Article Text

Unveiling contrasts in microbiota response: A1c control improves dysbiosis in low-A1c T2DM, but fails in high-A1c cases—a key to metabolic memory?
  1. Thiago Fraga Napoli1,2,
  2. Ramon V Cortez3,
  3. Luiz Gustavo Sparvoli3,
  4. Carla R Taddei3,
  5. Joao Eduardo Nunes Salles2
  1. 1Serviço de Endocrinologia e Metabologia, Hospital Servidor Público Estadual de São Paulo, São Paulo, São Paulo, Brazil
  2. 2Departamento de Clínica Médica, Disciplina de Endocrinologia, Faculdade de Ciências Médicas da Santa Casa de São Paulo, São Paulo, Brazil
  3. 3Department of Clinical Analysis and Toxicology, University of Sao Paulo, Sao Paulo, Brazil
  1. Correspondence to Dr Thiago Fraga Napoli; napoli38{at}gmail.com

Abstract

Introduction Type 2 diabetes mellitus (T2DM) is associated with dysbiosis in the gut microbiota (MB). Individually, each medication appears to partially correct this. However, there are no studies on the response of the MB to changes in A1c. Therefore, we investigated the MB’s response to intensive glycemic control.

Research design and methods We studied two groups of patients with uncontrolled T2DM, one group with an A1c <9% (18 patients—G1) and another group with an A1c >9% (13 patients—G2), aiming for at least a 1% reduction in A1c. We collected A1c and fecal samples at baseline, 6, and 12 months. G1 achieved an average A1c reduction of 1.1%, while G2 a reduction of 3.13%.

Results G1’s microbiota saw a decrease in Erysipelotrichaceae_UCG_003 and in Mollicutes order (both linked to metabolic syndrome and associated comorbidities). G2, despite having a more significant reduction in A1c, experienced an increase in the proinflammatory bacteria Megasphaera and Acidaminococcus, and only one beneficial genus, Phascolarctobacterium, increased, producer of butyrate.

Conclusion Despite a notable A1c outcome, G2 could not restore its MB. This seeming resistance to change, leading to a persistent inflammation component found in G2, might be part of the “metabolic memory” in T2DM.

  • diabetes mellitus, type 2
  • metabolism

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

WHAT IS ALREADY KNOWN ON THIS TOPIC?

  • Gut microbiota (MB) may be influenced by the presence or absence of dysglycemia.

  • However, the extent to which glycemic control, as proposed by clinical guidelines, affects MB across different glycemic ranges required investigation.

WHAT THIS STUDY ADDS?

  • This research reveals that improvement in MB following glycemic control in patients with an A1c >9% may not occur, or there could even be an increase in certain deleterious bacterial genera.

  • The persistence or exacerbation of inflammatory genera might be a source of metabolic memory.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • Addressing MB is complex and remains currently uncertain.

  • Understanding MB’s resistance to interventions could lead to the development of future treatment strategies tailored for these patients.

Introduction

The relationship between type 2 diabetes mellitus (T2DM) and the gut microbiota (MB) is complex and involves various possible mechanisms that could affect the course of the disease. One such mechanism is the potential for bacteria to potentially ameliorate T2DM, such as the production of short-chain fatty acids (SCFAs); conversely, their scarcity and association with an excess of branched chain amino acids1 and lipopolysaccharides2 may contribute to the progression of the disease. While there is ample evidence linking certain MB profiles to diabetes, such as a higher Firmicutes/Bacteroidetes ratio (F/B ratio),3 or reduced SCFA-producing bacteria,1 there is no current evidence that these patterns can be modified through T2D treatment. Instead, much of the research in this area has focused on the influence of specific antidiabetic medications.4

In the MB of mammals, microorganisms not only interact with each other but also influence the hosts and are, in turn, influenced by them. Certain animals can precisely select specific bacterial strains to thrive. Although humans may not have the same level of control,5 we can influence our MB through dietary choices, and the consumption of probiotics, prebiotics, and fiber-rich diets can promote the growth of beneficial gut bacteria such as Bacteroidetes and certain Firmicutes.6 These bacteria are associated with the production of SCFAs, such as butyrate, acetate and propionate, that exert a wide range of positive effects in the gut environment, including inhibiting iNOS (inducible nitric oxide synthase) and stimulating beta-oxidation through PPAR-gamma agonism, reducing nitrate (a substrate for pathological bacteria) and oxygen in the lumen, thereby favoring the increase of anaerobic bacteria, which protect the mucus and intestinal barrier.1 7

The immune system also plays an important role in the composition of the MB. Driven by dysglycemia, in T2DM an increased Th17:Treg activation has been observed, favoring a behavior of B lymphocytes toward dysfunctional IgA-dependent bacterial selection, shaping MB in a significant manner,8 denoted by alpha-diversity change, which demands a broader community imbalance to occur.

Referencing these data, a spectrum of MB changes was observed in relation to increasing A1c levels. Specifically, MB richness and diversity were noted to slightly increase, while the F/B ratio decreased with higher A1c levels. The A1c thresholds analyzed in this study were 7–9%, 9–11%, and >11%.9

Given that dysglycemia substantially shifts the MB balance, highlighted by the Th17:Treg response compared with patients with normal blood sugar levels,10 and considering the continuous MB spectrum correlated with A1c levels,9 we hypothesized that intensive glucose-lowering treatment might help reverse the MB’s dysglycemia-influenced immune selection.

Pursuing this hypothesis, we administered usual T2DM treatment to two patient groups with T2DM, differentiated by initial A1c levels (7.1–8.9% vs >9%). We theorized that varying dysglycemia levels would distinctively impact the immune system (and thus MB selection), and that glucose management could induce diverse responses within these disparate conditions.

Methods

Study design

We conducted a pilot study, enrolling 31 subjects, comprising two groups of patients: one with A1c levels ranging from 7.1% to 8.9% (18 subjects: group 1—G1) and another with A1c levels >9% (13 subjects: group 2—G2). The 9% A1c threshold has been proven to represent a dividing point in this MB versus A1c spectrum.9

This study is an interventional cohort study, in which our objective was to observe gut microbiome changes after decreasing A1c levels (through standard diabetes medical care) by at least 1% from baseline measurements in patients with decompensated T2DM (divided in two groups, as cited above), evaluating at baseline, 6, and 12 months, any improvements in microbiome community parameters (including alpha and beta diversity), as well as to identify which bacterial genera would be favored or unfavored by these changes in the environment.

Enhancements in alpha and beta diversity were both desirable and anticipated, as greater community richness is commonly linked with improved metabolic health. The individual genera that exhibited changes underwent further literature investigation to understand their metabolic characteristics and their association with improved or deteriorated metabolic control. Additionally, we examined whether these genera were linked to complications related to T2DM.

The patients enrolled in the study had T2DM and were being treated at our hospital and had their treatment intensified according to Brazilian Diabetes Guidelines (SBD/SBEM),8 adhering to a 12-month period of treatment. Drugs used per group are described in the online supplemental material (S8). Stool and laboratory samples were collected at three intervals: baseline, 6 months, and 12 months. Anthropometric evaluations in respect of body mass index (BMI), waist-to-height ratio (WHR), diabetes duration, use of medications, and HbA1c were recorded at each visit.

Supplemental material

All participants signed an informed consent form and had the opportunity to clarify any queries before signing. Recruitment was conducted at the Hospital do Servidor Público Estadual de São Paulo (HSPE) from 2016 to 2020.

Inclusion and exclusion criteria

The eligibility criteria were having T2DM, being aged between 18 and 65 years, being male or female, consenting to the study’s terms, and having a BMI higher than 26 kg/m2, to exclude potential cases of latent autoimmune diabetes in adults (LADA) and T1DM. Furthermore, candidates should have experienced less than a 5% weight loss in the past 3 months.

Exclusion criteria included active liver disease (hepatitis or cirrhosis), chronic kidney disease (ClCr CKD-EPI <50), a history of neoplasia within the last 5 years, corticosteroid use for more than 14 days in the past 3 months, antibiotic use in the 30 days prior to stool sampling, alcohol and drug addiction, a diagnosis of T1DM or LADA, a positive HIV status, pregnancy, or uncontrolled psychiatric disease. Patients were excluded if they had used prebiotics or probiotics in the 3 months before the trial and were advised not to use during the protocol.

Stool sampling and fractioning

Stool samples were collected on the evening prior to the clinical appointment on the following morning, transported to the hospital under frozen conditions, and subsequently subdivided and stored at −80°C under dry ice conditions at the HU-USP.

DNA extraction and 16S rRNA sequencing

Total DNA from the collected stool samples was extracted using the QIAamp DNA Stool Mini Kit (QIAGEN, Hilden, Germany) following the manufacturer’s instructions. After extraction, the samples were stored at −80°C until the time of use. After total DNA extraction, the samples were quantified using the Qubit Fluorometer (Thermo Fisher Scientific, Waltham, Massachusetts, USA).

Fecal microbiota characterization was carried out by amplifying the V4 domain of the bacterial 16S ribosomal segment, following the protocol, primers and PCR conditions described by Kozich et al.11 A negative control of the DNA extraction and PCR were used. The sequencing procedures were executed following the manufacturer’s protocol (Illumina-16S Metagenomic Sequencing Library Preparation). The amplicons were pooled and loaded onto Illumina MiSeq clamshell style cartridge kit V2 with 500 cycles for paired-end 250 sequencing at a final concentration of 10 pM. The library was clustered to a density of approximately 820 k/mm2. The MiSeq platform was used for image analysis, base calling, and data quality assessment.

Microbiota analysis using bioinformatics tools

After obtaining the sequences, the 16S rRNA libraries obtained were analyzed using the QIIME V.2-2020.2 software. Denoising was performed with the DADA2 tool. The forward sequences were then truncated at position 251, while the reverse ones were truncated at the 250 nucleotides to discard the positions for which the median nucleotide quality was below Q30. Samples with less than 1000 sequences were also excluded from further analyses. Taxonomy was assigned using amplicon sequencing variant (ASV) through the q2-feature classifier feature and the naïve Bayes classify-sklearn taxonomy classifier, comparing the ASVs obtained against the SILVA 132 reference database. Subsequent analyses were performed using SPSS V.26 software and R Studio V.4.0.4, using the phyloSeq, vegan, microbiome, ggplot2 packages.

Data deposition

Sequence data have been deposited in the National Center for Biotechnology Information (NCBI) under BioProject ID SUB13801375.

Statistical analysis

Baseline anthropometric data, gender, and HbA1c were analyzed using the Fisher’s test (gender) and general linear model (GLM)—for the continuous variables. For the GLM, a paired test was conducted and adjusted for multiple comparisons, following the choice between the linear or gamma model through the Akaike information criterion (AIC). Sidak’s correction was applied for all GLM analyses. The data are presented in table 1 and were analyzed using SPSS software V.26.

Table 1

Baseline clinical characteristics

For the analysis of beta diversity (Bray-Curtis, weighted UniFrac and unweighted UniFrac), microbiomeanalyst.ca was used, using QIIME2 data input (SILVA taxonomic database— V.138—https://www.arb-silva.de/). For filtering, sequences were used if present in at least two individuals (for singleton exclusion) and with a minimum prevalence of 10% through samples. The low variance filter was set to zero. Samples with less than 100 UTIs were excluded. For data normalization, rarefaction (minimum library size) was carried out. Neither data scaling nor data transformation was necessary.

For this, the Permanova test was used for most parameters. Continuous variables were converted to categorical as follows:

  • Age: 20–40, 40–60, >60 years.

  • Waist/height ratio (WHR): 0.5–0.6, 0.7–0.8, >0.8 at baseline, then grouped as ascending, descending, or neutral at 6 and 12 months.

  • BMI: Overweight, obesity degrees I, II and III at the beginning of the study, then grouped as ascending, descending or neutral at 6 and 12 months.

  • HbA1c: A decrease rate below 1%, 1–2%, and >2% (reassessed at 6 and 12 months).

The above beta-diversity data related to HbA1c, WHR, and BMI were analyzed considering the difference between groups G1 and G2, as they could possibly change over time. Gender and age were analyzed considering the differences between the complete set of patients.

For comparisons between groups and time points regarding alpha diversity and F/B ratio, a linear regression model with mixed effects (random and fixed effects) was proposed, controlling for the following potential confounding variables: gender, BMI over time, WHR ratio over time, duration of T2DM (<10 or >10 years), and the use of gliclazide, insulin, pioglitazone, dapagliflozin, empagliflozin, and other medications (dulaglutide, glyburide, alogliptin and glimepiride—the least frequently used drugs) in a binary analysis. Mixed-effects linear models, ideal for analyzing clustered responses with multiple measurements per individual, adjust for non-independence within groups.12 They require normally distributed residuals, analyzed via histograms, quantile-quantile plots, and scatter plots. When normality assumptions failed, response variable transformations were applied. For the comparisons, a post-test by orthogonal contrasts was used. The analyses were conducted using the PROC MIXED of the SAS software V.9.4.6.

For the comparisons involving bacterial genera, a generalized linear mixed-effects model with a negative binomial distribution and a log-link function was employed,13 controlling for the previously mentioned variables. For the comparisons, a post-test by orthogonal contrasts was used. The analyses were conducted using the PROC GLIMMIX of the SAS software V.9.4.6. For all analyses, a significance level of 5% was adopted.

Bacterial data were deemed relevant if:

  1. The alteration was observed at the genus or species level, except for data concerning the Mollicutes order, given the limited literature available and its possible metabolic role.

  2. A significant change was observed from baseline to 12 months or if a noteworthy shift occurred between 6 and 12 months, indicative of a continuous process.

Transient (0–6 months) and non-significant changes over the entire duration (0–12 months) were excluded from the results.

To analyze the variation of A1c, BMI and WHR over time, linear regression model with mixed effects (random and fixed effects) was used.

Results

Baseline clinical characteristics

Patients enrolled exhibited similar characteristics, including in respect of age, sex, initial BMI, WHR, and duration of diabetes. However, according to our protocol, G2 initially presented with a higher A1c (table 1). The groups showed similar distributions in terms of BMI and diabetes duration (online supplemental material).

Clinical parameters evolution

Regarding diabetes control, the higher A1c group (G2) experienced an average reduction of 3.13% in A1c levels over 1 year, compared with a 1.1% reduction in G1. The most significant decrease occurred up to 6 months, at which point both groups attained similar levels and maintained them through to 12 months (figure 1).

Figure 1

Longitudinal A1c evolution. Groups showed significant change until 6 months, reaching the same A1c value (p>0.05) and remained stable until 12 months.

Regarding variations in BMI and WHR, the groups were consistent throughout the entire observation period. While group 2 exhibited a decrease in both measures from 0 to 6 months (leading to a difference from 0 to 12 months), this did not produce a significant difference between the groups (online supplemental material).

In G1, there were seven dropouts (six for insufficient A1c drop and one for withdrawing from the protocol). In G2, there were six dropouts (four for insufficient A1c drop and two for withdrawing from the protocol).

Alpha and beta diversity

No significant differences were identified in alpha and beta diversities between groups at baseline, and this similarity persisted over time (online supplemental material). In examining beta diversity, none of the evaluated potential parameters indicated a shift in microbiota between the groups. It showed similarity between groups regardless of the analyzed clinical parameters (figure 2).

Figure 2

Principal Coordinates Analysis (PCoA) (A) Baseline beta diversity G1 vs G2. (B) Beta diversity at 6 months, according to A1c variation (< 1%, 1–1.9%, > 2%). (C) Beta diversity at 12 months, according to A1c. All analyses (A–C) were non-significant.

Microbiome description

Concerning the F/B ratio, there were no differences in baseline or longitudinal observations between the groups. At baseline, the 10 most prevalent genera were Streptococcus, followed by Erysipelotrichaceae_UCG_003, Catenibacterium, Peptostreptococcaceae_Family, Fusobacterium, Holdemanella, Phascolarctobacterium, Blautia, Acidaminococcus, and members of Mollicutes order. Only one genus, Phascolarctobacterium, exhibited a significant intergroup difference at baseline, with higher levels observed in G1. At 6 and 12 months, there was a shift in the abundance hierarchy; however, only a select few demonstrated statistically significant variation (figure 3).

Figure 3

Main bacterial genera at baseline, 6 and 12 months. *Significantly different 0–6 m (at 6 months) and 6–12 or 0–12 months (12 months). See figure 4 for specific data.

Figure 4

Intragroup genera variation. An asterisk denotes statistical significance.

Variation in bacterial genera

Significant changes in bacterial genera were identified based on the criteria specified in the Statistical analysis section. These changes primarily occurred within individual groups.

Intergroup analysis

On examining the intergroup differences, only Phascolarctobacterium showed significant variation as illustrated in the online supplemental material.

Phascolarctobacterium (baseline and at 6 months): G1 started higher than G2 and initially increased, but subsequently underwent a significant decline, with G1 being consequently higher than G2 at the 6-month mark.

Intragroup analysis

In group 1 (lower A1c), some specific genera showed an increase in relative abundance over the study time. Dialister demonstrated an overall positive trend from 0 to 12 months (p=0.02; CI −2.79 to −0.24). There were some decreasing genera: Phascolarctobacterium decreased over the 0–12 month period (p=0.03; CI 0.10 to 2.48), Mollicutes order continuously declined, resulting in an overall decrease over the 0–12 months (p=0.02; CI 0.45 to 4,53). Finally, Erysipelotrichaceae_UCG_003 decreased from 6 to 12 months (p=0.03; CI 0.06 to 1.40), culminating in a total reduction from 0 to 12 months (p=0.01; CI 0.28 to 1.59) (figure 4).

In group 2 (higher A1c), the rising genera were Acidaminococcus, which increased from 6 to 12 months (p=0.047; CI −3.37 to −0.02), and Megasphaera, which presented an increase from 6 to 12 months (p=0.01; CI −3.80 to −0.52). Ultimately, Phascolarctobacterium levels significantly increased from 6 to 12 months (p=0.03; CI −3.04 to −0.19), indicating an upward trend from 0 to 12 months (p=0.02; CI −3.11 to −0.27).

Not considered a significant change, but interesting concerning the whole scenario, Erysipelotrichaceae_UCG_003 (described above) showed a decrease in G2 from baseline to 6 months (p=0.04; CI 0.04 to 1.93). However, this declining trend did not continue beyond this period.

Discussion

Our analysis revealed two clinically distinct groups based on A1c values. However, other parameters, including BMI, diabetes duration, age, sex, and WHR, were equivalent between the groups, and all included as confounders in multivariate analysis. This similarity was also observed in both alpha-diversity and beta-diversity indices and the overall baseline measurements of genera, with the only difference concerning Phascolarctobacterium at baseline. Notably, our data underscore the almost complete similarity between two distinct A1c levels, marking our primary observation, which contrasts to previous data.9

There was no change in alpha and beta diversity following diabetes compensation. In humans, prior research supports this trend. While no other intervention study has compared distinct A1c starting values, alpha and beta diversity remained unaltered with glp-1 agonists, idpp4, or isglt2 (online supplemental material). Some studies identified differences, but only in metformin-naïve patients14 or in studies with larger sample sizes (online supplemental material). Traditionally, higher microbiota diversity has been associated with improved health outcomes. However, this perspective is currently undergoing scrutiny, primarily because diversity does not necessarily equate to quality: a more diverse pathobiont-based microbiota does not inherently indicate better health.10 Such observation could explain other groups data showing higher MB richness in patients with higher A1c.9

Similarly, our results did not show any alterations in the F/B ratio. While this ratio was previously considered an indicator of health, it is becoming less definitive.15 The Firmicutes phylum, for instance, is linked with SCFA production3 and improved health, but also houses certain pathogen producers of ammonia,16 17 which can compromise the gut barrier. Therefore, this ratio provides only a gross overview; what truly matters is the presence and activity of specific bacteria in the microbiota, whether they are beneficial or detrimental.

Determining the health impact of a particular bacterium is complex. A primary reason is that much research relies on 16S sequencing, which offers resolution only at the phyla-genera level.18 19 At the genera level, various pathways are documented for different species, leaving room for speculation regarding their contribution to observed results. Another consideration is the competition within ecological niches. The microbiota operates as an ecosystem, where components interact. These interactions extend beyond mere competition. They can involve intermediate substances that promote or inhibit certain genera, adding complexity to the interpretation of any fluctuations.1 7 So, as we also used 16S sequencing, here we propose an interpretation of the genera observed considering the data about each of them or its species previously cited in the literature.

Decreasing in G1, Erysipelotrichaceae_UCG_003 is very rarely recorded within the Erysipelotrichaceae genus, with little information available.20 It is a member of the Firmicutes phylum, and its genus is linked to Western diet,21 hepatic steatosis,22 23 and sleep deprivation24—all characteristics observed in our patient cohort. It decreased in G1, and given its pro-dysmetabolic nature, we interpreted this decline as favorable. Conversely, in G2, it initially experienced a decline but ceased this downward trend, not matching our definition of biologically significant change.

The Mollicutes order decreased in G1, but there is very little data in the literature about it. Studies have associated it with a Western diet in rats,25 and with obesity and increased visceral fat in humans.26 This reduction is compatible with the status of metabolic improvement underway by treatment. Although it is an order, and not a genus, given the scarcity of data about it, it is important to take it into account in respect of future research. Although deemed to be metabolically detrimental, Mollicutes order does not have a clear pathway for being so.25 26

In G1, a peculiar behavior could be observed in Veillonellaceae family: Two members of this family, Phascolarctobacterium and Dialister, exhibited notable changes. These two bacteria were grouped for discussion because they share similarities in their contradictions and share the same ecological niche,27 which could explain the mirror behavior observed, probably driven by competition: with an increase in Dialister and a reduction in Phascolarctobacterium. Phascolarctobacterium is associated with better insulin sensitivity, while Dialister is with worse IR in patients with obesity and overweight.28 Phascolarctobacterium is also associated with a better weight loss outcome in patients with obesity, while Dialister has been shown to be a predictor of a worse response.27 Phascolarctobacterium is an SCFA producer,29 especially butyrate,30 acetate, and propionate,31 which could justify its metabolically favorable profile. However, although Dialister is a pathogenic bacterium linked to oral infections27 and worsening of insulin resistance (IR), it produces acetate, lactate, and propionate, but not butyrate.32 In a human study, it was more linked to IR than to BMI variation.28

In G2, a parallel movement was observed with Acidaminococcus and Megasphaera, both members of the Firmicutes phylum. Megasphaera (genus) is associated with SCFA production,33 34 such as butyrate from lactate35 and propionate,36 which has the potential to improve T2DM control through the described pleiotropic effects. However, the pathway it uses for this synthesis produces ammonia as a byproduct, which is potentially harmful.26 Acidaminococcus is an amino acid fermenter, thus an SCFA producer37—acetate and butyrate.38 Like Megasphaera, it produces ammonia as a byproduct.37 Megasphaera has been associated with various states of higher metabolic-inflammatory risk, such as diabetes renal disease (DRD) versus healthy volunteers or T2DM non-DRD,39 as well as Acidaminococcus.40 Megasphaera has also been associated with diabetic neuropathy versus patients with T2DM without neuropathy.41 These two bacteria, both fermenters and both ammonia producers, had similar behaviors in group G2, increasing over time. The fact that they thrived within a potentially more inflamed environment—in the group with worse control—aligns with the profile of the patients they are usually associated with, with higher morbidity, such as DRD. The individual behavior of each is difficult to understand but probably depends on an interaction with other bacteria.

Phascolarctobacterium rised in G2. Deemed to be metabolically favorable as described above, it was the only genus associated with good health that changed significantly in G2. Conversely, no mirrored behavior concerning Dialister was observed in this group.

Considering the changes cited, the initially better controlled group (G1) had an improvement in its MB, with falling detrimental strains. Conversely, G2, while showing a more substantial improvement in A1c, manifested an increase in only one beneficial bacterium and an increase in two proinflammatory strains. Both cohorts underwent identical guideline-based treatment regimens and outcomes were adjusted for the use of medications. Both groups were previously on metformin, a prominent agent affecting MB. Yet, despite the pronounced A1c improvement in G2, only G1 appeared to create a conducive environment for a favorable MB shift. This occurred despite the increase in Phascolarctobacterium in G2, which could ameliorate the environment with SCFA synthesis. The halted decline in Erysipelotrichaceae_UCG_003 levels might also indicate a similar inability to improve the environment despite T2DM compensation.

In G2, two-thirds of the increasing bacteria, although SCFA producers, were also ammonia producers—a factor known to undermine gut barrier. This is an example of how F/B ratio can be misleading.

Our study has some limitations which should be noted. First, with 31 patients, our sample size could limit the statistical robustness required to discern subtle MB changes. This sample size was partly due to a high dropout rate, partly for the pandemics, when we stopped recruiting. Second, the broad inclusion criteria, spanning the overweight to those with grade 3 obesity, could give rise to questions in respect of the potential influence of obesity on our findings. However, our analyses were conducted using a robust linear regression adjusted for BMI, mitigating such biases. Third, we adjusted the results for T2DM duration setting a threshold of 10 years. Few studies investigate the relationship between T2DM duration and MB, and none explores the effects of longer durations of T2DM (over 10 years).42 43 Our choice of this duration threshold was guided by the distribution of our data (online supplemental material). Finally, we speculate over mechanisms that could have driven the changes observed, and this is the aim of this pilot study, which should bring ideas to future research. Certainly, trials with a larger sample size, and methods like shotgun genomic sequencing (for better MB characterization) and metabolomics for understanding functional aspects are needed to broaden the view over these phenomena.

Conclusion

The results of this study indicate that an exclusive glucose-controlling intervention can contribute to restoring gut microbiome balance, although this depends on the initial degree of T2DM decompensation. After glucose-lowering intervention, patients with an A1c >9%, G2, tended predominantly to a proinflammatory MB profile besides greater A1c drop, while those with an initial A1c <9% were associated with a greater likelihood of restoring the MB to a healthier level.

Given this resistance to restore MB in patients with higher A1c, and the permanence of an inflammatory component (Megasphera, Acidaminococcus) despite A1c compensation, it is inevitable to remind of metabolic memory, which happens in the same manner in long-decompensated T2DM. So, we hypothesize that gut microbiota could be one of the persistent focuses of inflammation in more patients with decompensated (A1c>9%) T2DM, besides glucose amelioration, contributing to metabolic memory. This is obviously an observation for further research and way beyond the scope and reach of our work but certainly not neglectable and an important path to be explored.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.

Ethics statements

Patient consent for publication

Ethics approval

The research was approved by the Ethics Committee of the HSPE (approval number 1.592.942) and the Hospital Universitário da Universidade de São Paulo (HU-USP) (approval number 1.744.732). Participants gave informed consent to participate in the study before taking part.

Acknowledgments

The authors would like to thank Drs Ana Beatriz Pinotti Pedro Miklos, Erika Ribeiro Barbosa, Larissa Bianca Paiva Cunha de Sá and Paula Paes Batista da Silva for their help in recruiting patients, and Dr Evandro Souza Portes for helping with logistics and support in several ways. We would also like to thank all the nurses from the Endocrine Department for helping with logistics and patient care. Finally, we are grateful to the microbiology department of IAMSPE for storing the samples. Artificial intelligence resources (ChatGPT-4) were used to check the grammar and clarity of the English in the text but was not used for generating any data.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • Contributors TFN conducted all steps of the research and wrote the manuscript. TFN, CRT and JENS designed the experiments and analyzed the data. CRT and JENS reviewed the manuscript and gave opinions and ideas. TFN, CRT, RVC and LGS were responsible for processing and analyzing the samples. RVC was responsible for bioinformatics processing. JENS and CRT were responsible for the thematic project, of which this trial is a part, and obtained the funding. JENS is

    responsible for the overall content as the guarantor.

  • Funding This study was funded by the Fundo de Amparo à Pesquisa (FAP) from Faculdade de Ciências Médicas da Santa Casa de São Paulo (FCMSCSP).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.