A SV-40 transformed bovine peritoneal macrophage cell line (BOMAC), was obtained from Judith Stabel (USDA, Ames, IA, US) and maintained as previously described (Stabel and Stabel, 1995). An ovine blood macrophage-like cell line (MOCL-4) was obtained from Michel Olivier (INRA, Nouzilly, France) and cultured as previously described (Olivier et al., 2001).

Two Map isolates from cattle (Bos taurus) and sheep (Ovis aries) were selected from the collection of isolates of the Mycobacteria laboratory, NEIKER-Tecnalia. The bovine K10 isolate of Map, a sequenced strain recovered from a clinical case of paratuberculosis, was obtained from the American Type Culture Collection. The Map 2349/06-1 isolate from sheep was kindly provided by A. C. Coelho. Isolate code, country of isolation, and genotype for each Map isolate are summarized in Table 1. Entry and intracellular growth of the K10 and 2349/06-1 isolates of Map in bovine and ovine macrophages was previously estimated (Abendaño et al., 2013, 2014) and is presented in Table 1. Both Map isolates were maintained as glycerol stock at −80°C (Adúriz et al., 1995; Sevilla et al., 2005, 2007). Aliquots of these glycerol stocks were utilized to directly inoculate all subsequent cultures for use in macrophages infections. The isolates of Map selected for our study were grown in T25 tissue culture flasks at 37 ± 1°C in 10 ml of Middlebrook 7 H9 broth (Difco Laboratories, Detroit, MI) supplemented with 10% (v/v) oleic acid-albumin-dextrose-catalase (Becton, Dickinson and Company, Franklin Lakes, NJ), 0.05% (v/v) Tween-80 (Sigma-Aldrch, St Louis, MO) and 2 mg l⁻¹ of Mycobactin J (Allied Monitor Inc., Fayette, MO) for 20 days at 37°C. Bacterial cells were harvested by centrifugation at 2000 × g for 20 min in a Beckman Coulter Allegra X-12 centrifuge. Bacterial pellets were washed three times with sterile Hank's balanced salt solution (HBSS), resuspended in 2 ml of HBSS, and the resultant suspension was passed 20 times through a 27-gauge needle, and large aggregates were allowed to settle. After 5 min, an aliquot was taken from the top half of the bacterial suspension and diluted in HBSS to a McFarland standard of 1 (3 10⁸ CFUs/ml) with a Densimat (bioMerieux, Marcy l'Étoile, France). Only the top fraction of the suspension containing dispersed bacteria was used for the infection assays.

Confluent monolayers of BOMAC and MOCL-4 cell lines grown in a 125-cm² cell culture flask (Corning Costar, New York, US) at 37°C in a 5% CO₂ were infected with each of the two Map isolates at MOI of 1:10 (cell:bacteria). At 4 h post-infection, the medium was collected, centrifuged at 2000 g for 15 min in a Beckman Coulter Allegra X-12 centrifuge and the resultant pellet containing extracellular bacteria was frozen at −80°C. Cell monolayers were washed twice with 20 ml of HBSB and then treated with 200 μg/ml amikacin (Sigma) in HBSB to kill extracellular bacteria. After 2 h at 37°C, the amikacin was removed, the monolayers were washed twice with HBSS and fresh culture medium was added to the monolayers. After 4 days at 37°C, intracellular bacteria were released by lysing the monolayers with sterile water. Cell debris and nuclear fractions were removed by low-speed centrifugation at 400 g for 5 min at 4°C. The bacterial fraction was recovered from the supernatant after additional centrifugation at 2000 g for 15 min.

FAMEs were extracted from bacterial pellets by saponification with sodium hydroxide in methanol, methylated with hydrochloric acid in methanol and then extracted with hexane in methyl-tert-butyl-ether (Sasser, 1990). Briefly, 1 ml of 15% (w/v) NaOH in 50% aqueous methanol was added to each bacterial pellet and incubated at 100°C for 30 min in a water bath. The saponified samples were then cooled in a pan of cold tap water for 2 min, acidified and methylated by the addition of 2 ml of 54% 6 N HCl in 46% methanol, and incubated at 80°C for 10 min in a water bath. This step drops the pH of the solution below 1.5 and cause methylation for the increased volatility in a partially polar column of the FAMEs. The methylated FAMEs were then incubated with 1.25 ml of an organic solvent [hexane:methyl-tert-butyl ether (1:1)]. Each sample was mixed for 10 min and the lower phase was removed with a Pasteur pipette. The upper phase was washed with 3 ml of 0.3 M NaOH to remove both free FAMEs and residual reagents. Following the wash step, the organic phase containing the FAMEs was transferred into a GC vial. FAMEs are more volatile than their respective FAs and therefore more suitable for GC analysis.

The FAMEs were analyzed by GC using the Agilent 6850 gas chromatographic unit equipped with a crossslinked 5% phenylmethyl silicone-fused silica capillary column (25 m × 0.2 mm, Agilent 19091B-102E), a flame ionization detector and an Agilent 6850 automatic liquid sampler. The column temperature ramped from 170 to 270°C at 5°C min⁻¹, then increased to 310°C at 40°C min⁻¹ and held for 1.5 min. Hydrogen served as the carrier gas at 0.5 ml min⁻¹. An external calibration standard, a mixture of straight chain saturated FAs from 9 to 20 carbons in length (9:0 to 20:0) and five hydroxyl acids (Sherlock MIS Calibration Standard, Part#1200-A, MIDI Inc., Newark, DE, US), was used. The hydroxyl compounds are especially sensitive to changes in pressure/temperature relationships and to contamination of the injection port lines. As a result, these compounds function as real-time quality control checks for the system. Retention time data obtained by injecting the calibration standard is converted to Equivalent Chain Length (ECL) data for bacterial FAs naming. The Retention factor (RFact) for each FA can be derived as a function of its elution time in relation to the elution time of the known series of straight chain FAs. GC allows windows to be set at 0.010 ECL units, giving great resolution of FAs isomers.

FAs analysis using the Sherlock® Microbial Identification software was used to automatically name and quantitate the peaks in an unknown sample (MIDI Inc, Newark, DE, US). The FAMEs extraction procedure may carry over sterols and other non-fatty acid material. Additionally, electronic noise may result in transient spikes, which might interfere with the chromatography. However, FAs peaks always have area/height ratios greater than 0.017 and less than 0.070, making it possible to set exclusionary parameters at these levels. Electronic noise spikes are typically <0.017 and non-fatty acid peaks are usually >0.070, allowing rejection of these artifacts. The Sherlock approach is set to use a “Summed in Feature” wherever imperfect peak separation occurs. Reproducibility of the profiles was calculated by comparing two different FAMEs extracts.

The Sherlock® Microbial Identification System with covariance matrix, principal component analysis and pattern recognition software was used to generate FAs profiles for our samples. The covariance matrix takes into account the mole-for-mole relationship of the conversion of one FA to another which might occur in relation to a temperature shift. The pattern recognition uses ratios between FAs amounts in addition to the principal component base. The dendrogram and principal component analysis (2-D plots) programs of the Sherlock® Microbial Identification System use data from FAs analyses to graphically illustrate relationships between the samples. The dendrogram analysis produces unweight pair matching based on FAs composition. In a dendrogram, the Euclidean distance is the distance in n-dimensional space between two bacterial samples when their FAs composition is compared. Lower linkages indicate greater similarity. In addition to dendrogram, the 2-D Plot is another cluster analysis tool which uses a principal components analysis of FAs profiles to group entries in a two dimensional space. The x-axis represents principal component 1, and the y-axis represents principal component 2.

The percentage of 21 FAs in the K10 (type-C) and 2349/06-1 (type-S) isolates of Map under different environmental conditions (extracellular, intracellular, or 7H9 medium) were compared with the General Lineal Model (GLM) procedure of the SAS statistical package version 9.3 (SAS Institute Inc., Cary, NC). Differences were considered significant when P-values were <0.05. Correlations between FAMEs profiles were examined with the principal components procedure of the SAS software.

The percentages of each of the 21 FAs in the K10 (type-C) and 2349/06-1 (type-S) isolates of Map under different environmental conditions (BOMAC, MOCL-4), and localizations within the host cell (extracellular and intracellular) were modeled using the NFL FormRules® software v4.03 (Intelligensys Ltd., 2013, Stokesley, UK). The FormRules software contains various statistical fitness criteria including Cross Validation, Minimum Description Length, Structural Risk Minimization, Leave One Out Cross Validation and Bayesian Information Criterion. All were analyzed in order to generate the model with the best predictability together with the simplest and more intelligible rules. The predictability of the model was assessed using correlation coefficient (R²) and ANOVA F-values for the percentage of each fatty acid. R2=(1−∑i = 1n(yi−y′i)2∑i = 1n(yi−y″i)2)×100% Where y” is the mean of the dependent variable, and the y' is the predicted value calculated by the model. ANOVA F-values over its critical values for the corresponding degrees of freedom are an indication of reasonable model predictabilities (Colbourn and Rowe, 2009). For each of the generated rule a “membership degree” or confidence level is defined which specifies how a “value” belongs to that fuzzy subset (from 0 to 1).

A representative chromatogram for each Map isolate (K10 and 2349/06-1) grown in Middlebrook 7H9 medium is presented in Figure 1S. The identities of 97.51 and 87.87% FAs of the K10 and 2349/06-1 isolates were confirmed respectively by demonstrating retention times similar to those of known standards. The peak, R Factor, systematic name, usual name and relative frequency of each identified FA are summarized in Table 2. FAs found ranged from 9 to 20 carbon atoms. They consisted of nine saturated FAs (9:0, 10:0, 12.0, 14:0, 15:0, 16:0, 17.0, 18:0, 20:0), six monounsaturated FAs (16:1 w9c, 16:1 w7c, 16:1 w6c, 17:1 w8c, 18:1 w9c, 18:1 w7c), one double-unsaturated FA (18:2 w6,9c), one 10-methyl branched of 18 carbon atoms (18:0 10Me) and three Sum in Feature FAs. The Palmitic acid (16:0) and Oleic acid (18:1 w9c) were the most abundant FAs and together represented more than 50% of the total cellular FA content of both Map isolates. From the 21 identified FAs, 14 were present in the profiles of both Map strains. The K10 isolate was characterized by the presence of four FAs (9:0, cis-10-Palmitoleic acid, Summed in Feature 1, and TBSA) which were absent in the 2349/06-1 isolate. The 16:0 2,4 DiMe was detected in very small amount only in the S-type isolate of Map. Statistical analysis of the FAs composition of both Map isolates grown in Middlebrook 7H9 medium (Figure 1) enabled the distinction between both Map strains based on the presence of the TBSA (18:0 10Me) exclusively in the C-type isolate of Map. Among the 14 FAs present in the profiles of both Map isolates, significant differences were observed in the abundances of Palmitic acid and TBSA between both Map isolates.

To evaluate whether the extra- and/or the intracellular environment influences Map FAs composition, BOMAC and MOCL-4 cells were infected with the K10 and the 2349/06-1 isolates of Map. At 4 h post-infection, the medium was collected, centrifuged at 2000 g for 15 min and the FAMEs of the bacterial pellet extracted. After 4 days at 37°C, intracellular bacteria were recovered by differential centrifugation after lysing the infected monolayers. FAMEs were extracted from extra- and intracellular bacteria and analyzed by GC (Figures 2S, 3S). The percentages of the FAs found in both isolates recovered from the extra- and intracellular environment of BOMAC and MOCL-4 cells are reported in Table 3. The FAs profiles of both Map isolates grown in 7H9 medium and recovered from the extracellular medium of BOMAC or MOCL-4 cells were found to be highly similar. As shown in Figure 1, the amount of Palmitic acid and TBSA was quite different in both isolates of Map recovered from the extracellular environment regardless of the cell line. Significant differences in the content of cis-7-Palmitoleic acid between both isolates of Map were only observed in the bacteria recovered from the extracellular environment of BOMAC cells. When comparing the FAs profiles of each Map isolate recovered from the extra- or intracellular environment of BOMAC and MOCL-4 we observed that Map FAs profiles change upon host macrophages infection and that these changes are strain-dependent (Figure 2). Since the K10 and 2349/06-1 isolates recovered from the extracellular environment showed a different FAs profile they had to re-align their FAs metabolism inside host macrophages in a different manner. Once within host macrophages, FAs profiles of both Map isolates were equivalents regardless of the cell line.

The dendrogram and 2D-plots derived from the FAs profiles of both isolates under the three assessed environmental conditions (extra, intra, and 7H9 grown) are presented in Figures 3A,B, respectively. FAs profiles were clustered into three groups. The first group consisted of the FAs profiles of the K10 and 2341/06-1 isolates recovered from the intracellular environment of BOMAC and MOCL-4 cells. The FAs profiles in the first group were clustered into two subgroups according to subtle differences of FAs depending of the host cell line.

The second group includes the FAs profiles of the K10 strain grown in 7H9 medium or recovered from the extracellular medium of infected BOMAC or MOCL-4 cells. The common feature of this group is that the amount of Palmitic acid and TBSA was different than the 2341/06-1 isolate which was clustered in a third group. Although the extracellular and 7H9-grown bacteria belonged to the same cluster, they formed two separate subgroups that reflected environmental divergence. The FAs profiles of each isolate of Map recovered from the extracellular environment of BOMAC and MOCL-4 cells were very similar, and the analysis could not differentiate these two extracellular environments. An excellent correlation between the two-dimensional plots obtained with the principal components procedures of the Sherlock Microbial Identification System and the SAS software was obtained.

In order to understand how the different environmental conditions (BOMAC, MOCL-4), and localization within the host cell (extra- and intracellular) contribute alone or in combination to the amount each of the 21 FAs identified in the K10 and 2349/06-1 isolates of Map, our data were modeled using the NFL FormRules® software v4.03 (Intelligensys Ltd., 2013, Stockesley, UK). Within the statistical fitness criteria included in this software, Structural Risk Minimization was selected to give our model with the best predictability and, simultaneously, the simplest and more intelligent rule sets. This approach allowed the discrimination of the inputs (host of origin of each isolate, genotype, cell line, and localization) which more accurately explained the variability of the amount of each identified FA. For instance, the effect of the Map isolate (K10 or 2349/06-1) and its localization within the host cell (extra- or intracellular) determined the variability in the amount of cis-7-Palmitoleic acid, Palmitic acid and TBSA (Figure 4). NFL technology generated a set of “IF…THEN” rules per submodel with their corresponding membership degrees (Table 4). These rule sets represent the cause-effect relationships of the different inputs on the percentage of each FA. For example, for Rule 9: “If the isolate is the K10 and if it is localized in the extracellular environment of the host cell then the amount of TBSA is high with a confidence level of membership of 0.92.” In contrast, if the isolate is the 2349/06-1 the amount of TBSA is low with a confidence level of membership of 1.00 (Rule 10). Interestingly, the only factor explaining the variability in the amount of the Stearic acid and Sum in Feature 3 of Map was the bacterial localization, extra- or intracellular.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.