Faecal D/L Lactate Ratio Is a Metabolic Signature of Microbiota Imbalance in Patients with Short Bowel Syndrome

Our objective was to understand the functional link between the composition of faecal microbiota and the clinical characteristics of adults with short bowel syndrome (SBS). Sixteen patients suffering from type II SBS were included in the study. They displayed a total oral intake of 2661±1005 Kcal/day with superior sugar absorption (83±12%) than protein (42±13%) or fat (39±26%). These patients displayed a marked dysbiosis in faecal microbiota, with a predominance of Lactobacillus/Leuconostoc group, while Clostridium and Bacteroides were under-represented. Each patient exhibited a diverse lactic acid bacteria composition (L. delbrueckii subsp. bulgaricus, L. crispatus, L. gasseri, L. johnsonii, L. reuteri, L. mucosae), displaying specific D and L-lactate production profiles in vitro. Of 16 patients, 9/16 (56%) accumulated lactates in their faecal samples, from 2 to 110 mM of D-lactate and from 2 to 80 mM of L-lactate. The presence of lactates in faeces (56% patients) was used to define the Lactate-accumulator group (LA), while absence of faecal lactates (44% patients) defines the Non lactate-accumulator group (NLA). The LA group had a lower plasma HCO3− concentration (17.1±2.8 mM) than the NLA group (22.8±4.6 mM), indicating that LA and NLA groups are clinically relevant sub–types. Two patients, belonging to the LA group and who particularly accumulated faecal D-lactate, were at risk of D-encephalopathic reactions. Furthermore, all patients of the NLA group and those accumulating preferentially L isoform in the LA group had never developed D-acidosis. The D/L faecal lactate ratio seems to be the most relevant index for a higher D- encephalopathy risk, rather than D- and L-lactate faecal concentrations per se. Testing criteria that take into account HCO3− value, total faecal lactate and the faecal D/L lactate ratio may become useful tools for identifying SBS patients at risk for D-encephalopathy.


Introduction
Micro-organisms colonizing the digestive tract (microbiota) play a key role in nutrition and health. The diversity and abundance of this microbial population makes it difficult to define a ''healthy intestinal flora''. In general, adult human microbiota is mainly composed of a phylogenic core containing Firmicutes, Bacteroidetes, and Actinobacteria [1,2]. It is well known that the composition and mostly, molecular functions of microbiota are altered by age, nutrition, environment and pathologic status [3,4,5,6]. Dysbiosis is characterised by a divergent ratio between the major groups and/ or by the emergence of specific microorganisms. The protective or harmful role of bowel microbiota depends on the pathology, and thus microbiota specifically influences the aetiopathogenesis of each human disease.
In patients with type II short bowel syndrome (SBS), the aetiopathogenesis has been associated with an extreme dysbiosis [7,8,9,10]. These patients have had a massive small bowel resection, leaving less than 150 cm joining with a partial or an entire colon [11]. These patients have such a reduced intestinal surface that they suffer from major water and nutrient malabsorption, requiring oral and parenteral nutrition [12,13,14,15]. Microbiota of SBS patients is markedly disturbed with a high prevalence of Lactobacillus/Leuconostoc group along with a drop in numbers and a poor diversity of Clostridium leptum, Clostridium coccoides and Bacteroides [7,9]. Bacterial groups abundant in healthy humans nearly disappear in SBS, while Lactobacillus/Leuconostoc group, sub-dominant in healthy controls, become over-represented in SBS [9]. In addition to these unusual microbial features, we previously detected a SBS-associated species, Lactobacillus mucosae which has never been found in healthy controls [9]. Intestinal resection probably favours the Lactobacillus/Leuconostoc group and the emergence of L. mucosae because they are aerobic-tolerant micro-organisms and moreover these bacteria resist the acidic environment and nutrient load, using mal-absorbed nutrients as an energy source. The Roux-en-Y gastric bypass (RYGB), a current surgical treatment for morbid obesity, is also associated with changes in microbial composition and activities [16,17]. Short bowel syndrome and gastric bypass illustrate how gut remodelling due to surgery impacts the microbiota that in turn becomes central to understanding the clinical outcome.
In SBS patients, it has been shown that the presence of colon determines decreased dependence on parenteral nutrition. The number of calories provided by microbiota in healthy humans was up to 150 Kcal [18]; since the colonic digestion via microbiota activity may provide an energy supply of up to approximately 1000 kcal per day in SBS [19,20,21]. The remaining colon adapts following resection by increasing cell numbers and crypt depth. It plays a key role to salvage energy via the fermentative activities of endogenous microbiota [22]. Some SBS patients develop severe, symptomatic D-lactic acidosis, with metabolic acidosis and neurological disorders. The D-lactic acidosis is linked to both the metabolic activity of microbiota and the individual ability to metabolize D-lactic acid [23,24,25]. The occurrence of D-lactic acidosis remains sporadic and non-predictable. D-lactic acidosis may be more common than is thought and should be looked for in cases of metabolic acidosis in which the cause of acidosis is not apparent. The clinical presentation is often characterised by episodes of unusual neurological manifestations and severe metabolic acidosis. Plasmatic bicarbonate assays are done routinely to monitor SBS patients and evaluate ionic balance. But, the relationship between neurological disorders, microbiota composition, fermentative activity, metabolism and systemic acidosis has not been yet completely understood [7,26]. Moreover, a survey of SBS patients is difficult since measurement of serum Dlactate concentration is not routinely available in most hospital laboratories or the results come back after the D-lactic acidosis crisis [27,28]. Our objective was to understand the functional link between the composition/metabolic activity ratio of the faecal microbiota and the SBS clinical characteristics.

SBS Patients
Ethics statement: The Human Investigations Committee of Hospital Saint-Louis in Paris approved the study (n: 031048) in January 2004. All patients gave their written informed consent to participate in this study.
This prospective study was performed between January 2006 and January 2011. Patients were enrolled and monitored in our nutrition support unit, an approved centre for intestinal failure and Home Parenteral Nutrition, located at Lariboisiere and Beaujon hospitals (Paris and Clichy respectively). Clinical and demographic information were collected from the case records. Data included: sex, age, and medical history, SBS aetiology, characteristics of past surgical procedures, and characteristics of intestinal tract. All patients received gastric anti-secretory drugs (IPP) but none had pancreatic enzyme treatment and cholestyramine.

Inclusion criteria
(a) SBS type II with a post duodenal remnant jejunum ,150 cm, no remnant ileum and re-established jejuno-colic continuity for more than 2 years. (b) a non restricted oral diet. sepsis and/or antibiotics or aspirin in the 4 weeks before the study. (k) treatments with steroids or immunosuppressive therapy in the 4 months preceding the study. (l) pregnancy (m) a D-lactic encephalopathy episode in the 3 months preceding the study.
The remaining colon in continuity, expressed in terms of percent of the usual length according to the method of Cummings et al., was estimated on the basis of per-operative records [29].
General procedures. Following patient screening, eligible patients were invited to participate in the study during their regular out-patient check up. When patients had given their informed consent, biological analyses were performed.

Dietary
The patients were asked to maintain their usual intake. Under the supervision of trained dieticians, patients noted the type, amount or weight of foods they ingested at each meal, including snacks and nibbling for 3 days during the week preceding sample collection. The daily intake of total calories, proteins, carbohydrates (total, complex and simple), fat, fibre and alcohol were calculated with BilnutH software (Bourgerette P, Rolshansen M. BILNUT 4.0. SCDA Nutrisoft. Le Hallier 37390 Cerellers, France). The mean intake for each macronutrient and fibre were expressed as the percentage of total g/day oral intake.

Absorption study
Over a 6-day period, the first 3-day baseline period confirmed that patients were continuing their spontaneous intake of energy, carbohydrates, fat, protein and fibres. Absorption was measured during the following 3 days. Stools were collected daily and frozen at 220uC. Protein, lipid, and total energy were determined by nitrogen elemental analysis (N analyser Flash EA1112, Thermo Scientific, Waltham, MA) [30], and the method of Van de Kamer and bomb calorimetry (PARR 1351 bomb calorimeter, Parr instrument company, Moline, IL) [31], respectively. Quantification of carbohydrate-derived energy was calculated by subtracting the energy associated with the protein and lipid components from the total energy. The calorie-conversion factors used were 4.2, 9.35 and 5.65 kcal/g for carbohydrates, lipids and proteins respectively [32]. The coefficient of net intestinal absorption represented the proportion of ingested energy not recovered in stool output. REE kcal/day (resting energy expenditure): basal metabolic rate assessed with Harris and Benedict equations 61.5 required to obtain energy balance equilibrium in patients with short bowel syndrome.

Biological analysis
Routine biological data included serum albumin, transthyretin, electrolytes and C Reactive Protein (CRP). The anion gap was calculated from measured electrolytes as the sum of serum chloride and bicarbonate concentrations subtracted from the sodium and potassium concentrations [(Na + +K + )-(Cl 2 +HCO3 2 )]. Faecal sam-ples (200 mg fractions) were stored in sterile tube at 280uC until further processing.

D-encephalopathy description
We considered a patient having a past history of D-lactic encephalopathy when characteristic clinical presentation of recurrent episodes of encephalopathy was associated or not with metabolic acidosis. The symptoms and signs included altered mental status, slurred speech, ataxia, gait disturbance, confusion, weakness, and/or impaired motor coordination. Hostile, aggressive, and abusive behaviour as well as inability to concentrate are also characteristic neurologic features. Nystagmus is sometimes a sign of the disease. The correction of neurological symptoms by fasting is an additional diagnostic element.

D and L-Lactate analyses on in vitro Lactobacillus cultures and faeces
D and L-lactates were measured from different Lactobacillus species in vitro cultures and directly from stools. The medium used for the pre-cultures and cultures of Lactobacillus was MRS supplemented with 0.5 g/l Cys. A preculture of each species was platted on solid medium and then, individual colony of each strain was grown in MRS-Cys at 37uC in anaerobic conditions (Anaerocult A) during 24 hours. D and L-lactates were measured on 24 h in vitro culture from each strain using 1 ml of culture supernatant firstly precipitated with TCA 6 N (10%). For faecal dosage, 100 to 250 mg frozen stools were 2 fold diluted in 0.1 M triethanolamine buffer (pH 9.15) and precipitated with TCA 6 N (10%). D and L-Lactates were measured in TCA-supernatants using Biosentec D/L lactic acid enzymatic kit according to manufacturer instructions (Biosentec, Toulouse, France) and as we described previously [33].

Identification of bacterial groups that were quantified by real-time qPCR
The dominant groups (C. leptum, C. coccoides, Bacteroides/Prevotella, Bifidobacterium, E. coli and Lactobacillus/Leuconostoc) were quantified by real-time qPCR using primers and probes (Table 1), as previously described [34,35,36,37]. Amplification efficiencies in real-time qPCR of all taqman and sybergreen systems were reported in Table 1.
The Lactobacillus species, quantified by real-time qPCR was chosen following our previous results obtained by sequencing a ''Lactobacillus/Leuconostoc'' amplicon. This amplicon was obtained by amplifying faecal DNA with primers of the V6-V8 region designed on the 16S ribosomal DNA from Lactobacillus/ Leuconostoc group. Primers (Lac 352-f and Lac 679-GC-r) and PCR conditions used were described previously [9,38,39] The PCR amplicon was cloned into TOPO4 (invitrogen, Cergy-Pontoise, France) and 40 clones were sequenced. Sequences were compared to known sequences in the Ribosomal Database Project using Blast2seq and Seqmatch programs. The amplicon sequences detected with high blast hits (95 to 99% identity) were : L. delbrueckii subsp. bulgaricus, L. crispatus, L.gasseri, L. mucosae, L. reuteri.

Lactobacillus-specific primer design
Primers targeting different Lactobacillus species (L. mucosae, L. delbrueckii subsp. bulgaricus, L. crispatus, L. gasseri/johnsonii, L. reuteri) were based on specific region of the 16S ribosomal DNA sequences aligned using the Blastn program of NCBI database and design using Primer-express Version 2.0 (Applied-biosystems) ( Table 1). Primer specificities were challenged in silico against sequences in Ribosomal Database project II using the probe match program. Primer specificities were then firstly assessed by real-time qPCR amplification against DNA from dominant faecal species or closely related species and secondly using PCR amplification followed by visualisation on agarose gel (Table S1). Amplification and detection were carried out in 96-well plates with Syber-Green PCR 26 master mix (Applied-Biosystems).

Experimental conditions of real-time qPCR
Total bacterial DNA was extracted from 0.15 g to 0.2 g of faecal samples using a bead-beating method G'NOME DNA extraction kit (Mp Biomedicals, Illkirch) as described by Furet et al [34]. Faecal DNAs were dissolved in 150 ml Tris-EDTA buffer and stored at 220uC.
The real-time qPCR was performed using the ABI7000 sequence detection system apparatus and 7000 system software version 1.2.3 (Applied-Biosystems). Amplification and detection were carried out in 96-well plates with Taqman universal 26PCR master mix or with Syber-Green PCR 26 master mix (Applied-Biosystems). Each reaction was run in duplicate in 25 ml with 0.2 mM primers, 0.25 mM of each probe and 10 ml DNA samples. Amplifications were carried out using the following profile: 1 cycle at 95uC for 10 min, followed by 40 cycles of 95uC for 30 s, 60uC for 1 min. For syber-green amplification, a melting curve was added to show the amplification specificity. Amplification efficiencies were listed in Table 1. No qPCR inhibition was detected using 10 23 dilution of faecal samples using the Taqman exogenous internal positive control (Applied Biosystems), thus we used this dilution to perform all qPCR amplifications from faecal samples.
Bacterial genomic DNA of each reference strain (Table 1) was extracted as for faecal DNA and used to generate standard curves. Each reference strain was numbered from dilutions of each genomic DNA extract using an ''All bacteria-qPCR'' reaction with an E. coli standard curve (10-fold serial dilutions of genomic DNA from a culture numbered E. coli suspension). New DNA standard curves were obtained for each reference strain DNA (10-fold serial dilutions). Standard curve was generated by plotting cycle threshold (Ct) versus cell number of reference strain in the qPCR assay. Using cycle threshold (Ct) values in the linear range of the standard curve, faecal bacterial cells were interpolated from each standard curve generated in the same experiment. Then, results were expressed in bacterial cell/g faeces as mean and standard errors. When results were expressed in percentage of total bacteria, bacteria (expressed in bacterial cell/g faeces) was compared to all bacteria (expressed in bacterial cell/g faeces) and expressed as %.

Statistical analysis
Data were expressed as mean 6 standard deviation or percentage. Clinical and microbial data were compared between LA (Lactate Accumulator) and NLA (Non Lactate Accumulator) groups with Fisher exact test for qualitative variables and Mann Whitney test for quantitative variables. All statistical tests were two-sided. The critical level of statistical significance was set at p,0.05. Data were analyzed with the SAS 9.1 statistical software for windows (SAS Institute Inc., Cary. NC).

SBS patient cohort
All 16 patients had undergone a major resection including a part of the jejunum, the entire ileum and part of the colon ( Table 2). Aetiology of SBS was mainly venous/arterial mesenteric infarction (69%, n = 11/16). The study was performed 1068 years following re-establishment of bowel continuity.
Oral intake and macronutrient repartition are listed in Table 3. SBS patients had a total oral intake of 266161005 Kcal/day. The faecal wet weight was 236761140 g/day. The carbohydrate absorption (83612%) was better than for proteins (42613%) and fat (39626%). BMI was 20.963.2 kg/m 2 and levels of albumin, transthyretin protein and protein C reactive were normal (data not shown). Although S12 and S11 patients were known to be at risk for D-encephalopathy recurrent crisis, they did not present any clinical symptoms at the time of sampling.
To determine the species in the SBS-related population; we sequenced the amplicon obtained by PCR from faecal DNA amplification by using universal Lactobacillus/Leuconostoc primers. This amplicon was mainly composed by L. delbrueckii subsp. bulgaricus, L. crispatus, L. gasseri, L. johnsonii, L. reuteri, L. mucosae. Therefore, we designed specific qPCR conditions and primers (Tables 1 and S1) for these Lactobacillus species. In all SBS patients, these species were indeed detected with a high inter-individual variability ( Figure 1B). L. mucosae that we previously described in SBS patients [9], was present from 1.2610 2 to 2.0610 9 bacterial cell/g. Over a period of 8 months, faecal microbiota was stable for both dominant groups and the different species of Lactobacillus in S11 patient ( Figure S1A and S1B). L. mucosae was detected at similar level in S12 patient over 4 years (data not shown). Thus, the overall composition of faecal microbiota seems to be specific for each patient and rather stable over time.
Each Lactobacillus species identified were characterized for its specific fermentative activity. It was monitored in vitro by the total amount of D and L-lactates expressed in percentage of total lactates ( Figure 1C) or expressed in mM in the in vitro culture medium ( Figure S2). L. delbrueckii subsp. bulgaricus synthesized in vitro specifically D-lactate (up to 70 mM). L. mucosae produced in vitro 30 mM D-lactate and 57 mM L-lactate, which respectively accounted for 35 and 65% of total lactate. Thus, each patient was characterized by the diversity of lactic acid bacteria in their faeces, and each lactic acid bacteria had a specific lactate production profile in vitro ( Figure 1C).

Faecal lactates amount in SBS patients
The total amount of faecal lactates, the ratio of D/L lactates isoforms, faecal pH and the plasma concentration of HCO3 2 are presented for each patient in Figure 2. Faecal pH values were in acidic range from 4.4 to 6.8. Of 16 patients, 9/16 (56%) accumulated lactate in their faecal samples (S12, S3, S10, S11, S8, S4, S9, S1, S7) (Figure 2A). In the faeces, the total amount of D-lactate ranged from 2 to 110 mM and the L-lactate ranged from 2 to 80 mM. The presence of lactates in faeces was used as criteria to define the Lactate-accumulator group (LA). In this LA group, 6/9 (67%) patients preferentially accumulated the D lactate isoform (with a D/L lactate ratio .1). In S12, we observed over 3 years a high amount of faecal D and L lactate with a D/L lactate ratio .2 (Table 4). S12 and S11, which are known to be at risk for D-encepahlopathy belong to the LA sub-group. S12 and S11 had a D/L lactate ratio above 2, and their plasma HCO3 2 concentrations were respectively 17 and 13 mM (Figure 2). S13, S2, S6, S14, S15, S16, S5 had too lower level of lactates to be detected in their faecal sample, this sub-group (7/16, 44%) was defined as the non-lactate-accumulating group (NLA).

Comparison between LA and NLA sub-groups
Between both LA and NLA groups, there were no significant differences of all clinical data listed in Tables 2 and 3, such as sex, age, remnant small bowel length, remnant colon and nutritional parameters such as oral intake, caloric and nutrients absorption and home parental nutrition supply. Although it tended to be higher in the LA group, the percentage of faecal total Lactobacillus was not statistically different in LA and NLA groups (54.9623.1% vs 30.1619.1%) (Figure 3). Faecal pH (5.560.8 vs 5.560.4), faecal weight (24256955 g vs 212261426 g), and the anion gap (13.862.7 vs 12.663.4) were not statistically different in LA and NLA groups (Figure 3). On the other hand, the LA group had a lower concentration of plasma HCO3 2 (17.162.8 mM) than the NLA group (22.864.6 mM) (p,0.01). In addition, the faecal Dlactate and L-lactate amounts were inversely correlated to plasma HCO3 2 in the entire cohort of 16 patients (respectively spearman correlation 20.50, p = 0.04 and 20.53, p = 0.03). We did not find any correlation between faecal D or L-lactate amount and the different faecal bacteria except for L. mucosae that was significantly (p = 0.03) more present in the NLA (12.7611.7% of Lactobacillus/ Leuconostoc) vs LA group (2.864.7% of Lactobacillus/Leuconostoc) ( Figure 3).

Discussion
We recruited 16 type II SBS patients and described their nutritional and pathophysiologic status, their absorptive capacity, the composition of their faecal microbiota, the amount of faecal D/L lactate and routine serum values.
In healthy human stools, we have early described that microbiota is characterized by high dominant anaerobic bacteria (C. leptum 4.1610 9 63.9610 9 bacterial cell/g faeces, C. coccoides 3.1610 9 62.9610 9 bacterial cell/g faeces and Bacteroides/Prevotella 7.9610 8 67.8610 8 bacterial cell/g faeces) and sub-dominant bacteria (Bifidobacteria 2.8610 8 62.7610 8 bacterial cell/g faeces, Lactobacillus/Leuconostoc 3.7610 7 63.4610 7 bacterial cell/g faeces, and Enterobacteria) [9]. In contrast, SBS patients harboured a specific faecal microbiota that we called ''lactobiota'' because it is enriched in the Lactobacillus/Leuconostoc group and depleted in anaerobic micro-organisms (especially Clostridium) [7,8,9]. The number of patients enrolled in this study was comparable to our previous work [9,22]. In both cohorts (see [9,22] and Tables 2 and 3), SBS patients displayed similar clinical characteristics: they adopted a hyperphagic behavior (intake/REE.1) and they had similar absorptive capacities with better absorption of carbohydrates. Although stool samples were collected after a PEG treatment in the first cohort [9,22] and without a PEG treatment in this study, we confirmed a shift in microbial communities between dominant and sub-dominant groups in SBS patients. Compared this cohort with that previously described [9,22], both cohorts presented similar percentage of microbial groups and the emergence of L. mucosae in their lactobiota.
Lactobiota plays a crucial role in the digestion process mainly through Lactobacillus-related fermentative activities [40]. Although all patients are rich in Lactobacillus/Leuconostoc group, only 56% displayed an accumulation of faecal lactate. We thus defined two sub-groups of SBS patients: those accumulating lactate (LA) and those who did not accumulate lactate (NLA). In the cohort  Table 4. Faecal D and L lactate concentrations (mM) in S12 patient over time from sequential sampling (T0, T0+3 years and T0+3.5 years). previously studied [9,22], we also measured faecal lactate. Of 12 patients described in Joly et al [9,22], 6 (50%) accumulated D and L-lactates in their faeces with a total amount of D-lactate ranging from 6 to 25 mM and the L-lactate ranging from 3 to 23 mM. LA and NLA groups of SBS patients are found in our both distinct cohorts. Since faecal L mucosae is less present in the LA subgroup, we could hypothesize that LA and NLA subgroups defined specific and different ecologic niches. Bustos et al. also described relatively high concentrations of faecal D-and L-lactate in some paediatric and adult SBS patients, who did not develop metabolic D-lactic acidosis [41]. Thus, LA and NLA groups of SBS that are also distinguished by their mean serum HCO3 2 , may constitute clinically relevant sub-types revealing two distinct metabolic signatures. Lactate is produced both by eukaryotic and prokaryotic cells as a final metabolite of fermentation. In healthy humans, lactate produced as a key intermediate for the overall gut fermentation process isn't accumulated in faeces, it is absorbed by intestinal cells [42] or is used by lactate consuming bacteria [43,44,45,46]. We found, coherently with the literature, that faecal D-lactate was undetectable in healthy human faeces and the faecal L-lactate was around 1.5 mM [41,47,48]. D-lactate enantiomer is toxic for neurons in a manner that is independent of acidosis [49] while Llactate is metabolized at high level by neurons. D-and L-lactate share similar pathways of transport and metabolism in the digestive tract, but in human brain tissue, D isoform seems to impair the L lactate availability, thus provoking energy deficit in neurons [23,50]. In SBS, the lactate accumulation in faeces could result, non-exclusively from 1) a predominance of lactateproducing bacteria over lactate-consuming ones, 2) an intense fermentation activity of lactobiota, 3) an imbalance between microbial production and intestinal absorption 4) and individual capacity of lactate metabolism. Lactate concentration in the colon depends mostly on microbial formation (main input) and microbial utilization or absorption from the host (main output). So, a lactate accumulation reflects an imbalance among these factors. As Duncan et al reported [51], D-L lactate induce .90% growth inhibition of dominant bacteria (such as B. thetaiotaomicron) at low pH, so the link between the accumulation of faecal lactate and the composition of lactobiota remains to be deeply explored by a deep sequencing. It could also be interesting to specifically quantify lactate consuming bacteria, such as Eubacterium hallii, Anaerostipes caccae, Coprococcus catus, bacteria related to Megasphaera elsdenii, Eubacterium limosum and Anaerostipes coli [52,53,54,55,56]. All these lactate consuming bacteria (mainly Clostridium cluster XIVa), representing around 10 7 bacterial cell/g faeces (2 to 3% of the total bacteria) in healthy adult humans are highly oxygen-sensitive anaerobes [45,52,56]. In our study, C. coccoides group represented only 10 4 bacterial cell/g faeces in microbiota of all SBS patients (Figure 1), thus, these lactate consuming bacteria were under the detection threshold by PCR amplification (note that in our PCR conditions, the probes C. coccoides detect 3/6 of these lactate consuming bacteria [34]). In SBS patients, considering the short length of remnant small intestine and colon, it can be supposed that the level of oxygen might be too high to favor growth of strict anaerobic bacteria. The low faecal pH of SBS patients may also participate to the selection of growth and activities of some lactate consuming bacteria [57]. Even if lactate consuming bacteria were present in extremely low amount, would they be sufficiently active to decrease faecal lactates amount? In this context, it could be informative to evaluate, in vitro, the potential of lactate metabolism from fresh faecal microbial population in SBS patients.

Faeces
In the LA sub-group, the total amount of lactate and the D/L ratio varied in each individual. S8 and S9 patients had the higher total amount of lactate, but the ratio between D/L was low (,1). On the contrary S12, S3, S10 and S11 preferentially accumulated D-lactate with a D/L lactate ratio over 2. This ratio remained over 2 during three years in S12 patient. Compared to healthy humans where plasma D-lactate concentration was undetectable (personal communication and [25]), plasma D and L-lactate concentrations in S12 patient were respectively 0.3960.03 mM and 2.2660.06 mM, and were stable overtime without D acidosis crisis (+3.5 years from this study). S12 and S11 had no D-acidosis crisis at the moment of the faecal sampling, but they were known to be at high risk for D-lactic acidosis. So, S12 and S11 patients, who preferentially accumulated D lactate in faeces, were at risk of D-encephalopathy crisis. We observed in our study an inverse correlation between plasma HCO3 2 and faecal lactate levels. Furthermore, all patients who did not accumulate lactate and those who accumulate preferentially L isoform did not develop Dacidosis. In our previous study [9,22], patients having a D/L ratio under 1 did not either develop D-encephalopathy (data not shown). D-lactate accumulation in faeces seems to indicate a microbiota dysbiosis that may be associated with a higher risk of metabolic acidosis and D-encephalopathy.
The D/L faecal lactate ratio seems to be the most relevant index first for an imbalanced microbiota in SBS patients and second for a higher D-encephalopathy risk, rather than D-and L-lactate faecal concentrations per se. Because D lactate amount in blood is not routinely dosed at hospital, we emphasize that the faecal amount of D and L lactate will be a useful tool to detect disturbed activity or composition of intestinal faecal microbiota after intestinal surgery in clinical management of short bowel syndrome or intestinal bypass [24]. To be more informative on the in vivo metabolic efflux, in the future, we will study the D-and L-lactate in faeces, blood and urine, in a longitudinal study, in a larger population of SBS patients who develop recurrent episodes of Dencephalopathy.

Conclusions
We put in evidence a connection between intestinal dysbiosis and metabolite changes in the faeces of SBS patients. The plasma HCO3 2 value, the faecal lactate amount and the faecal D/L lactate ratio constitute clinically relevant parameters that allow a classification of two sub-types of SBS patients: those having no detectable lactate in faeces (NLA sub-group), and those accumulating faecal D and L lactate in faeces (LA sub-group). In the LA sub-group, patients with high faecal D/L lactate ratio and low plasma HCO3 2 value were at risk to develop recurrent Dencephalopathy episodes. All these criteria may be useful for identifying nutritional strategies to prevent or alleviate the deleterious consequences of D lactate accumulation in faeces in short bowel disease. Figure S1 Faecal microbiota of S11 patient over time from sequential collection (T0, T0+7 months and T0+8 months). A-Dominant faecal microbiota in S11 patient was stable over 8 months; B-Lactobacillus species were relatively stable in S11 faeces over 8 months. (TIF) Figure S2 In vitro D-and L-lactate production of different reference species of Lactobacillus, during 24 hours. Results were expressed in mM.

(TIF)
Table S1 Specificity of Lactobacillus specific primers used in this study. Specificity of primers designed for this study was tested by performing: i) real-time qPCR on related and unrelated strains using a dilution 100 for DNA. Each primer set was named for its intended target bacteria. +++, Ct,20; ++, Ct,25; +, Ct.25; +/2, Ct.30; 2, no Ct detected; nd , not done; and ii) PCR amplification against all groups and species and detection of the amplicon on an 1.5% agarose gel. (DOC)