Ethical approval for this study was obtained from the Institutional Review Board (Comité de Ética de Investigación Clínica del Principado de Asturias) in compliance with the Declaration of Helsinki. All participants were informed and gave a signed informed consent prior their inclusion in the study.

Our study involved 21 SLE patients, all fulfilling classification criteria for SLE. According to the 1982 revised criteria from the American College of Rheumatology, a definitive SLE diagnosis can be established when a patient exhibit at least 4 out of the 11 SLE criteria (malar rash, discoid lesions, photosensitivity, oral ulcers, arthritis, serositis, renal disorders, neurological disorder, cytopenia, raised anti-DNA titers, and positivity to antinuclear antibodies) (23). A complete clinical examination, including Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) calculation and anti-dsDNA autoantibodies assessment, was performed at the time of sampling. All patients were in remission (SLEDAI <8) at the sampling time. Information on clinical features along the disease course as well as the therapies received during the last 6 months was obtained from their clinical records. A group of 25 age- [median 43.50 (range 23.00–63.00) years] and gender-matched (23 females) healthy individuals recruited from the general population was included as the healthy controls (HC). Patients and controls did not differ in age (p = 0.293) and gender distribution (p = 0.495). Exclusion criteria were the history of recent infections, diagnosis of metabolic diseases, or the use of antibiotics, glucocorticoids, or monoclonal antibodies in the previous 6 months.

Upon acceptance of the individuals to participate in the study, a strict overnight fast (more than 8 h) blood sample was obtained in tubes without anticoagulant. Serum was collected, divided into aliquots, and samples were stored at −80°C until experimental procedures were performed. Additionally, basic serum blood lipid analyses were carried out on fresh samples at the time of sampling, by standardized procedures.

Total FFA serum levels were analyzed by a colorimetric enzymatic assay using a commercial kit (NEFA kit half-microtest, Roche Life Sciences, Penzberg, Germany) following the instructions from the manufacturer. Final absorbance was measured at 546 nm, and the detection limit was 0.2 mM.

Individual FFA were analyzed in serum samples following a methyl-tert-butylether-based extraction protocol (MTBE) as previously described (24), with minor modifications. Briefly, serum samples (100 µl) were spiked with 5 µl of internal standard (600 ppm heptadecanoic acid). Proteins were precipitated by the addition of 200 µl methanol chromasolv grade (Sigma Aldrich, MO, USA). Organic phases were obtained by the addition of 1,200 µl MTBE chromasolv grade (Sigma) followed by an incubation in an ultrasound water bath at 15°C for 30 min. Finally, organic phases were isolated by centrifugation at 5,000 rpm (7 min, 15°C) after the addition of 200 µl milliQ water. The extraction protocol was repeated once with 100 µl MetOH, 500 µl MTBE, and 100 µl milliQ H₂O. Lipid extracts were dried in a miVac centrifugal evaporator (Genevac Ltd., UK) and redissolved in 100 µl of water:acetonitrile (38:62).

The analyses of fatty acids in the samples were performed in a Dionex Ultimate 3000 HPLC system (Thermo Scientific, Bremen, Germany) equipped with a column Zorbax Eclipse Plus C18 (50 mm × 2.1 mm, 1.8 μm). Mobile phases A and B were water and acetonitrile, respectively, both containing 0.1% of formic acid. Fatty acids were separated in an injection volume of 2 µl by a gradient program as follows: 62% B (held for 4.5 min) followed by a linear increase up to 100% B in 10 min (held for 1 min). The column temperature was set at 45°C. Mass detection was carried out in a Bruker Impact II q-ToF mass spectrometer with electrospray ionization, operating in the negative mode. The settings of the mass spectrometer were as follows: spray voltage 4.5 kV; drying gas flow 12 l/min; drying gas temperature 250°C; and nebulizer pressure 44 psi.

For quantification, calibration curves for each compound were prepared by dissolution of the pure standards in methanol to adequately encompass the expected concentration of the analytes in the samples. The calibration ranges were as follows: 0.4–12.5 µg/ml for EPA and γ-linolenic; 1.2–37.5 µg/ml for DHA and linolenic; 2.3–75 µg/ml for AA and palmitoleic; 3.9–125 µg/ml for linoleic; and 7.8–250 µg/ml for oleic, palmitic, and stearic acids. A good linearity was observed in all cases (r² > 0.994). Heptadecanoic acid was used as internal standard to account for potential biases during the extraction protocol.

Fecal sample collection and metagenomic analyses of fecal microbiota were performed as previously reported (22). Briefly, fresh fecal material was processed within 3 h from collection and immediately homogenized and stored at −80°C. Fecal DNA was extracted with a QIAampDNA stool minikit (Qiagen, Strasse, Germany). Then, 16S rRNA gene sequences were amplified, and 16S rRNA and gene-based amplicons were sequenced by an Ion Torrent PGM sequencing platform as described elsewhere (22).

Analysis of SCFA (acetate, propionate, and butyrate) was performed by gas chromatography. Briefly, 1 g of fecal samples was diluted 1:10 in sterile PBS and homogenized in a LabBlender 400 stomacher (Seward Medical, London, UK) at full speed for 4 min. Then, supernatants were obtained by centrifugation (10.000 × g, 30 min, 4°C), filtered through 0.2-µm filters, mixed with 1:10 of ethyl butyric acid (2 mg/ml) as an internal standard, and stored at −80°C until analysis.

A gas chromatograph 6890N (Agilent Technologies Inc., Palo Alto, CA, USA) connected to a mass spectrometry (MS) 5973N detector (Agilent Technologies) and to a flame ionization detector was used for identification and quantification of SCFA. Data were collected using the Enhanced ChemStation G1701DA software (Agilent). Samples (1 µl) were injected into the gas chromatograph equipped with an HP-Innowax capillary column (60-m length by 0.25-mm internal diameter, with a 0.25-µm film thickness; Agilent) using He as a gas carrier (flow rate of 1.5 ml/min). The temperature of the injector was kept at 220°C, and the split ratio was 50:1. Chromatographic conditions were as follows: initial oven temperature of 120°C, 5°C/min up to 180°C, 1 min at 180°C, and a ramp of 20°C/min up to 220°C to clean the column. In the MS detector, the electron impact energy was set at 70 eV. The data collected were in the range of 25 to 250 atomic mass units (at 3.25 scans/s).

SCFA were identified by comparison of their mass spectra with those held in the HP-Wiley 138 library (Agilent) and by comparison of their retention times with those of the corresponding standards (Sigma Aldrich, St. Louis, MO, USA). The peaks were quantified as relative abundances with respect to the internal standard. The concentration (in millimolar) of each SCFA was calculated using the linear regression equations (R² ≥ 0.99) from the corresponding standard curves.

Height was measured using a stadiometer with an accuracy of ±1 mm (Año-Sayol, Barcelona, Spain). The subjects stood barefoot, in an upright position and with the head positioned in the Frankfort horizontal plane. Weight was measured on a scale with an accuracy of ±100 g (Seca, Hamburg, Germany).

Dietary intakes were assessed by means of an annual semiquantitative validated food frequency questionnaire including 160 items (25). During an interview by trained dietitians, subjects were asked, item by item, whether they usually ate each food and, if so, how much they usually ate. For this purpose, three different serving sizes of each cooked food were presented in pictures to the participants so that they could choose from up to seven serving sizes (from “less than the small one” to “more than the large one”). For some of the foods consumed, amounts were recorded in household units, by volume, or by measuring with a ruler. Information on the cooking practices, number and amount of ingredients used in each recipe, and other relevant information for the study was collected. Methodological issues concerning dietary assessment have been detailed elsewhere (25). The consumption of foods was converted into energy intake (kilocalories per day), macronutrients (carbohydrates, lipids, and proteins, grams per day), and total fiber (grams per day) using the nutrient food composition tables developed by the Centro de Enseñanza Superior de Nutrición Humana y Dietética (CESNID) (26). CESNID, a foundation, involves different institutions, universities, and companies related to the food and nutrition area, and its food composition databases are supported by the Spanish Association of Nutrition and Dietetics.

Continuous variables were expressed as median (interquartile range) or mean ± SD. Mann–Whitney U, Student’s t or Kruskal–Wallis tests were performed to assess statistical differences. Correlations were analyzed by Spearman’s rank or Pearson tests, depending on the distribution of the data. Categorical variables were summarized as n (%), and differences were analyzed by χ² tests. A principal component analysis (PCA) was performed to avoid potential bias due to collinearity. The adequacy of the data was studied by Kaiser–Meyer–Olkin test and Bartlett test of sphericity. The number of components retained was based on eigenvalues (>1), and loadings greater than 0.5 were used to identify the variables comprising each component. Unsupervised cluster analysis was performed based on squared Euclidean distances, and Ward’s Minimum Variance method was used to identify the clusters. Heatmaps were generated under R package heatmap.2. SPSS 21.0, R 3.0.3, and GraphPad Prism 5.0 for Windows were used.

The concentration of total FFA was measured in serum samples from 21 SLE patients and 25 matched HC (Table 1). SLE patients exhibited higher FFA serum levels (Table 1). Differences between groups in the levels of FFA remained significant after adjusting for age and gender (p = 0.024). Moreover, no associations were found with demographical parameters, cholesterol and triglycerides levels, and dietary intakes (Table 2). Furthermore, FFA were neither related to clinical manifestations (Table 3) (all p > 0.050) nor disease activity (r = −0.349, p = 0.169), duration (r = −0.005, p = 0.982), or anti-dsDNA levels (r = −0.350, p = 0.130) in SLE. Therefore, parameters other than those indicated may explain the altered serum FFA levels registered in SLE.

Then, we wondered whether gut microbial composition may account for the increased FFA serum levels in SLE. To this aim, the associations between FFA levels and the main intestinal microbial groups analyzed by a metagenomic approach as already described (22) were assessed. As previously reported, diminished Firmicutes to Bacteroidetes (F/B) ratio characterized the intestinal dysbiosis found in SLE compared to healthy subjects [1.94 (1.51) vs 4.27 (5.93), p < 0.001] (22). Among the phyla analyzed (Actinobacteria, Bacteroidetes, Firmicutes, Cyanobacteria, Euryarchaeota, Fusobacteria, Lentisphaerae, Proteobacteria, Tenericutes, TM7, Verrucomicrobia, and Synergistetes), FFA levels displayed opposite correlations with the Firmicutes and Bacteroidetes groups in HC, but not in lupus patients (Table 4). Consequently, a negative association with the F/B ratio was observed in HC, but not in SLE (Figure 1). Moreover, this association remained significant after adjusting for potential confounders (Table 5). Therefore, F/B ratio was found to be the main predictor of FFA serum levels in HC, whereas this effect was not seen in lupus patients, hallmarked by a decreased F/B ratio and elevated FFA levels.

Since some heterogeneity within groups in the FFA levels was observed and in order to gain more insight into the connections between gut microbiota and serum FFA, further analyses were performed. Focusing on the main microbial groups at the level of phyla, individuals were classified into groups by means of a cluster analysis. Interestingly, two main clusters were identified (thereafter referred to as clusters I and II) (Figure 2A), mainly differing in the F/B ratio (8.84 ± 5.53 vs 1.70 ± 0.76, respectively; p < 0.0001). Notably, a distinct distribution of individuals was observed, as HC were mainly found within the cluster I (14/25), whereas SLE patients were marginally present in this group (3/21, p = 0.004). On the one hand, this result highlights a shift in the microbiota composition in SLE patients compared to HC, hence supporting an association between a biased distribution of the intestinal microbial groups and elevated FFA serum levels. More importantly, when FFA levels were compared among HC and SLE subjects stratified by microbial clusters, it was noted that HC grouping within the cluster II exhibited similar FFA serum levels as SLE patients (Figure 2B), thus reinforcing the relevance of the microbiota composition on the FFA serum levels.

Overall, our findings disclose a strong association between FFA levels in serum and specific groups of the gut microbiota in healthy individuals, but not in lupus patients where a profound intestinal dysbiosis was registered.

Our results point to a relationship between the gut microbiota and the host metabolism at the systemic level, but the actual mediators are unclear. Since SCFA may affect the human metabolism and an altered gut microbiota composition leads to a dysregulated SCFA production, the associations between fecal SCFA levels and those of serum FFA were analyzed.

On the one hand, higher levels of all SCFA studied were observed in lupus patients compared to HC (Table 6). However, no differences were found between SLE and HC when relative proportions were compared (all p > 0.050). On the other hand, all SCFA exhibited a positive correlation with FFA serum levels in HC (Table 6).

Importantly, the F/B ratio was negatively correlated to the fecal levels of propionate and butyrate in HC but not in SLE patients (Figure 3). Notably, stronger associations were found in HC when only Bacteroidetes abundance was considered: propionate (r = 0.653, p < 0.001) and butyrate (r = 0.623, p = 0.002).

In order to gain further insight into the relevance of the intestinal microbiota and the SCFA production and serum FFA levels, we performed additional analyses by stratifying subjects according to the clusters obtained from the microbiota analysis (Figure 2) instead of their clinical condition. Interestingly, a negative association between F/B ratio and fecal SCFA was found in cluster I (propionate: r = −0.621, p = 0.024 and butyrate: r = −0.654, p = 0.015), but not in those grouped in cluster II (propionate: r = −0.015, p = 0.940 and butyrate: r = 0.220, p = 0.271). Overall, this picture mirrors that of found for the HC vs SLE populations according to our previous findings and confirms a pivotal role of the intestinal microbiota in this scenario. However, SCFA and FFA levels did not remain associated after stratifying the whole population by the clusters, hence suggesting the involvement of additional factors, such as the clinical condition, to explain the connection between gut microbiota composition, SCFA, and serum FFA levels.

All these results highlight a role for the gut microbiota in the maintenance of serum FFA levels, SCFA having a potential role orchestrating this interaction. Additional factors, such as disease status, may also influence the outcome of the associations between gut microbiota composition and the interaction SCFA–FFA. Indeed, altered gut microbiota composition found in SLE patients was linked to an altered SCFA production and increased FFA levels in serum, thus supporting this hypothesis.

Although elevated FFA levels were found in SLE, whether a global increase in all FFA species underlies this finding, or if some specific FFA were altered was not clear. To address this issue, a number of FFA species were measured, and the differences between SLE and HC were studied.

Overall, no striking differences were observed between patients and controls (Table 7). However, since some collinearity among FFA species existed, a PCA was carried out on the FFA species analyzed to avoid potential biases. PCA demonstrated a good adequacy of the data (KMO statistic = 0.781, Barlett sphericity test p = 10⁻⁴⁴), and three components were identified (eigenvalues >1) explaining 77.53% of the total variance. Based on their loadings, PC1 (53.32% variance explained) retained γ-linolenic, palmitoleic, palmitic, oleic, linolenic, linoleic, and arachidonic acids, and PC2 (13.95% variance explained) mainly retained EPA and DHA. Finally, PC3 (10.26% variance explained) only retained stearic acid (Figure 4).

On the one hand, these results underlie the outstanding heterogeneity of FFA species, a clearly distinct pattern of grouping depending on their chemical structure (chain length or double-bond position) not being found. Globally, saturated, monounsaturated, and w6 fatty acids were grouped together within the PC1, whereas the most important anti-inflammatory w3 FFA did in the PC2. Stearic acid, of controversial immunological and metabolic role, was grouped in the third component. Hence, these findings support a functional, rather than structural, relationship of FFA.

On the other hand, no differences in the PCA scores were registered between SLE and HC groups (PC1: p = 0.169, PC2: p = 0.378, and PC3: p = 0.916), thus suggesting that SLE patients did not exhibit a different FFA profile compared to HC.

Finally, whether PCA scores could be related to gut microbiota composition was analyzed. Notably, F/B ratio was positively correlated with the PC2 score (Table 8) in HC but not in SLE, thereby suggesting a beneficial effect of gut microbiota composition on the serum FFA pool in healthy individuals. Again, when the phyla were independently studied, Bacteroidetes exhibited a stronger correlation with PC2 score (r = −0.433, p = 0.039) than that of Firmicutes (r = 0.411, p = 0.052) in HC individuals. Thus, gut microbiota seems to quantitatively and qualitatively impact the FFA serum pool.

Since some associations between gut microbiota and specific FFA PCA scores were observed, we aimed to evaluate whether these parameters could be related to some relevant serum biomarkers in lupus. To this end, a panel of serum biomarkers of endothelial activation (VEGF, GM-CSF, EGF, IL-8, TNFα, MCP-1, IP-10, and leptin) and oxidative stress (MDA) were measured (Table 9).

First, the associations between these biomarkers and the gut microbiota composition were analyzed. Interestingly, the F/B ratio was negatively associated to leptin (r = −0.369, p = 0.006) and MCP-1 serum levels (r = −0.304, p = 0.025) in the whole group.

On the other hand, divergent associations were noted among FFA PCA scores and these biomarkers in SLE patients. Whereas PC1 was positively correlated with biomarkers of endothelial activation [VEGF (r = 0.444, p = 0.044), IL-8 (r = 0.522, 0.015), and EGF (r = 0.400, p = 0.070)], negative associations were observed for PC2 [EGF (r = −0.425, p = 0.055), MCP-1 (r = −0.640, p = 0.002), IP-10 (r = −0.397, p = 0.075), TNFα (r = −0.410, p = 0.065), and MDA (r = −0.375, p = 0.057)]. Interestingly, negative associations were also found for PC3 [VEGF (r = −0.534, p = 0.013) and IL-8 (r = −0.459, p = 0.036)]. No associations were observed in the HC.

All these findings seem to point to a link between gut microbiota, FFA serum pool, and biomarkers of endothelial activation in lupus, thus emphasizing the systemic effect of the gut microbiota in this condition. Additionally, differences among FFA PCA are in line with their proposed functional diversity.