n-3 long-chain PUFA promote antibacterial and inflammation-resolving effects in Mycobacterium tuberculosis-infected C3HeB/FeJ mice, dependent on fatty acid status

Abstract Non-resolving inflammation is characteristic of tuberculosis (TB). Given their inflammation-resolving properties, n-3 long-chain PUFA (n-3 LCPUFA) may support TB treatment. This research aimed to investigate the effects of n-3 LCPUFA on clinical and inflammatory outcomes of Mycobacterium tuberculosis-infected C3HeB/FeJ mice with either normal or low n-3 PUFA status before infection. Using a two-by-two design, uninfected mice were conditioned on either an n-3 PUFA-sufficient (n-3FAS) or -deficient (n-3FAD) diet for 6 weeks. One week post-infection, mice were randomised to either n-3 LCPUFA supplemented (n-3FAS/n-3+ and n-3FAD/n-3+) or continued on n-3FAS or n-3FAD diets for 3 weeks. Mice were euthanised and fatty acid status, lung bacterial load and pathology, cytokine, lipid mediator and immune cell phenotype analysed. n-3 LCPUFA supplementation in n-3FAS mice lowered lung bacterial loads (P = 0·003), T cells (P = 0·019), CD4+ T cells (P = 0·014) and interferon (IFN)-γ (P < 0·001) and promoted a pro-resolving lung lipid mediator profile. Compared with n-3FAS mice, the n-3FAD group had lower bacterial loads (P = 0·037), significantly higher immune cell recruitment and a more pro-inflammatory lipid mediator profile, however, significantly lower lung IFN-γ, IL-1α, IL-1β and IL-17, and supplementation in the n-3FAD group provided no beneficial effect on lung bacterial load or inflammation. Our study provides the first evidence that n-3 LCPUFA supplementation has antibacterial and inflammation-resolving benefits in TB when provided 1 week after infection in the context of a sufficient n-3 PUFA status, whilst a low n-3 PUFA status may promote better bacterial control and lower lung inflammation not benefiting from n-3 LCPUFA supplementation.

response to infection, rather than treatment strategies directed at bacterial killing, has lately been suggested for improving current TB treatment regimens (3) . Since TB is characterised by excessive, non-resolving inflammation, various anti-inflammatory drugs have been investigated for use as possible HDT options (4,5) . These medications have been shown to reduce lung lesions and bacillary load, favouring host survival (4,6,7) . However, they are not without side effects and, therefore, a nutritional approach may be considered a safer alternative (8) .
Dietary n-3 long-chain PUFA (n-3 LCPUFA) consumption alters membrane phospholipid fatty acid (FA) composition of blood and tissue cells that play a role in immune and inflammatory responses (9)(10)(11) . It is well known that various lipid mediators, synthesised from n-3 LCPUFA, contribute to inflammation resolution. Eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) serve as precursors for specialised pro-resolving mediators (SPM), including resolvins, protectins and maresins. These SPM play a role in significantly reducing pro-inflammatory lipid mediator, chemokine and cytokine production and altering immune cell recruitment, whilst promoting anti-inflammatory cytokine release (12) . The incorporation of dietary EPA and DHA into cell membranes has also been found to enhance the phagocytosis of apoptotic cells and bacteria, whilst SPM promote bacterial killing (12,13) . Although these functions have not been proven in TB specifically, n-3 LCPUFA have been successfully used as anti-inflammatory and inflammation-resolving agents in other conditions driven by inflammation (9) .
Considering this, it is reasonable to hypothesise that n-3 LCPUFA supplementation would benefit TB patients, but research on the application of n-3 LCPUFA as HDT in TB is limited at present. Moreover, the effects of n-3 LCPUFA supplementation after the acute inflammatory response in Mycobacterium tuberculosis (Mtb) infection have not yet been investigated. The aim of the present study is, therefore, to determine the effects of EPA and DHA supplementation, administered 1 week after Mtb infection for 28 d, on inflammatory, immune and clinical outcomes in C3HeB/FeJ mice. The well-established C3HeB/FeJ mouse model has been reported to be the closest representative murine model of human pulmonary TB lung histopathology (14) . Furthermore, the n-3 LCPUFA status of the general human adult population is not considered optimal, owing to insufficient dietary n-3 PUFA consumption and high dietary n-6 (n-6)/n-3 PUFA ratios, often resulting in low n-3 PUFA status (15,16) . We further aim to mimic this scenario of possible suboptimal n-3 PUFA intakes among TB patients to determine whether supplementation outcomes depend on n-3 PUFA status before Mtb infection (interaction effects between n-3 PUFA status and n-3 LCPUFA supplementation).

Animals and ethics statement
Male C3HeB/FeJ mice (Jackson Laboratory), aged 10-12 weeks, were bred and housed at the Institute of Infectious Diseases and Molecular Medicine, University of Cape Town, Cape Town, South Africa. Following infection, mice were housed in a biosafety level 3 containment facility, five per individually ventilated cage with filter tops (type 2 long), as well as dried wood shavings and shredded filter paper as floor coverings. The temperature range was set at 22-24°C and 12-to-12 h light cycles. The experiments were performed in accordance with the South African National Guidelines and University of Cape Town practice guidelines for laboratory animal procedures. The protocol was approved by the Animal Ethics Committee, Faculty of Health Sciences, University of Cape Town (AEC 015/040) and the AnimCare Animal Research Ethics Committee of the North-West University (NWU-00260-16-A5).

Experimental design and animal diets
Mice had ad libitum access to food and water. The experimental design of the present study is illustrated in Fig. 1. Mice were randomly allocated to an n-3 PUFA-deficient (n-3FAD) (n 20) orsufficient diet (n-3FAS) (n 20) and kept on these diets for 6 weeks prior to infection, in order to establish a sufficient or a low n-3 PUFA status. The n-3FAS diet contained the essential n-3 PUFA α-linolenic acid. Mice were then infected via the aerosol route (described below) and their respective diets maintained for an additional week. One week post-infection (week 7), mice that were conditioned on the n-3 PUFA-sufficient diet (n-3FAS) were randomised to continue on this diet (n-3FAS) (n 10) or were switched to the same diet supplemented with n-3 LCPUFA (EPA plus DHA) (n-3FAS/n-3þ group, n 10) (Fig. 1). Similarly, the mice in the n-3FAD group either continued on the n-3FAD diet (n 10) or were switched to the n-3 LCPUFA-supplemented diet (n-3FAD/n-3þ group, n 10). The mice received these diets for an additional 3 weeks until euthanasia at 28 d after infection (as described below). The welfare of the mice was assessed daily and body weight and food intake were measured weekly. The daily food intake per mouse was calculated by dividing the weekly food intake by seven (days) and then by five (five mice per cage). The results of this experiment were reproduced in a second experiment (resulting in ten mice per treatment group). The data of one experiment (five mice per group) are presented in this article.
All the purified experimental diets were obtained commercially (Dyets) and were based on the AIN-93G (17) formulation, all containing 10 % fat, but with modifications in the fat source (Table 1). All the diets were isoenergetic with identical macronutrient contents. The mice in the n-3FAS group received the AIN-93G diet, which provides both n-3 and n-6 PUFA at amounts found to induce optimal tissue saturation of DHA and arachidonic acid (AA), in rodents (17) . The EPA-and DHAsupplemented diets (n-3þ) contained commercially obtained Incromega TG4030 oil (Croda Chemicals) supplemented at amounts that could reasonably be achieved in humans. GC-MS analysis was performed by the manufacturer to confirm the FA composition of the diets (Table 1). From this composition, the actual EPA and DHA intake could be calculated and was expressed as percentage of total energy intake.

Aerosol infection
A virulent Mtb H37Rv strain was cultured and stocks were prepared and stored at −80°C, as described elsewhere (18) . Mice were exposed to aerosol infection for 40 min by nebulising 6 ml of a suspension that contained 2·4 × 10 7 live bacteria in an inhalation exposure system (model A4224, Glas-Col). One day following infection, four mice were euthanised to confirm the infection dose, which was 500 colony-forming units/mouse.

End point blood and tissue collection
At the end of the 3 weeks of receiving intervention diets, mice were euthanised by halothane exposure, followed by trunk blood collection by heart puncture. The blood was collected into EDTA-coated Microtainer® tubes (K 2 EDTA, 1000 μl, BD), and then centrifuged. The plasma and buffy coat were removed for FA analysis. The erythrocytes were washed twice with saline before storage at −80°C and subsequent FA analysis. The lung lobes were removed aseptically and weighed prior to preparation. The left lung lobe was homogenised in saline and 0·04 % Tween-80 for the analysis of the bacillary load and lung cytokines. The right superior and post-caval lung lobes were snapfrozen in liquid N 2 and stored at −80°C for lung FA and lipid mediator analysis. The right middle lobe was submerged in 10 % neutral buffered formalin for histology analysis and the right inferior lobe prepared for flow cytometry.
Total phospholipid fatty acid composition analysis FA were extracted from ∼20 mg lung tissue, homogenised in 10 μl PBS with protease inhibitor (homogenisation buffer) per 1 mg tissue, or from ∼200 μl erythrocytes or peripheral blood mononuclear cells (PBMC) collected as buffy coat. Lipids were extracted from each lipid pool with chloroform-methanol (2:1, v:v; containing 0·01 % butylated hydroxytoluene) by a modification of the method of Folch et al. (19) The lipid extracts were Week 5 Week 10 Week 11 Week 9 n-3FAD diet (n 22) Week 1 n-3FAS diet (n 22)

Bacterial load determination
The bacterial loads of lungs were determined at euthanasia (28 d after infection). The left lung of each mouse was aseptically removed, weighed, homogenised and serial dilutions were plated onto Difco TM Middlebrooks 7H10 Agar (BD Biosciences) medium with oleic acid-albumin-dextrosecatalase supplementation and 0·005 % glycerol. The colonyforming units were determined 21 d following incubation at 37°C. Data are expressed as log 10 colony-forming units.

Histopathology analysis
Right middle lobes of the lungs were dissected out and fixed in 10 % neutral buffered formalin. The tissue was processed using the Leica TP 1020 Processor for 24 h and subsequently embedded in paraffin wax. The Leica Sliding Microtome 2000R was used to cut 2-μm thick sections of the embedded tissues. Three sections with 30 μm distance apart per section were cut, deparaffinised and subsequently stained with the haematoxylin/eosin stain. The images were acquired in Nikon Eclipse 90i microscopes and analysed with NIS-Elements AR software (Nikon Corporation) to determine the granulomatous area and alveolar space as a percentage of the total lung tissue (20) .

Statistical analysis
Using the G*Power statistical package version 3.1.9.7, a two-way ANOVA power analysis was done. A total sample size of 34 was calculated for an α of 0·05, a power of 80 % and an effect size estimated at 0.5. Therefore, a total sample size of 40 mice was included in this research in two experiments (n 20 each) of five mice per group. Data are presented as means and standard errors of the means. Statistical analyses were performed using IBM SPSS statistics software (version 25; IBM Corporation). To determine the differences between FA composition at baseline in the n-3FAD and n-3FAS group, the Student Fischer t test for independent variables was used. The main effects of n-3 LCPUFA supplementation (n-3FAS/n-3þ and n-3FAD/n-3þ v. n-3FAS and n-3FAD) and a low pre-infection n-3 PUFA status (n-3FAD and n-3FAD/n-3þ v. n-3FAS and n-3FAS/n-3þ), and their interaction (pre-infection status × n-3þ), on all outcome variables, were analysed by using two-way ANOVA. Significant treatment effects in the absence of a significant interaction effect indicate additive effects of the treatments, whereas a significant interaction implies synergism or antagonism. In the presence of a significant main effect or interaction, between-group differences n-3 improves tuberculosis outcomes were examined using the Bonferroni correction for multiple comparisons.
The total phospholipid fatty acid composition of erythrocytes, peripheral blood mononuclear cells and crude lung homogenates Table 2 presents the phospholipid FA composition of erythrocytes following the 6-week dietary conditioning period on either n-3FAS or n-3FAD diets. Erythrocyte FA composition has been reported to be representative of the FA content of other tissues (23) . Following the conditioning period, the n-3FAD group had lower EPA, DHA and total n-3 LCPUFA, and higher AA, osbond acid and total n-6 LCPUFA compositions, as well as a higher total n-6/n-3 LCPUFA ratio, in comparison with the n-3FAS group (P < 0·001 for all). There was no significant difference between the n-3FAS and n-3FAD groups in terms of erythrocyte saturated fatty acid composition following the conditioning period of 6 weeks (n-3FAS, 34·97 (SE 2·71); n-3FAD, 34·62 (SE 2·53)). The phospholipid FA composition of erythrocytes, PBMC and crude lung homogenates of Mtb-infected mice after 3 weeks of dietary intervention is presented in Table 3. In addition to recruited immune cells, lung epithelium also synthesises lipid mediators, and therefore, the modification of the FA composition of lung tissue and immune cells may exert local immune-and inflammation-modulatory effects (11,24) . There were antagonistic pre-infection status × n-3þ interactions for DHA, total n-3 LCPUFA, osbond acid, total n-6 LCPUFA and n-6/ n-3 LCPUFA ratios in erythrocytes, PBMC and lung homogenates (P < 0·001 for all) and AA in erythrocytes and PBMC (P < 0·001 and P = 0·001) ( Table 3). n-3 LCPUFA supplementation resulted in higher phospholipid EPA, DHA and total n-3 LCPUFA (P < 0·001 for all), whilst there was an effect of a low n-3 PUFA pre-infection status for lower EPA, DHA and total n-3 LCPUFA in erythrocytes, PBMC and lung homogenates (P < 0·001 for all, except for EPA in lung homogenates P = 0·82).
With regard to n-6 PUFA, n-3 LCPUFA supplementation lowered AA, osbond acid, total n-6 LCPUFA and total n-6/n-3 LCPUFA ratios in erythrocytes, PBMC and crude lung homogenates (P < 0·001 for all). In contrast, there was an effect of a low n-3 PUFA pre-infection status for higher AA, osbond acid, total n-6 LCPUFA and n-6/n-3 LCPUFA ratios (P < 0·001 for all, except for AA in lung homogenates P = 0·27). Respective differences between groups are shown in Table 3. Fig. 2 shows the lung bacterial loads, percentage of free alveolar space and lung histology images. There was an antagonistic preinfection status × n-3þ interaction on lung bacterial load (P = 0·006, Fig. 2(a)). Within the n-3 PUFA-sufficient arm, the n-3FAS/n-3þ group had a lower lung bacterial load when compared with the n-3FAS group (P = 0·003). However, this lowering effect was attenuated by a low n-3 PUFA status (in the n-3FAD/n-3þ group). The n-3FAD group had a lower bacterial load compared with the n-3FAS group (P = 0·037). The quantification of the percentage of free alveolar space revealed no significant main effects for neither n-3 PUFA pre-infection status nor n-3 LCPUFA supplementation ( Fig. 2(b) and (c)).
In addition, neutrophils appeared to remain unaffected by n-3 LCPUFA supplementation and pre-infection status in n-3FAS and n-3FAD groups (Fig. 3(f)).

Discussion
The present study provides evidence that n-3 LCPUFA supplementation, commenced 1 week post-infection, reduced bacterial burden, altered the local lung immune response and assisted in weight gain in a C3HeB/FeJ mouse model of TB. Importantly, these findings applied only to mice conditioned to have an n-3 PUFA-sufficient status before infection, whereas the low n-3 PUFA status mice also showed a lower bacterial load compared with the sufficient n-3 PUFA status group and did not benefit from n-3 LCPUFA supplementation.
The finding that n-3 LCPUFA supplementation lowered bacterial burden in n-3 PUFA sufficient mice is similar to that published by Jordao et al., who found lower bacterial loads in the lungs and spleens of BALB/c Mtb-infected mice fed n-3 PUFA-rich diets, compared with mice that were fed a fat-free diet (25) . The incorporation of n-3 LCPUFA into phagocytic cell membranes changes membrane fluidity, in addition to receptor expression, thereby enhancing bacterial phagocytosis, which has also been shown in TB (26,27) . This is confirmed by the higher n-3 LCPUFA composition found in crude lung homogenates and PBMC in our study, and subsequently, higher EPA incorporation would be expected in the macrophage and neutrophil phospholipid bilayers as well. This may partly explain the lower lung bacterial loads of the n-3FAS/n-3þ group. Additionally, the changes in FA composition resulted in a more pro-resolving lipid mediator profile.  = 1000 μm). The values represent means and standard errors of the means. Results repeated in two experiments, data shown for one experiment (n 5 per group). A two-way ANOVA was used to test effects of n-3þ (n-3FAS/n-3þ plus n-3FAD/n-3þ v. n-3FAD plus n-3FAS), pre-infection status (n-3FAS plus n-3FAS/n-3þ v. n-3FAD plus n-3FAD/n-3þ) and pre-infection status × n-3þ interactions. Bonferroni correction for multiple comparisons was used, *P < 0·05, **P < 0·01. CFU, colony-forming units; n-3FAD, n-3 fatty acid-deficient diet; n-3FAS, n-3 fatty acid sufficient diet; n-3þ, n-3 long-chain PUFA-supplemented diet; /, switched to.
The n-3FAS/n-3þ group presented with higher lung concentrations of the pro-resolving 18-HEPE, which is an intermediate of the E-series resolvins (SPM) synthesised from EPA (28,29) . Since SPM aid in the differentiation and activation of macrophages and neutrophils for phagocytosis and bacterial killing (12,13,30) , this may further explain the bactericidal effects of n-3 LCPUFA supplementation observed in the present study.
bacterial loads (27,35) . Therefore, the timely initiation of n-3 LCPUFA supplementation was an important contributor to positive outcomes. Furthermore, the dietary composition provided in previous studies differed from that which we used. Whilst the EPA/ DHA ratio in the n-3þ diet groups was comparable to that of Jordao et al., who also found antibacterial effects of n-3 LCPUFA supplementation in TB, other studies that found negative effects provided either higher DHA concentrations or DHA only (25,27,32,33,36) . Previous studies also used in vitro cell culture models (27) or endogenously enriched mice (fat-1 mice) (31) , and differences in the genetic backgrounds of the mice may also have contributed.
As lung inflammation is central in lesion formation, granuloma liquefaction, cavity formation and clinical outcomes, we hypothesised that the resolution of inflammation would also improve lung pathology (2,37) . However, confirming previous evidence, no effect of n-3 LCPUFA supplementation could be found in terms of percentage of free alveolar space in the present study (32) . On the other hand, n-3 LCPUFA supplementation has previously been found to inhibit T cell proliferation, elsewhere and in TB, specifically (32,38) . Consistent with this, we also found a lower percentage of lung T cells and CD4 þ T cells in the n-3FAS/ n-3þ group, which may have been driven by the effects of n-3 LCPUFA supplementation causing structural changes to cell membranes, producing subsequent alterations in cell signalling and lipid mediator synthesis (29) . These changes, together with the lower bacterial burden in this group, may explain the lower T cell percentages in the n-3FAS/n-3þ mice.
Concerning lung cytokines, IFN-γ is important in the protection against TB; however, higher concentrations have been correlated with cavitary TB, higher bacterial loads and delayed culture conversion (2,39) . We found that IFN-γ concentrations were lower in the n-3FAS/n-3þ group, which is consistent with the findings of others in TB (32) . Similarly, n-3 LCPUFA supplementation reduced lung IL-6 and IL-1α tended to be lowered. This complements our findings on T cell numbers mentioned above and confirms previous findings (40) . As expected, there was also a trend towards n-3 LCPUFA supplementation elevating the concentrations of the anti-inflammatory IL-10, therefore promoting inflammation resolution (12) .
Supplementation of n-3 LCPUFA was successfully confirmed by elevated cell membrane compositions and a pro-resolving lung lipid mediator profile of the n-3 PUFA sufficient status arm of the study. This translated into the lowering of some pro-inflammatory lung cytokines and lipid mediators, but not in all markers. A similar result to ours was found in a rat model injected with Salmonella enteritidis endotoxin, where the administration of fish oil altered pro-resolving lipid mediators without significantly changing the cytokine concentrations in bronchoalveolar lavage fluid (41) . The fact that n-3 LCPUFA have been reported to affect the Th1/Th2 balance mainly by Mtb-infected mice with n-3FAS, n-3FAS/n-3þ, n-3FAD or n-3FAD/n-3þ diets for 3 weeks. The values represent the means. Results repeated in two experiments, data shown for one experiment (n 5 per group). A two-way ANOVA was used to test effects of n-3þ (n-3FAS/n-3þ plus n-3FAD/n-3þ v. n-3FAD plus n-3FAS), pre-infection status (n-3FAS plus n-3FAS/n-3þ v. n-3FAD plus n-3FAD/n-3þ) and pre-infection status × n-3þ interactions. Bonferroni correction for multiple comparisons was used, *P < 0·05, **P < 0·01, ***P < 0·001. HDHA, hydroxydocosahexaenoic acid; HEPE, hydroxyeicosapentaenoic acid; n-3þ, n-3 long-chain PUFA-supplemented diet; n-3FAD, n-3 fatty acid-deficient diet; n-3FAS, n-3 fatty acid-sufficient diet; /, switched to. n-3 improves tuberculosis outcomes inhibiting the production of Th1 type cytokines (including IFN-γ) may serve as an explanation for the current findings (42) . Furthermore, Kroesen and colleagues found a more pronounced effect on systemic (serum) cytokine concentrations as compared with lung cytokines when administering aspirin in the same animal TB model as in our study (4) . In contrast with our results, previous studies on n-3 LCPUFA treatment in Mtb-infected animals, macrophages and peritoneal cells showed reduced PGE 2 , leukotriene B 4 , TNF-α, IL-6, IL-1β and monocyte chemotactic protein-1 synthesis (25,27,31,33) . Nevertheless, irrespective of the fact that some of the pro-inflammatory lipid mediators and cytokines were not significantly altered in the n-3FAS/n-3þ group, the higher pro-resolving lipid mediator concentrations were a positive finding, demonstrating the pro-resolving properties of n-3 LCPUFA. Therefore, our results suggest that n-3 LCPUFA supplementation does not inhibit the host's natural immune and inflammatory responses necessary to protect against bacteria. This supports the notion that SPM are not immunosuppressive and do not block inflammation, but instead elicit pro-resolving effects (12) .
On the other hand, the low n-3 PUFA status mice also presented with lower bacterial loads, similar to that seen in the n-3 PUFA sufficient group, supplemented with n-3 LCPUFA. Bonilla et al. also reported that n-3 PUFA-deficient mice had a lower susceptibility to TB when compared with fat-1 transgenic mice, with an endogenous abundance of n-3 PUFA (31) . This may indicate that n-3 PUFA deficiency is protective against TB. Nevertheless, the clinical relevance of these findings for humans is questionable. It would be unrealistic to promote low n-3 PUFA consumption in TB infection as a protective measure, considering the other important biological functions that n-3 PUFA would have in these individuals. However, considering that there may be populations with a low n-3 PUFA status at risk for TB, the interaction between a low n-3 PUFA status, TB medication and treatment outcomes require further investigation, before continuing human trials.
As expected, the lipid mediator profile of the low n-3 PUFA status group was in congruence with their FA status. A low n-3 PUFA status promoted lower concentrations of n-3 PUFA-and higher n-6 PUFA-derived lung lipid mediators. However, the n-3FAD group presented with lower lung concentrations of IFN-γ, IL-1α, IL-1β and IL-17 compared with the n-3FAS group, which is conflicting with the FA status results and the less pro-resolving lipid mediator profiles found in these mice. The  11-HETE in crude lung homogenates after providing Mtb-infected mice with n-3FAS, n-3FAS/n-3þ, n-3FAD or n-3FAD/n-3þ diets for 3 weeks. The values represent the means. Results repeated in two experiments, data shown for one experiment (n 5 per group). A two-way ANOVA was used to test effects of n-3þ (n-3FAS/n-3þ plus n-3FAD/n-3þ v. n-3FAD plus n-3FAS), pre-infection status (n-3FAS plus n-3FAS/n-3þ v. n-3FAD plus n-3FAD/n-3þ) and pre-infection status × n-3þ interactions. Bonferroni correction for multiple comparisons was used,
reasons why the low n-3 PUFA status mice presented with lower levels of some of the inflammatory cytokines may be related to the timing of the cytokine measurement. An initially higher inflammatory response due to a higher n-6 PUFA status and pro-inflammatory lipid mediator profile may have resulted in lower cytokine concentrations by the time assessed (4 weeks after infection). Another plausible explanation is that the lower bacterial loads of these mice likely provoked a lower inflammatory response. Seemingly, in contrast, the low n-3 PUFA status in our study promoted higher percentages of certain immune cells, including the natural killer cells, interstitial macrophages and dendritic cells which were higher in the n-3FAD group compared with the n-3FAS group. This could have contributed to bacterial control of the n-3 PUFA low-status group via cell-intrinsic killing functions independent of cytokine levels. The higher percentages of dendritic cells and macrophages can be explained by the fact that PGE 2 concentrations were higher in the n-3FAD group, which have been implicated to induce human DC and mice macrophage recruitment, whilst in a peritonitis mouse model COX-2 deficient mice presented with reduced macrophage recruitment (43)(44)(45) . n-3 LCPUFA supplementation of the low n-3 PUFA status group (n-3FAD/n-3þ) did not have the same beneficial effects as in the n-3FAS/n-3þ group. Our results show that both a low n-3 PUFA status and n-3 LCPUFA supplementation had lowering effects on pro-inflammatory lung cytokines, but combining a low status, and supplementation attenuated these lowering effects. This was despite the successful alteration of the n-3 LCPUFA cell membrane composition and lipid mediators towards a more pro-resolving lung profile in the n-3FAD/ n-3þ group. Moreover, n-3 LCPUFA supplementation in the low n-3 PUFA status mice (n-3FAD/n-3þ) led to a more pronounced increase in PGE 3 and 5-HEPE than supplementation in n-3 PUFA sufficient mice. Also, different from the n-3FAS/ n-3þ group, the n-3FAD/n-3þ group showed significantly lower lung concentrations of the pro-inflammatory lipid mediators PGF2α, PGD 2, 8-, 9-and 11-HETE. Still, n-3 LPCUFA supplementation in n-3FAD mice resulted in higher lung IL-6 and IL-1α concentrations. Possible reasons why n-3 LCPUFA supplementation did not exert the same beneficial effects in the n-3FAD/ n-3þ group may be related, firstly, to the dosage and duration of supplementation and secondly, to possible epigenetic adaptation to deficiency. As discussed previously, the n-3FAD group itself also presented with low lung cytokine concentrations and possible clinical benefit to start with, which may be the reason why n-3 LCPUFA supplementation in this group did not improve cytokine concentrations or bacterial load. Nevertheless, with this in mind, it cannot be said with certainty that a low n-3 PUFA status improves TB outcomes due to the inconsistent immune and inflammatory findings of this group, or that n-3 LCPUFA should not be supplemented under conditions of a low n-3 PUFA status. Further investigation into these findings is warranted.
One of the strengths of the present study was that we used a murine model that is well-established and reflective of human pulmonary TB. Furthermore, our experimental design, including the timing of supplementation, comparison of n-3 PUFA sufficiency and low status and the EPA/DHA ratio of our supplement, also strengthens our findings. However, in the n-3FAD group, specifically, the dose of n-3 LCPUFA supplementation may have been too low and/or the duration too short. Future prospects would be to perform the present study with euthanasia time points at the different phases of the inflammatory and immune response, also including systemic markers of inflammation. Additionally, the possible beneficial effects of n-3 LCPUFA, when administered in combination with standard TB treatment, are yet to be determined.

Conclusions
In conclusion, the present study showed that n-3 LCPUFA supplementation, administered after the initial inflammatory response in Mtb-infected mice, lowered the bacterial burden in n-3 PUFA-sufficient mice, but not in mice with a low n-3 PUFA status. It further promoted a more pro-resolving lipid mediator profile, lower production of inflammatory cytokines and tended to enhance weight gain. Considering this, n-3 LCPUFA supplementation in the context of a sufficient n-3 PUFA status may be a promising approach as an HDT in TB. The present study emphasises, however, that the timing, the EPA/DHA ratio administered and n-3 PUFA status before supplementation are critical considerations. It further shows that a low n-3 PUFA status before TB infection may be protective, which requires further investigation.