Huang et al. report on a series of experiments designed to show that saturated fatty acids (SFA), lauric (C:12) and palmitic (C:16) acids specifically, activate inflammatory signaling pathways through toll-like receptors (TLRs, TLR2 & TLR4) 1. Additionally, this article is a response to an article I previously wrote about, Erridge and Samani’s “Saturated Fatty Acids Do Not Directly Stimulate Toll-Like Receptor Signaling” 2. In pursuance of that goal, the author’s repeated several of the experiments of Erridge and Samani, adding a few techniques to try and show limitations in the earlier work that led to an incorrect conclusion.
The central finding of Erridge and Samani was that the commonly used solubilizing vehicles for free fatty acid delivery to cultured cells, bovine serum albumin (BSA) and fetal bovine serum (FBS), were largely contaminated with lipopolysaccharides, lipopeptides, flagella and possibly other molecules of bacterial origin. Thus experiments showing that SFAs activate TLR4 are really showing that bacterial components activate them, and that simply incubating cultured cells with most commercially available BSAs will cause TLR4 activation. They also suggested that the usual reagent used to sequester and inactivate LPS in BSA, polymixin B, is not effective for lipopeptides or flagella. Also, a negative limulus amoebocyte lysate (LAL) test is not sufficient for showing absence of endotoxin as it, too, is only effective for LPS. For more details on that paper, see the original review.
The primary technique used by Huang et al. was a MyD88 -/- macrophage cell line in tandem with polymixin B treatment. As stated previously, polymixin B sequesters TLR4 activating LPS, and MyD88 is a gene whose product is considered necessary for TLR2 signaling. Without the ability to cause inflammation through TLR2, which is what lipopeptides and flagella bind to, and without active LPS, any inflammation produced through SFA treatment of that cell line will be by the SFAs alone. They did in fact show that sodium laurate (lauric acid, C:12) by itself (C:12 is sufficiently soluble without vehicle), as well as C:16 in polymixin B treated BSA and FBS activated inflammatory signaling pathways in MyD88 deficient macrophages. They also confirmed that many commercial BSAs were contaminated with LPS and other endotoxins, but that the TLR activation by BSA treatment alone was attenuated by polymixin B and MyD88 deficiency. Docosahexaenoic acid (DHA, 22:6 (n-3)) attenuated TLR activation by SFA.
The first question to ask is why, if this is a response to Erridge and Samani, Huang et al. did not treat their vehicles with C. antarctica or C. cylindracea lipases, in combination with polymixin B, to reduce endotoxin activity, instead chosing the double knockout macrophage line. That would have been a direct comparison, whereas the MyD88 -/- experiments rest on the concept that all BSA and FBS contaminants not inactivated by polymixin B work solely through TLR2, and that MyD88 is an absolute necessity for TLR2 downstream signaling. If any point in that chain of reasoning is incorrect, the experiment is refuted. Looking more deeply into the results, however, there are more interesting concepts to question.
Figure 1 (B & C)
Figure 1 B, C: TNF-α expression, measured by enzyme-linked immunosorbent assay (ELISA), from RAW 264.7 (murine monocyte line) treated with LPS, synthetic lipopeptides (Pam), muramyl dipeptide (MDP), vehicle (BSA), C:12 (B), or C:16 (C) at different concentrations.
Considering Figure 1, it is true that both C:12 and C:16, in a dose-dependent manner, increase TNF-αin this cell line. Looking closer, one can see that even at the highest dose (150 uM for C:12, 500 uM for C:16) TNF-α is about 3 to 4 times higher than from vehicle, whereas treatment with any of the three bacterial molecules at lower doses (.2 ng/mL to 100 uM) result in TNF-α levels 3-5 times higher than either SFA. Three thousand units are cut out of the Y-axis in order to make visible the increase from SFA, given the peak results from LPS, Pam and MDP. This is not in itself a refutation of their results, but it fits into an alternate explanation of them.
Figure 2 C
Figure 2 C: Same as Figure 1 B & C, but with Polymixin B (PM) treatment
The same pattern is seen in the TNF-α ELISA numbers in Figure 2. Results from SFA are very close to BSA until the highest dose (150 uM for C:12, 500 uM for C:16). PM treatment reduces TNF-α to vehicle levels from LPS but not PAM, C:12 or C:16.
Figure 3 B
Figure 3 B: Similar to Figure 1, but in MyD88 -/- macrophages. Experiment is performed with and without polymixin B treatment
Using MyD88 -/- macrophages, Pam and BSA were equally ineffective at stimulating TNF-α. Polymixin B treatment reduced LPS efficacy. Neither treatment reduced TNF-α from SFA.
Figure 4 E & F
Figure 4 E, F: Interleukin 8 (IL-8) production in THP-1 monocytes from C:16 treatment with and without PM (E) or from C:16 and DHA treatment (F).
Results were largely the same for IL-8 production in THP-1 cells as for TNF-α production in the previous figures. C:16 increased IL-8 in a dose-dependent manner, still much lower than LPS, but was not attenuated with PM treatment as LPS was. DHA attenuated C:16 induced IL-8 in a semi-dose dependent manner, peaking at 5 uM.
One last finding, which they used to explain their results, was that higher concentration FBS attenuated inflammation from SFA and that serum starvation prior to treatment was needed for significant cytokine production. They theorized that reactive oxygen species (ROS) from NADPH oxidase was necessary for SFA induced inflammation, and provided evidence for this by showing apocynin treatment, an NADPH oxidase inhibitor, blocked inflammation from SFA. They also wrote that they believe it is the ability of SFA to stabilize lipid rafts, upon which TLRs “float” in the cell membrane, and PUFAs destabilization of them, which caused the pro and anti-inflammatory actions of those fatty acids, respectively. They also discussed the physical binding of TLR4 by LPS and how it might work with SFA, admitting that there was no evidence for the mechanics. In that way, their ultimate explanation of the results is not far removed from mine, but their conclusion that SFA is inherently inflammatory is.
Before providing an alternate explanation of these results, more background is needed. Firstly, as explained in the series on PUFA as enzyme inhibitors…PUFA are enzyme inhibitors. That is to say that any enzyme yet studied in the context of treatment with PUFA, SFA or fatty-acid free albumin has shown inhibition by PUFA. Receptors, including TLRs, are enzymes. Secondly, there is a phenomenon studied in the relationship of gut physiology to Diabetes whereby endotoxin produced in the gut by Gram-negative bacteria gets into circulation more effectively from meals with SFA than with PUFA or fat free 3. The mechanism is not fully understood, but probably involves bile acids, SFAs, and the micelles they create stabilizing LPS transport into enterocytes, where they are complexed with chylomicrons and released into circulation. This is similar to the description by Huang et al. of SFAs ability to stabilize lipid rafts, and PUFAs inability to do so, in order to facilitate TLR signaling.
I find it most likely that there is, even in the polymixin B treated BSA/FBS, residual endotoxic material. The authors say as much, admitting that LAL assay showed some contamination in the C:12 and C16 reagents themselves, but that this was probably due to interference by the fatty acids. The relatively low rate of activation by SFA compared to endotoxin, even at high concentrations, coupled with the inhibitory action of DHA, point to SFA facilitating the trace endotoxin binding with TLRs and PUFA acting as inhibitor of the enzymatic activity of the receptor. Another piece of supporting evidence is that high concentration FBS attenuated SFA and LPS inflammatory signaling. The high levels of albumin would bind fatty acids and endotoxin with fatty acid moieties, sequestering them both from interaction with TLRs. This fits with the hypothesis that trace endotoxin is the proximal activator of TLRs in these experiments.
Finally, Huang et al. did not address one of the points of Erridge and Samani that I found to be quite compelling. In previous studies, the amount of SFA needed to activate TLRs in a similar manner to LPS was a physiological level, in the hundreds of uM. Fasting can induce FFA levels > 500 uM in healthy, lean subjects 4, and if a large portion of that is SFA you would expect, from the experimentally analogous numbers, to routinely see acute endotoxemia in humans and animals. Of course, there are many other factors involved, such as lipoproteins and albumin binding of endotoxin in circulation, but the point should still be addressed, especially since humans preferentially produce SFAs from carbohydrate, a very strange thing to do if they are indeed so inflammatory. Add that to the list of questions unlikely to be asked by biomedical researchers in the near future.
1. Huang, S. et al. Saturated fatty acids activate TLR-mediated proinflammatory signaling pathways. J. Lipid Res. 53, 2002–13 (2012).
2. Erridge, C. & Samani, N. J. Saturated fatty acids do not directly stimulate toll-like receptor signaling. Arterioscler. Thromb. Vasc. Biol. 29, 1944–1949 (2009).
3. Mani, V., Hollis, J. H. & Gabler, N. K. Dietary oil composition differentially modulates intestinal endotoxin transport and postprandial endotoxemia. Nutr. Metab. (Lond). 10, 1–9 (2013).
4. Vasodilation, E. et al. Elevated Circulating Free Fatty Acid Levels Impair. J. Clin. Invest. 100, 1230–1239 (1997).