig. 3a) and 0.4912, the one associated with mucositis with more complexity (grades > = 2). These data indicated that an increased MD-Index value correlates with more complex mucositis. The SparCC analysis of microbial flora associated with oral mucositis of any grade indicated that 29 genera produce positive/negative correlations (threshold 0.3, p < = 0.05). Among these genera, two third (at engraftment) and approximately three-quarters of 30 genera (after 30 days from HSCT, MD-Index 0.9946) showed positive correlations among them. Overall, the Prevotella genus showed the highest number of correlations with other taxa (seven), computed by the SparCC network analysis (Fig. 3a). The Capnocytophaga showed six correlations all with positive sign. Interestingly, the species Atopobium parvulum and Capnocytophaga sputigena confirmed their association with mucositis (Fig. 3b), as already suggested in patients treated with radiotherapy for haematologic malignancies [18, 19]. More, the analysis of the microbial species associated with oral mucositis (grades > = 2) in patients subjected to MAC regimen showed a network of 178 species (threshold of 0.3 and a p < = 0.05), 106 of them producing positive correlations and the MD-Index of cases over controls after 30 days from transplant was 0.6979 (Fig. 3b). Streptococcus, Prevotella, and Veillonella were among the genera showing the highest correlation numbers with other taxa in the network analysis after 30 days from transplant. In addition, twelve, ten, and eight were the species found to belong to these genera. Thus, indicating their importance in the association with oral mucositis. Overall, these network analyses on oral microbiome pinpointed that Prevotella correlates positively with oral mucositis samples, while Streptococcus was always associated with a negative correlation.
Acute GVHD.
Regarding aGvHD, the longitudinal analysis of Chao1, Shannon, and Simpson's alpha diversity indexes showed higher median values on aGvHD samples (Fig. 4). Simpson index was the only measure to reach statistical significance, at least in the pre-HSCT specimens (p = 0.0350) (Fig. 4). At engraftment the compositional microbiome analysis showed a sturdy increase in the relative abundance of Proteobacteria in patients with a-GvHD from 17–49% (Fig. 5). More in detail, the Firmicutes/Proteobacteria ratio was 3.2 in samples from patients without cutaneous aGvHD and 0.5 in cutaneous aGvHD patients, thus indicating an unbalanced increase of Proteobacteria compared to the Firmicutes in aGvHD patients. Differently from the patients who developed mucositis, the fecal microbiome of subjects with cutaneous aGvHD was associated with the Proteobacteria phylum (Fig. 5). This data might indicate either a direct dependency of the showed microbial family with aGvHD or an association with pro-inflammatory taxa, as suggested by others [20]. More in detail, different genera and species showed a positive correlation with cutaneous aGvHD in stool samples, like some species of Enterobacteriaceae like Klebsiella, Kluyvera, Yersinia, Serratia, and Enterobacter, as it was also confirmed by the network correlation analysis performed with SPARCC algorithms (Fig. 6). It is to be stressed that only a relatively small number of microbial taxonomies showed differences in the oral microbiome compared to the results of fecal samples (data not shown). At 30 days after HSCT, among the bacterial taxonomy with higher abundance in stool samples from patients suffering from cutaneous aGvHD, different of them belong to the Alphaproteobacteria (Gemminger formicilis) and Deltaproteobacteria (Bilophila wadsworthia) (Fig. 5 and Supplementary material Table S6), followed by the ones belonging to the Bacteroides genus (Supplementary material Table S6). Indeed, the Firmicutes/Bacteroidetes ratio always showed a lower value in patients with cutaneous aGvHD compared to the ones that did not develop it at all the time points studied (Fig. 5).