Microbiota dynamics and prediction of aGvHD post-allo-HSCT

Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is a curative treatment for several hematologic (and non-hematologic) diseases, though it carries the risk of complications that are potentially fatal, such as infection and graft-versus-host disease (GvHD).¹

Acute GvHD (aGvHD) is a common complication of allo-HSCT that is characterized by diarrhea, abdominal pain, anorexia, hyperbilirubinemia, and maculopapular rash.¹ This clinical syndrome occurs due to inflammatory cytotoxic activity against healthy host tissue—including the gut—by donor T cells and is influenced by the donor source and conditioning regimen.¹ The first step in the pathogenesis of aGvHD is thought to be increased permeability of the intestines caused by the conditioning regimen, leading to extraintestinal movement of bacteria and resulting in bacteremia.²

Loss of diversity of the intestinal microbiota (IM) has been associated with both the risk and intensity of aGvHD as well as aGvHD-related mortality, and data have suggested that the influence of certain bacterial species following allo-HSCT may promote the development of aGvHD.¹ Bacteria from the oral cavity have also been shown to translocate to the gut, driving IM dysbiosis.²

Heidrich and colleagues recently published a study that aimed to directly evaluate the effect of allo-HSCT on the oral microbiota (OM) and the influence of OM dysbiosis on the risk of aGvHD as a way to further understand how the bacterial compositions of the OM and IM impact the development of aGvHD in the post-allo-HSCT setting.¹ Similarly, Ingham and colleagues have just reported their work on long-term microbiota dynamics of the gut, oral, and nasal cavities in pediatric patients who had undergone allo-HSCT.²

Study by Heidrich et al.¹

Study design

Thirty patients who underwent allo-HSCT for hematologic disorders from a single center were consecutively enrolled, and supragingival biofilm samples were collected from all patients at three phases: at preconditioning, at aplasia, and at engraftment. Bacterial cells were recovered, and DNA extraction and sequencing were performed.

Cumulative incidence (CMI) rates were calculated for aGvHD (Grade II‒IV) and severe aGvHD (Grade III–IV) with death as a competing event, and relative risks for developing aGvHD and severe aGvHD were estimated and adjusted for graft source and intensity of the conditioning regimen.

Baseline characteristics

Most patients received reduced-intensity conditioning (60%) and grafts from peripheral blood (67%), and the most common underlying disease was acute leukemia (60%). Clinical characteristics are shown in Table 1.

Table 1. Clinical characteristics*

Results

Dental biofilm microbiota dysbiosis during allo-HSCT

Dental biofilm microbiota (DBM) alpha diversity was assessed using the Shannon index, and a statistically significant decrease in DBM alpha diversity was observed during allo-HSCT:

• Engraftment samples had the lowest overall bacterial diversity (median, 2.75) compared with preconditioning (median, 4.15) and aplasia (median 3.39) samples (p < 0.001).
• Notable changes in DBM general composition were seen for all patients during allo-HSCT; while there was a high average relative abundance of several commensal organisms (including Streptococcus, Veillonella, Actinomyces, and Capnocytophagia species), their average relative abundance decreased during allo-HSCT.
• There was an increase in the average relative abundance of potentially pathogenic species such as Enterococcus and Lactobacillus.
• The most statistically significant differences in abundance were observed for Enterococcus, Lactobacillus, and Mycoplasma.

DBM diversity and aGvHD risk

Median alpha diversity value was used to stratify patients into equal-sized high- and low-diversity cohorts to evaluate the association between DBM diversity and aGvHD risk. No association was found between DBM diversity and the risk of aGvHD at preconditioning, aplasia, or engraftment.

DBM composition and aGvHD risk

Only genera present at relative abundance ≥0.1% in at least 25% of samples were considered in the evaluation of whether abundance of specific genera was associated with risk of aGvHD at any of the three time points. Patients were then stratified into equal-sized high- and-low abundance cohorts.

Patients with high Veillonella relative abundance at preconditioning had a lower CMI of aGvHD, which was significant even after adjusting for graft source and intensity of conditioning regimen. Patients with high Streptococcus or Corynebacterium relative abundance at preconditioning had a higher CMI of aGvHD, though only Streptococcus remained significantly associated with aGvHD risk after adjusting for graft source and intensity of the conditioning regimen (Table 2).

Table 2. Risk analyses for the association of acute graft-versus-host disease with relevant microbiota variables*

Veillonella and Streptococcus had the highest relative abundance at preconditioning, and patients with a Veillonella/Streptococcus ratio ≥1 at preconditioning had a lower CMI of aGvHD (p = 0.004). Notably, the association between this ratio and aGvHD risk was stronger than the association seen for each genus separately, and it remained significant after adjusting for graft source and conditioning regimen intensity (p = 0.005). There was no risk between the Veillonella/Streptococcus ratio at aplasia or engraftment and aGvHD risk.

Enterococcus faecalis bloom and aGvHD risk

The final analysis conducted by the investigators was to determine whether bloom—defined as the sudden expansion of a genus from near absence to dominance—of potentially pathogenic bacteria seen during allo-HSCT was associated with aGvHD risk. During allo-HSCT, 23 blooms involving 12 genera were observed, affecting 20 patients, 3 of whom experienced more than one blooming event.

• Patients experiencing any genus bloom did not have altered aGvHD risk.
• Enterococcus bloom was the most frequent event (observed in 20% of patients).
  • Enterococcus bloom was attributed to Enterococcus faecalis in all but one patient; there was no association between E. faecalis bloom and antibiotic usage. 
  • All patients who experienced E. faecalis bloom developed aGvHD, and E. faecalis bloom was significantly associated with higher CMI of aGvHD (p = 0.0007) and severe aGvHD (p = 0.014).

Study by Ingham et al.²

Study design

The study included 29 pediatric patients at a single hospital center. Each patient underwent a myeloablative conditioning regimen prior to allo-HSCT, and the patients were grouped into categories based on conditioning regimen:

• Total-body irradiation + cyclophosphamide or total-body irradiation + etoposide (n = 6)
• Combinations of busulfan, cyclophosphamide, etoposide, and melphalan (n = 6)
• Fludarabine + busulfan + thiotepa, fludarabine + treosulfan + thiotepa, fludarabine + thiotepa, or fludarabine + cyclophosphamide + thiotepa (n = 12)
• Fludarabine + busulfan, fludarabine + cyclophosphamide, or fludarabine + treosulfan (n = 5)

Sampling timepoints were pre-examination, around the start of conditioning, at the time of allo-HSCT, and weekly during the first 3 weeks after transplantation. Fecal samples (212), buccal swabs (248), and anterior naris swabs (249) were collected from patients at each sampling timepoint for a total of 709 patient samples. Severity of aGvHD was grouped into Grades 0‒I and II‒IV and was graded via daily clinical assessment of skin, liver, and gastrointestinal manifestations.

Baseline characteristics

• The median age of the 29 patients in the study was 8.2 years (range, 2.5–16.4 years) at the time of transplant.
• There were 31% patients with no or mild aGvHD (Grade 0 or I) and 69% with moderate to severe aGvHD (Grade II‒IV) at median +14 days after HSCT.
• Skin was involved in all patients who experienced aGvHD, while three had intestinal tract involvement and two had liver involvement.
• During the follow-up period of 21.4 months on average (range, 10.1–32.7 months),
  • two patients experienced disease progression;
  • one patient underwent donor lymphocyte transfusion; and
  • three patients died (one relapse-related death and two treatment-related deaths).
• All patients were treated prophylactically with trimethoprim and sulfamethoxazole prior to HSCT.
• Patients with fever or clinical signs of infection were treated with meropenem (n = 28), vancomycin (n = 24), ciprofloxacin (n = 20), phenoxymethylpenicillin (n = 14), or other antibiotics, depending on clinical presentation and culture results.

Results

Bacterial alpha diversity

• Bacterial alpha diversity was highest overall in the gut, followed by the oral cavity and the nose.
• The lowest alpha diversity for all three body sites was observed within the first month following HSCT.
• The decrease in microbial diversity was significant for the nasal cavity, decreasing from 4.43 at the start of conditioning to 2.65 in Week +1 (p = 0.02).
  • While alpha diversity increased again at all body sites thereafter, it was lower again at Month 12 in the nasal cavity.

Microbial community composition in patients prior to HSCT vs healthy controls

The gut microbiota of the patients at pre-examination was compared with age-matched healthy children and demonstrated an alpha diversity 2.4-fold lower in the patient group, potentially due to treatment given to these children prior to enrollment in the study. Bacterial composition also differed between the two groups, with several taxa that were significantly more abundant in the patient group compared with healthy controls, including Lactobacillus, Enterococcus, Erysipelotrichaceae, and Klebsiella. Contrastingly, Prevotella, Ruminococcus, and Akkermansia were more abundant in the healthy controls.

Temporal microbial community dynamics

A tree-based sparse linear discriminant analysis (LDA) was performed to characterize samples from the gut, oral, and nasal sites at different time points, identifying three partly entwined phases:

• Phase I included samples from pre-examination and the start of conditioning.
• Phase II included samples from the day of HSCT to Month +1.
• Phase III included samples from Month +3 to Month +12.

Samples from the oral and nasal cavities in phases I and III overlapped, which suggested that microbial communities from later timepoints had possibly returned to a state similar to that prior to HSCT. Interestingly, the nasal community composition at Month +12 was different from the composition from the nasal cavity at Week +1, though both had low alpha diversity.

The 12 most abundant families at each site were examined to provide a more detailed view of the abundance dynamics:

• In the gut, there was a reduction in Lachnospiraceae in phase II, immediately post-HSCT, from 13% at pre-examination to 4.7% in Week +1, followed by recovery to 27.5% in Month +3 at the start of phase III.
• In the oral cavity, there was a reduction in the relative abundance of Actinomycetaceae at several timepoints in phase II compared with phase I and later follow-up timepoints, with abundances of 9.7% at pre-examination and 2.9% at Week +3.
• Additionally, Streptococcaceae abundances were lower from the day of HSCT to Week +2 compared with pre-HSCT and late follow-up timepoints (44.6% at pre-examination, 23.3% at Week +1, and 51.3% at Month +3).
• In the nasal cavity, there was a reduced relative abundance of Corynebacteriaceae and Moraxellaceae at most timepoints in phase II compared with samples from phases I and III; Corynebacteriaceae abundances were 28.7% at pre- examination and 0.7% in Week +1.

Distinct bacterial lineages

Individual discriminating amplicon sequencing variants (ASVs) were examined to determine which specific gut taxa drove the differences between samples seen in the LDA. This analysis revealed 19 clades in the gut that best separated the samples based on timepoint. The two most discriminating clades with positive LDA coefficients were Enterococcaceae and Lactobacillaceae, which increased in abundance from the day of HSCT (Enterococcaceae) and Week +1 (Lactobacillaceae); each clade decreased in abundance from Month +3 to levels similar to pre-examination (Table 3). The two most discriminative clades with negative LDA coefficients were two individual ASVs, one Lachnospiraceae clade, and two Ruminococcaceae clades; the abundance of these clades decreased in Week +1 and recovered after Month +3. Enterococcaceae were more abundant in phase II samples, and Lachnospiraceae and Ruminococcaceae were more abundant in phase I and III samples.

Table 3. Distinct discriminatory lineages*

Ten clades of 71 total ASVs were identified in the oral cavity, and 30 discriminating nasal clades of 36 total ASVs were revealed in the nasal cavity (Table 3).

Severity of aGvHD predicted by microbiota

Patients with Grades 0–I aGvHD had lower relative abundances of Tannerellaceae in the gut pre-HSCT compared with patients with Grades II-IV aGvHD, particularly at pre-examination and start of conditioning. In addition, three predictive ASVs were identified in the gut demonstrating that high abundances of ASV 128 (Parabacteroides distasonis, Tanerellaceae, p < 0.01), ASV 268 (Lachnospiraceae NK4A136 group sp., Lachnospiraceae, p = 0.01) and ASV 3 (Lactobacillus sp., Lactobacillaceae, p < 0.01) pre-HSCT were associated with development of aGvHD Grades II–IV.

The bacterial community of the oral cavity pre-HSCT in patients with Grades II–IV aGvHD was characterized by a lower relative abundance of Neisseriaceae and higher relative abundances of Aerococcaceae and Prevotellaceae when compared with Grades 0‒I aGvHD, particularly at the pre-examination and start of conditioning timepoints. High abundances of the ASVs 568 (Actinomyces sp., Actinomycetaceae), 226 (Prevotella melaninogenica, Prevotellaceae), and 500 (Pseudoproprionibacterium propioninum, Proprionobacteriaecae) (all p < 0.001) pretransplantation predicted the development of aGvHD Grades II–IV after HSCT.

In the nasal cavity, the proportion of Neisseriaceae pretransplant was higher in patients with aGvHD Grades 0–I compared with Grades II–IV, while Actinomycetaceae and Corynebacteriaceae were more abundant in patients with aGvHD Grades II–IV compared with Grades 0–I. Regarding specific ASVs, the partial 16S rRNA gene sequence of ASV 66 (with a high sequence similarity to Actinomyces viscosus) predicted development of Grades II–IV aGvHD (p = 0.03) when occurring in abundance pre-HSCT. In addition, pre-HSCT levels of this organism were 2.3-times higher in patients with aGvHD Grades II–IV compared with Grades 0–I. High pre-HSCT abundance of the partial 16S rRNA gene sequence of ASV 47 (exhibiting a high sequence similarity to Rothia aeria), on the other hand, predicted that a patient would not develop aGvHD (p = 0.03).

T-cell reconstitution

• In the gut, high numbers of CD4⁺ T cells and T_H17 cells were associated with high abundances of Lachnospiraceae, Ruminococcaceae, and Lactobacillaceae ASVs.
• In the oral cavity, the same T cell subsets were positively associated with specific Flavobacteriaceae, Prevotellaceae, Veillonellaceae, and Neisseriaceae ASVs.
• In the nasal cavity, it was primarily Veillonellaceae that was associated with high T cell counts.

B-cell reconstitution

• In the gut, high B cell counts were positively correlated with high abundances of Ruminococcaceae, Lachnospiraceae and Rikenellaceae ASVs, and few Veillonellaceae and Lactobacillaceae ASVs.
• In the oral cavity, Actinomyces odontolyticus and Veillonella parvula ASVs were positively correlated with high B cell counts, and Staphylococcaceae and Lactobacillaceae ASVs showed negative correlations.
• In the nasal cavity, Streptococcaceae, Moraxellaceae, and Corynebacteriaceae ASVs were positively correlated with high B cells counts, particularly in Month +3.

Discussion

aGvHD is a significant cause of mortality after allo-HSCT, and the current first-line therapy—corticosteroids—has a response rate ranging from 40% to 70%, highlighting the importance of being able to predict aGvHD risk and develop preventive therapy.¹ Heidrich et al. and Ingham et al. have identified that changes in the oral, nasal, and gut microbiomes, and their immune environments prior to and during allo-HSCT, may be predictive of aGvHD after transplantation, though it is important to note that these were single-centered studies with limited sample sizes. These studies indicate that future research may allow for the development of more precise treatment strategies in this area.