Alteration of subgingival microbiome in patients with type 2 diabetes induced by diabetes treatment Authors: Moe Inoue, Akito Sakanaka, Naoto Katakami, Hitoshi Nishizawa, Kazuo Omori, Naohiro Taya, Asuka Ishikawa, Shota Mayumi, Hiroki Takeuchi, Atsuo Amano, Toru Takeshita, Eiichiro Fukusaki, Iichiro Shimomura, Masae Kuboniwa Published: 2026 ABSTRACT ABSTRACTAims/IntroductionDiabetes and periodontal disease, a bacterial infection involving the oral microbiome, have a bidirectional relationship. However, the relationship between poor glycemic control and the oral microbiome remains unclear. The aim of this study was to assess the direct effect of intensive glycemic control on the oral microbiome in patients with type 2 diabetes mellitus (T2DM).Materials and MethodsTwenty‐nine patients with T2DM and 29 controls were enrolled in this study. Patients underwent a 2‐week intensive glycemic control regimen. We collected subgingival dental plaque and conducted full‐length 16S ribosomal RNA gene sequencing to identify bacterial species. The focus was on the subgingival microbiome, particularly 29 species of oral bacteria related to periodontal disease, and the characteristics of the diabetic cohort. Periodontal inflammation was quantified using the periodontal inflamed surface area (PISA).ResultsPrior to diabetes treatment, the relative abundance of Porphyromonas gingivalis (P = 0.018), Tannerella forsythia (P < 0.001), Fusobacterium nucleatum (P = 0.041), and Prevotella intermedia (P < 0.001), known for their high pathogenicity in periodontal disease, was higher in the diabetes group than in the control group. The bacterial species whose relative abundance changed significantly before and after treatment differed between the PISA improved and non‐improved groups in patients with diabetes.ConclusionsOral bacteria changed pre‐ and post‐diabetes treatment, correlating with improvements in PISA.Keywords: Diabetes mellitus, microbiota, periodontal disease INTRODUCTION Diabetes mellitus is characterized by a relative or absolute deficiency of insulin action, leading to hyperglycemia 1 . According to the polymicrobial synergy and dysbiosis model, periodontal disease is initiated by a synergistic and dysbiotic subgingival microbial community, which leads to changes in the host–microbiome immune response and causes periodontium destruction 2 . These bacterial species are categorized into six closely associated groups: red, orange, yellow, green, purple, and blue complexes 3 . Of these, species in the red and orange complexes are considered late and mid colonizers, while those in the other four complexes are early colonizers of the tooth surface and precede the multiplication of other bacterial species 4 . Among them, Porphyromonas gingivalis (P. gingivalis) is a keystone pathogen of the disease‐provoking microbiota, impairs immune surveillance, and tips the balance from homeostasis to dysbiosis, leading to overgrowth of commensal bacteria 2 , 5 , 6 . Diabetes and periodontal disease are known to have a bidirectional relationship 7 , 8 , 9 . Some reviews show that periodontal treatment improves hyperglycemia in poorly controlled diabetics 8 , 9 . By contrast, our previous study showed that diabetes treatment alone significantly improved periodontal inflammation 10 . Some researchers suggest that microbiome changes may influence this bidirectional relationship 11 , 12 , 13 . However, evidence regarding microbiota changes in poorly controlled diabetes remains inconsistent 1 , 11 , 12 , 14 . Our previous study showed that plasma metabolites changed after diabetes treatment, potentially altering the serum‐derived gingival crevicular fluid, which supplies nutrients for the subgingival microbiome 15 , 16 , 17 , 18 . We hypothesized that this potentially changes the bacterial species in the subgingival microbiome and improves periodontal inflammation. The aim of this study was to investigate the direct effect of intensive diabetes treatment during hospitalization on the oral microbiome, particularly regarding periodontal disease and the degree of improvement in periodontal inflammation. MATERIALS AND METHODS Study participants Patients with type 2 diabetes mellitus (T2DM), aged 20–75 years, who were hospitalized for 2 weeks between December 2017 and March 2019 at the Division of Endocrinology & Metabolism, Osaka University Hospital, were recruited for this study. T2DM was diagnosed based on the criteria established by the World Health Organization National Diabetic Group 2006 and/or diabetes treatment status. The exclusion criteria were (i) inability to achieve strict glycemic control due to severe hypoglycemia or advanced diabetic retinopathy, in which case a substantial reduction in glucose levels was deemed inappropriate; (ii) severe nephropathy, defined by a serum creatinine level >2.0 mg/dL; (iii) presence of acute infection, severe trauma, or active malignancy; (iv) undergoing medical examinations during the preoperative and/or postoperative periods; and (v) deemed unsuitable for participation by the attending physician. During hospitalization, comprehensive risk management for diabetes mellitus was implemented for all patients. It included intensive glycemic control, blood pressure regulation, lipid management, and body weight control. Following the Japanese treatment guidelines for diabetes, fasting blood glucose <130 mg/dL (7.2 mmol/L) and 2‐h postprandial blood glucose <180 mg/dL (10.0 mmol/L) were maintained. Volunteers without diabetes, aged 20–75 years, were recruited for comparison. Their exclusion criteria were (iii)–(v) above. Periodontal treatment, including oral hygiene instructions, was not provided to any participant. The Human Ethics Committees of the Osaka University Hospital (no. 16374‐6), Osaka University Graduate School of Dentistry, and Osaka University Dental Hospital (no. H28‐E40‐2) approved this study. This study was conducted in accordance with the Declaration of Helsinki and STROBE guidelines for human observational studies. Written informed consent was obtained from all participants. Systemic and oral indicators Thirty systemic and six oral indicators were assessed on the 2nd and 14th day of hospitalization, respectively. Blood samples were collected in the morning after a 15‐h fasting period and analyzed using standard techniques. Vital signs, body weights, blood pressure, and estimated glomerular filtration rate (eGFR) were measured as previously described 10 . A dental survey conducted by four calibrated dentists was performed in the afternoon. All participants were instructed to avoid eating, drinking, or brushing for at least 1 h before the examination. The number of residual teeth was counted, and periodontal parameters, including probing pocket depth (PPD), bleeding on probing (BOP), gingival recession, and clinical attachment level (CAL), were measured at six sites on all teeth. Periodontal inflamed surface area (PISA) was used to quantify the amount of inflamed periodontal tissue 19 . In the present study, PISA per tooth (PISA/tooth) was used for adjusting the number of residual teeth. Concurrently, plaque accumulation was assessed using the average score of plaque index (PlI), an indicator of overall oral hygiene status 20 , 21 . For participants without diabetes, systemic and dental indicators were measured once, following the same procedure. Dental plaque collection, deoxyribonucleic acid (DNA) extraction, and 16S ribosomal ribonucleic acid (rRNA) gene sequencing Following the periodontal assessment, subgingival dental plaque was collected. After wiping the tooth surface near the gingival margins with sterilized cotton balls, four sterilized paper points (Morita, Japan) were inserted into the deepest periodontal pockets and held in place for 10 s. Subsequently, the paper points were removed, immersed in 700 μL of phosphate‐buffered saline, and stored at −80°C until DNA extraction. DNA extraction was conducted using a bead‐beating method 22 , 23 , with detailed procedures described in Method S1. DNA purification was conducted via phenol extraction, phenol‐chloroform‐isoamyl alcohol extraction, and ethanol precipitation. The precipitates were resuspended in 50 μL of 10 mM Tris‐hydrochloric acid. After DNA extraction, full‐length 16S rRNA gene sequencing analysis was performed for species identification. Library preparations, DNA sequencing, and species identification with Silva 138 were conducted using PacBio technology by AZENTA LIFE SCIENCE (Suzhou, China) (Method S2). Definition Values obtained on the second and 14th day of the general and oral examinations were defined as pre‐ and post‐treatment values for glycemic control, respectively. For oral microbiome analysis, relative abundance represents the ratio of the reads detected for a specific bacterium to the total number of reads detected in a sample. Statistical analysis Four groups were analyzed in this study: patients with T2DM, participants without diabetes (control group), PISA improved group (Imp), and PISA non‐improved group (N‐Imp). Categorization into Imp and N‐Imp groups for patients with T2DM was based on PISA/tooth improvement: Imp group (<−5.0 mm2) and N‐Imp group (≥−5.0 mm2) 10 . The distribution of data was assessed using the Shapiro–Wilk test. Variables were considered parametric when both compared samples yielded a P value ≥0.05; otherwise, they were classified as nonparametric. The homoscedasticity of parametric data was assessed using Levene's test. Variables were considered homoscedastic when P value was 0.05 or more; otherwise, they were classified as heteroscedastic. The differences in values between the two groups were analyzed using Student's t‐test for parametric and homoscedastic variables, Welch's test for parametric and heteroscedastic variables or the Mann–Whitney U‐test for nonparametric variables. Differences in qualitative values were analyzed using the chi‐square test. Paired samples were analyzed using paired t‐tests for parametric variables or Wilcoxon signed‐rank tests for nonparametric variables. For oral microbiome analysis, bacterial species diversity was compared among the control, T2DM before treatment, and T2DM after treatment groups. The number of identified species and the Shannon index were analyzed using the Kruskal–Wallis test. Principal coordinate analysis (PCoA) was conducted using the Bray–Curtis distance matrix. Orthogonal partial least squares discriminant analysis (OPLS‐DA) was used to assess the bacterial species characterizing the T2DM group. OPLS‐DA is a modeling approach that discovers variables with the largest discriminatory power 24 . The validity of the model was judged using R2 values, which represent the goodness of the model fit, and Q2 values, which show the predictive ability of the model: the higher these values, the better the model 24 , 25 . Additionally, the variable influence on projection (VIP) value, showing the total contribution of x variables to the OPLS‐DA model, and the S‐plot, which highlights the bacterial species that most significantly influenced the group separation, were used to identify the target bacterial species 24 , 26 . VIP values greater than 1 indicate relevance for group discrimination 27 . The S‐plot shows the relationship between the magnitude of change (p [1]) and the reliability (p(corr) [1]) for each variable in the model 26 . The following criteria were used to select the bacterial species for the Shannon index, PCoA, and OPLS‐DA: (i) species with a relative abundance >0.01% and (ii) species identified in >10% of all participants. General analyses were performed using SPSS Statistics version 29 (IBM Corp., Armonk, NY, USA). Oral microbiome analysis was performed using R version 4.4.1. OPLS‐DA was performed using SIMCA 16 (SARTORIUS, Göttingen, Germany). Statistical significance was set at P‐value <0.05. RESULTS Study population Thirty‐one patients with T2DM, aged 41–73 years, and 30 controls, aged 30–71 years, were enrolled. Two patients with diabetes declined an oral examination on the 14th day of hospitalization, and one control did not provide a subgingival plaque sample. Therefore, these individuals were excluded from the analysis, leaving a final cohort of 29 participants in each of the T2DM and control groups. Participants in the T2DM group were the same patients analyzed in our previous study 10 . Clinical characteristics of participants The systemic and oral disease indicators observed in the T2DM group before treatment and those in the control group are shown in Table 1. Patients in the T2DM group were significantly older (P < 0.001) and exhibited higher levels of body mass index (P < 0.001), waist circumference (P < 0.001), triglycerides (P < 0.001), high‐sensitivity C‐reactive protein (P = 0.025), fasting plasma glucose (P < 0.001), glycated hemoglobin (HbA1c) (P < 0.001), glycated albumin (GA) (P < 0.001), acetoacetic acid (P < 0.001), beta hydroxybutyrate (P < 0.001), and PISA/tooth (P < 0.001) compared with participants in the control group (Table 1). Additionally, patients in the T2DM group had a higher prevalence of smoking (P = 0.005), hypertension (P < 0.001), and dyslipidemia (P < 0.001) (Table 1). By contrast, the levels of eGFR (P < 0.001) and high‐density lipoprotein cholesterol (P < 0.001) in the T2DM group were lower than those in the control group (Table 1). The median values for the number of residual teeth, BOP rate, and CAL were 23.0, 29.2%, and 3.81 mm, respectively, in the T2DM group and 28.0, 17.2%, and 3.23 mm, respectively, in the control group. Table 1Values for systemic and oral disease indicators in the control and T2DM groups before diabetes treatmentClinical indicatorsControl (n = 29)T2DM (n = 29) P‐valueSex (Female/Male)9 (31.0)/20 (69.0)16 (55.2)/13 (44.8)0.063Age † 37.0 (34.0–47.0)68.0 (65.0–70.0)<0.001BW † (kg)61.0 (52.0–70.0)66.4 (61.7–72.9)0.078BMI † (kg/m2)22.0 (19.7–23.9)26.3 (23.7–28.0)<0.001Waist ‡ (cm)81.5 ± 10.598.1 ± 8.7<0.001Smoking history4 (13.8)14 (48.3)0.005Hypertension2 (6.90)22 (75.9)<0.001Dyslipidemia1 (3.45)25 (86.2)<0.001Macroangiopathy0 (0.0)11 (37.9) ¶ Coronary arterial disease0 (0.0)7 (24.1) ¶ Cerebral infarction0 (0.0)5 (17.2) ¶ Peripheral arterial disease0 (0.0)2 (6.9) ¶ UA ‡ (mg/dL)5.59 ± 1.225.77 ± 1.090.55Cre † (mg/dL)0.750 (0.640–0.860)0.700 (0.640–0.850)0.852eGFR ‡ (mL/min/1.73 m2)85.9 ± 9.867.9 ± 16.8<0.001AST † (U/L)22.0 (18.0–26.0)22.0 (17.0–27.0)0.645ALT † (U/L)23.0 (15.0–30.0)22.0 (17.0–37.0)0.755γ‐GTP † (U/L)22.0 (17.0–30.0)27.0 (19.0–47.0)0.107CPK † (U/L)100 (77–118)93.0 (71.0–129.0)0.388TC † (mg/dL)205 (176–228)200 (174–215)0.388TG † (mg/dL)70.0 (53.0–104.0)125 (83–235)<0.001HDL‐cho § (mg/dL)68.2 ± 20.452.5 ± 12.1<0.001LDL‐cho † (mg/dL)123 (95–145)117 (91–137)0.514hs‐CRP † (ng/mL)388 (132–670)710 (436–1,170)0.025FPG † (mmol/L)4.89 (4.61–5.28)7.50 (6.28–9.50)<0.001HbA1c † (%)5.30 (5.20–5.60)8.50 (7.70–10.30)<0.001GA § (%)13.7 ± 1.423.8 ± 5.8<0.001AcAc † (μmol/L)7.00 (5.00–10.00)66.0 (41.0–101.0)<0.001BHB † (μmol/L)45.0 (30.0–70.0)103 (46–208)<0.001uAlb/Cr † (mg/gCr)4.90 (3.87–8.80)12.5 (3.4–40.0)0.142PlI ‡ 0.778 ± 0.4061.02 ± 0.500.051PISA/tooth † 8.12 (4.56–10.57)18.5 (11.7–26.9)<0.001No symbols = Data are presented as counts (percentages) using the chi‐squared test. † Data are presented as the median of all participants' values (interquartile range) using the Mann–Whitney U‐test. ‡ Data are given as the average of all participants' values (one standard deviation) using Student's t‐test. § Data are given as the average of all participants' values (one standard deviation) using Welch's test. ¶ No P‐value. The chi‐squared test could not be used because no patients in the control group had macroangiopathy, coronary arterial disease, cerebral infarction, or peripheral arterial disease. The P‐value indicates the result of comparing the control and T2DM groups before diabetes treatment.AcAc, acetoacetic acid; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BHB, beta hydroxybutyrate; BMI, body mass index; BW, body weight; CPK, creatine phosphokinase; Cre, creatinine; eGFR, estimated glomerular filtration rate; FPG, fasting plasma glucose; GA, glycated albumin; HbA1c, glycated hemoglobin; HDL‐cho, high‐density lipoprotein cholesterol; hs‐CRP, high‐sensitivity C‐reactive protein; LDL‐cho, low‐density lipoprotein cholesterol; PISA/tooth, periodontal inflamed surface area/number of residual teeth; PlI, plaque index; TC, total cholesterol; TG, triglycerides; UA, uric acid; uAlb/Cr, urine albumin/creatinine ratio; γ‐GTP, γ‐glutamyl transpeptidase. Values for systemic and oral disease indicators in the control and T2DM groups before diabetes treatment No symbols = Data are presented as counts (percentages) using the chi‐squared test. Data are presented as the median of all participants' values (interquartile range) using the Mann–Whitney U‐test. Data are given as the average of all participants' values (one standard deviation) using Student's t‐test. Data are given as the average of all participants' values (one standard deviation) using Welch's test. No P‐value. The chi‐squared test could not be used because no patients in the control group had macroangiopathy, coronary arterial disease, cerebral infarction, or peripheral arterial disease. The P‐value indicates the result of comparing the control and T2DM groups before diabetes treatment. AcAc, acetoacetic acid; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BHB, beta hydroxybutyrate; BMI, body mass index; BW, body weight; CPK, creatine phosphokinase; Cre, creatinine; eGFR, estimated glomerular filtration rate; FPG, fasting plasma glucose; GA, glycated albumin; HbA1c, glycated hemoglobin; HDL‐cho, high‐density lipoprotein cholesterol; hs‐CRP, high‐sensitivity C‐reactive protein; LDL‐cho, low‐density lipoprotein cholesterol; PISA/tooth, periodontal inflamed surface area/number of residual teeth; PlI, plaque index; TC, total cholesterol; TG, triglycerides; UA, uric acid; uAlb/Cr, urine albumin/creatinine ratio; γ‐GTP, γ‐glutamyl transpeptidase. As our previous study showed, diabetes treatment could improve glycemic control 10 . GA and HbA1c decreased significantly in the T2DM group after treatment (GA: P < 0.001; HbA1c: P < 0.001). Diversity in subgingival microbiome The 16S rRNA gene sequencing identified 583 bacterial species, with 234 meeting the criteria outlined in the statistical analysis section. The number of identified species and the Shannon index in the control and T2DM before and after diabetes treatment groups are shown in Figure 1a,b. No significant differences were observed between groups (number of identified species, P = 0.80; Shannon Index, P = 0.53). PCoA did not distinguish the groups (Figure 1c). Figure 1d shows the OPLS‐DA model for bacterial species in control and T2DM before treatment groups (R2X = 0.236; R2Y = 0.410; Q2 = 0.230). These groups were separated along the first principal component (t [1]) on the horizontal axis. This demonstrates that the composition of the subgingival microbiome tended to differ between the two groups. VIP values for the target bacterial species in the control and T2DM before‐treatment groups are shown in Figure 1e. The bacterial species with the highest VIP values (>1.0) are listed in Table S1. They included some bacterial species regarded as new candidate periodontal pathogens, such as Treponema socranskii, Selenomonas sputigena, and Filifactor alocis 28 . The VIP values of the top five bacterial species were markedly higher than those of the remaining species (Table S1, Figure 1e). These five species also exhibited a high magnitude of change and reliability in the S‐plot (Table 2, Figure 1f). These bacterial species were prominent targets for discrimination and belong to the red, orange, and yellow complexes. Therefore, we focused on 29 species associated with periodontal disease 3 among the 234 species in the subsequent analysis (Table S2). Diversity in subgingival microbiome. Box plots show the first quartile, median, and third quartile. Green, pink, and purple dots indicate the value of control, T2DM before diabetes treatment group and after diabetes treatment group, respectively. (a) The number of identified species. (b) Shannon index. (c) Principal coordinates analysis. The distance between the dots indicates the similarity of the microbial community in the samples. (d) Orthogonal partial least squares discriminant analysis modeling. (e) Variable influence on projection value. Dots indicate each bacterial species' values. The five dots within the circle represent the top five bacterial species with VIP values markedly higher than those of the others. (f) S‐plot analysis. Dots indicate the values for each bacterial species. Five red dots indicate the bacterial species with the highest VIP values. NS, not significant. Table 2S‐plot features of the bacterial species with the top five VIP valuesSpecies nameVIPMagnitude (P [1])Reliability (P(corr) [1])Trend Streptococcus oralis 7.61−0.541−0.787Increased in Control Fusobacterium nucleatum 4.230.3010.601Increased in T2DM Porphyromonas gingivalis 3.820.2720.497Increased in T2DM Parvimonas micra 3.620.2570.539Increased in T2DM Tannerella forsythia 2.990.2130.654Increased in T2DMVIP values and S‐plot features were obtained from the OPLS‐DA model comparing the control and T2DM before treatment groups.Control, control group; T2DM, patients with type 2 diabetes mellitus before treatment group; VIP, variable influence on projection. S‐plot features of the bacterial species with the top five VIP values VIP values and S‐plot features were obtained from the OPLS‐DA model comparing the control and T2DM before treatment groups. Control, control group; T2DM, patients with type 2 diabetes mellitus before treatment group; VIP, variable influence on projection. Characteristic features of the T2DM group Before diabetes treatment, the T2DM group had higher relative abundance of red and orange complexes compared to the control group. Specifically, P. gingivalis (P = 0.018), Tannerella forsythia (T. forsythia) (P < 0.001), Fusobacterium nucleatum (F. nucleatum) (P = 0.041), and Prevotella intermedia (P. intermedia) (P < 0.001) were significantly more abundant (Figure 2a). Additionally, the relative abundance of Eubacterium spp. (P = 0.001) in the T2DM group was significantly higher than in the control group. Conversely, the significantly lower relative abundance of Streptococcus oralis (S. oralis) (P < 0.001), Streptococcus sanguinis (S. sanguinis) (P < 0.001), Capnocytophaga sputigena (C. sputigena) (P < 0.001), Haemophilus parainfluenzae (P < 0.001), and Actinomyces naeslundii (A. naeslundii) (P = 0.032) was observed in the T2DM group (Figure 2a). Additionally, when comparing the relative abundance before and after diabetes treatment, Campylobacter gracilis (C. gracilis) (P = 0.045) and C. sputigena (P = 0.018) increased significantly, while Eubacterium nodatum (E. nodatum) (P = 0.022), P. intermedia (P = 0.018), Streptococcus intermedius (S. intermedius) (P = 0.038), Campylobacter concisus (C. concisus) (P = 0.007), and Eubacterium spp. (P = 0.012) decreased significantly (Figure 2b). Oral bacteria characterization in the T2DM group. Box plots show the first quartile, median, and third quartile. (a) Bacterial species whose relative abundances were significantly different between the T2DM and control groups before diabetes treatment. Green and orange dots indicate the values for the control and T2DM groups, respectively. *P < 0.05, **P < 0.01. (b) Bacterial species changed significantly between before and after diabetes treatment in the T2DM group. Dots indicate each patient's values, with a line connecting before and after treatment data for each individual. *P < 0.05, **P < 0.01. Characteristic features of Imp and N‐Imp groups The relative abundance of S. sanguinis (P = 0.016) was significantly lower in the Imp group than in the N‐Imp group before diabetes treatment (Figure 3a). Analyzing bacterial changes before and after diabetes treatment showed differences in both groups. In the Imp group, the relative abundances of C. gracilis (P = 0.01), Streptococcus constellatus (S. constellatus) (P = 0.021), and S. sanguinis (P = 0.041) increased significantly (Figure 3b). Moreover, a significant reduction was observed in S. intermedius (P = 0.012), C. concisus (P = 0.021), and Veillonella parvula (V. parvula) (P = 0.018) (Figure 3b). By contrast, only P. intermedia (P = 0.034) was significantly decreased in the N‐Imp group (Figure 3b). After diabetes treatment, Treponema denticola (T. denticola) (P = 0.045), F. nucleatum (P = 0.031), and Eubacterium spp. (P = 0.035) were more abundant, while S. sanguinis (P < 0.001) and A. naeslundii (P = 0.002) were significantly less abundant in the Imp group (Figure 3c). Oral bacteria characterizing the Imp and N‐Imp groups. Box plots show the first quartile, median, and third quartile. (a) Bacterial species whose relative abundances were significantly different between the Imp and N‐Imp groups before diabetes treatment. Nineteen blue dots and 10 red dots indicate the value of the Imp and N‐Imp groups, respectively. *P < 0.05. (b) Bacterial species changed significantly between before and after diabetes treatment in the Imp and N‐Imp groups. Dots indicate each patient's values, with a line connecting the before and after treatment data for each individual. Blue and red lines indicate the Imp and N‐Imp groups, respectively. *P < 0.05, **P < 0.01. (c) Bacterial species whose relative abundances were significantly different between the Imp and N‐Imp groups after diabetes treatment. Nineteen blue dots and 10 red dots indicate the value of the Imp and N‐Imp groups, respectively. *P < 0.05, **P < 0.01. DISCUSSION In this study, the bacterial species diversity did not differ significantly in the control and T2DM groups before and after diabetes treatment. Among bacteria associated with periodontal disease, in the T2DM group, the relative abundance of bacterial species in the red and orange complexes was higher than in the control group before comprehensive diabetes treatment, consistent with previous findings 12 , 29 . Comparing the Imp and N‐Imp groups, the bacterial changes before and after diabetes treatment were different. Numerous studies have investigated the subgingival microbiome of patients with diabetes. However, findings on diabetes‐associated bacteria are inconsistent, and the relationship between glucose metabolism and microbiome remains unclear 1 , 11 , 12 , 14 . This inconsistency is likely attributed to confounding factors, such as diabetes duration, environmental factors, sample collection, and analytical methods 1 , 14 , 30 . By contrast, patients were treated with a hospital‐based glycemic control regimen in this study, and no significant differences in PlI, indicating oral hygiene level, were observed between the T2DM and control groups. Moreover, our previous study showed that the duration of diabetes was not significantly different between the Imp and N‐Imp groups 10 . These factors minimized confounders and allowed for direct assessment of the impact of diabetes treatment on subgingival plaque composition. In the OPLS‐DA model for bacterial species in the control and T2DM before treatment groups, five bacterial species were identified as prominent targets for discrimination, characterized by high VIP values, a large magnitude of change, and strong reliability. Among these, F. nucleatum, P. gingivalis, Parvimonas micra, and T. forsythia contributing to the T2DM before treatment group are observed in severe periodontal disease, whereas S. oralis associated with the control group is reported as a prevalent species in healthy sites 31 , 32 , 33 . Additionally, when comparing the control and T2DM before treatment groups, the relative abundance of P. gingivalis, which plays a key role in the shift from symbiosis to dysbiosis 2 , 6 , and the PISA/tooth value were significantly higher in the T2DM before treatment group. These findings suggest that the microbiome in the T2DM pretreatment group tends to be more dysbiotic and to damage periodontal tissue more severely than in the control group. Biofilm formation detected before diabetes treatment in the T2DM group may have been influenced by advanced glycation end products (AGEs). Hyperglycemia promotes AGEs accumulation in tissues and cells, which induces free radical production and subsequent oxidative stress 34 , 35 . Under oxidative stress, F. nucleatum and red complex species increase resistance by upregulating stress response proteins, chaperones, regulators, and antioxidants to lower reactive oxygen species and limit damage effects 36 . Porphyromonas gingivalis is known to coaggregate with F. nucleatum and P. intermedia by outer membrane protein FomA and the gingipain–adhesin complex, respectively 37 , 38 . Moreover, oxidative stress may influence other bacterial interactions. A previous experiment with Streptococcus mutans (S. mutans), S. oralis, and A. naeslundii showed that only high hydrogen peroxide‐tolerant S. mutans strains increased and outnumbered the other two 39 . Additionally, S. mutans produces mutacins inhibiting S. sanguinis 40 , 41 , potentially explaining its lower prevalence in the T2DM group. Therefore, we hypothesize that, under oxidative stress in diabetes, bacteria with high oxidative stress tolerance accounted for a significant proportion of subgingival plaques in the T2DM group, affecting other bacterial interactions. Notably, these bacteria may exacerbate diabetes. Porphyromonas gingivalis proteolytically degrades insulin receptors through gingipain, leading to insulin resistance 42 . Additionally, T. forsythia produces methylglyoxal, which is a precursor of AGEs, which promotes oxidative stress 43 , 44 . Observing changes before and after treatment revealed that diabetes treatment alone altered oral bacterial composition, suggesting its potential to mitigate periodontal disease. The relative abundances of C. gracilis and C. sputigena increased significantly, whereas those of E. nodatum, P. intermedia, S. intermedius, C. concisus, and Eubacterium spp. decreased significantly in the T2DM group. Gingival crevicular fluid derived from serum is a primary nutrient source for subgingival plaque 17 , 18 . Previously, we observed an increase in several amino acids and a reduction in triglycerides in plasma following diabetes treatment 15 , 16 . Subgingival bacteria, including Fusobacterium, Prevotella, and Porphyromonas, can degrade and ferment amino acids to produce energy 17 , 45 . An increase in plasma amino acids may have influenced the composition of subgingival bacteria; however, F. nucleatum and P. gingivalis did not exhibit significant changes. Additionally, evidence supporting the potential role of fat in oral microbiome composition is lacking 46 . This relationship between plasma metabolites and subgingival bacteria likely involves bacterial interactions and exposure to bacterial metabolites. Further research is needed to clarify how plasma metabolite changes affect gingival cervical fluid and subgingival bacteria. The five bacterial species that decreased are linked to promoting inflammation 47 , 48 , 49 , 50 , 51 , 52 . A reduction in these bacteria may help reduce inflammation, potentially improving the PISA. By contrast, C. gracilis and C. sputigena are considered opportunistic pathogens and have been identified in both healthy and periodontitis‐affected sites 50 , 53 . The balance of these bacteria in dental plaques may be associated with improving periodontal inflammation. Compared to the N‐Imp group, the Imp group had different bacterial species whose abundance changed before and after diabetes treatment. This suggests that different bacterial shifts may relate to PISA improvements. In the Imp group, the relative abundance of S. intermedius and C. concisus, which promote inflammation, decreased significantly 49 , 50 . Additionally, a reduction in V. parvula, a key early colonizer and central to the metabolic interactions within the microbial community, was observed in the Imp group 54 , 55 . By contrast, in the N‐Imp group, only the relative abundance of P. intermedia, a late colonizer associated with inflammation, decreased significantly 48 , 56 . These findings suggest that patients in the Imp group experienced reduced pathogenicity of the biofilm and a greater change in biofilm formation before and after treatment compared with the N‐Imp group. This finding supports the hypothesis that alterations in the oral microbiome affect periodontal disease in diabetes 10 , 11 . Regarding C. gracilis and S. constellatus, which increased in the Imp group, their pathogenic roles remain poorly understood 50 , 57 . Notably, the relative abundance of S. sanguinis, linked to healthy microbiomes, was lower in the Imp group before treatment but increased significantly afterward 41 , 58 . This observation suggests that the oral microbiome in the Imp group had high pathogenicity before treatment but shifted toward a composition comparable to a healthy microbiome, leading to improved PISA. However, S. sanguinis can also serve as a binding target for pathogen localization within dental plaque 41 , 59 . Further research is needed to clarify its dual role in periodontal disease. The composition of subgingival plaques in the Imp group after diabetes treatment showed a higher proportion of T. denticola (red complex), F. nucleatum (orange complex), and Eubacterium spp. and a lower proportion of S. sanguinis (yellow complex) and A. naeslundii than that in the N‐Imp group. This result shows the relative abundance of bacterial species; the total absolute amount in the subgingival microbiome was not observed. This study has some limitations. The small sample size, limited by time‐consuming exams, constrained multivariate models to analyze the relationship between diabetes‐related factors and PISA improvements. Second, some patients had poorly controlled diabetes before glycemic control treatment, limiting the generalizability of the results to typical clinical settings. Additionally, the age distribution differed between the T2DM and control groups. However, some studies show that aging does not influence the composition of subgingival biofilms 13 , 60 . Further research with a larger sample size is needed to clarify the relationship between diabetes, inflammation, the microbiome, and its metabolites. Overall, only diabetes treatment changed the oral microbiome. The bacteria with significant abundance changes differed between the Imp and N‐Imp groups. These bacteria could help understand the interaction between diabetes and periodontal disease, offering new insights into treatment. Diabetes management should be considered a strategy for periodontal disease in diabetic patients. DISCLOSURE The authors declare no conflict of interest regarding the research, authorship, and/or publication of this article. Approval of the research protocol: The protocol for this research project has been approved by a suitably constituted Ethics Committee of the institution and it conforms to the provisions of the Declaration of Helsinki: The Human Ethics Committees of the Osaka University Hospital, No. 16374‐6; Osaka University Graduate School of Dentistry; and Osaka University Dental Hospital, No. H28‐E40‐2. Informed consent: Informed consent was obtained from all the participants. Registry and the registration no. of the study/trial: N/A. Animal studies: N/A. Supporting information Table S1. Oral bacteria with the highest VIP values (>1.0). VIP values were obtained from the OPLS‐DA model comparing the control and T2DM before treatment groups. The bacterial species whose VIP values were greater than 1 were listed. VIP values greater than 1 indicate relevance for group discrimination. Abbreviation: VIP, variable influence on projection. Table S2. Oral bacteria related to periodontal disease. The criteria for dividing bacterial species into groups were based on reports by Socransky et al. Abbreviations: red, red complex; orange, orange complex; yellow, yellow complex; green, green complex; purple, purple complex; blue, blue complex; accessory, accessory pathogen. Method S1. DNA extraction by bead‐beating method. Method S2. Full‐length 16S rRNA gene sequencing analysis.