Methodological Issues in Studying PAHSA Biology: Masking PAHSA Effects

Branched fatty acid esters of hydroxy fatty acids (FAHFAs) are structurally novel, bioactive lipids that are synthesized in humans, animals, and plants and that have beneficial metabolic and anti-inflammatory effects (Kuda et al., 2016; Yore et al., 2014; Zhu et al., 2018). An in silico analysis predicted that >1,000 FAHFAs could exist in nature (Ma et al., 2015), and >20 FAHFA families have been identified in mammalian tissues (Kuda et al., 2016, 2018; Ma et al., 2015; Yore et al., 2014), with many additional families in plants (Zhu et al., 2018). These families differ by their acyl-chain composition and consist of multiple isomers distinguished by the position of the ester bond between acyl chains. Palmitic acid esters of hydroxy stearic acids (PAHSAs) have been studied most extensively. Levels of multiple PAHSA isomers are reduced in serum and subcutaneous adipose tissue of insulin-resistant mice and humans, and serum PAHSA levels correlate highly with insulin sensitivity in humans (Yore et al., 2014). A single oral dose of 5-PAHSA or 9-PAHSA improves glucose tolerance in aged, glucose-intolerant chow-fed and high-fat diet (HFD)-fed mice (Yore et al., 2014), whereas chronic subcutaneous PAHSA treatment improves both insulin sensitivity and glucose tolerance in chow-fed and HFD-fed mice (Syed et al., 2018). The anti-inflammatory effects of PAHSAs are seen in mouse models of insulin resistance (Syed et al., 2018; Vijayakumar et al., 2017; Yore et al., 2014), colitis (Lee et al., 2016), and type 1 diabetes (Syed et al., 2017).

Pflimlin et al. (2018) recently challenged the finding that PAHSAs have beneficial metabolic effects. We appreciate the attempt to corroborate our findings as reproducibility of results by different laboratories is critical. While these different results may be perceived as a reproducibility issue, numerous methodological differences, some of which were mentioned by Pflimlin et al. (2018) in their “limitations” section, can explain why Pflimlin et al.’s experiments did not detect anti-diabetic effects (Tables S1 and S2). None of the experiments in Pflimlin et al. repeated the protocols used in any previous PAHSA/FAHFA papers.

Choice of vehicles for in vivo studies.

Pflimlin et al. used olive oil, which contains many bioactive, metabolically beneficial molecules including omega-3 and −6 fatty acids, tocopherols, polyphenols, hydroxytyrosols, flavones, lignans (Guasch-Ferré et al., 2017), and FAHFAs (Figure S1A). Hence, olive oil is not an inert vehicle. Figures S1B–S1D shows striking effects of olive oil itself to improve oral glucose tolerance tests (OGTTs) in HFD-fed mice compared to no effects of PEG400/Tween80 used in other studies. We recapitulated the conditions used by Pflimlin et al. (same lard HFD, volume of vehicle, and glucose concentration for OGTT) and found that olive oil markedly reduces glucose excursion (Figure S1B), so it is similar to that of mice on a low-fat diet (LFD) (Figure S1C). In fact, the increase in glycemia at 8 min is even lower than in LFD-fed mice. In contrast, in mice on the same lard HFD, the PEG400/Tween80 vehicle has no effect on glucose tolerance with the conditions used in Yore et al. (1 g/kg glucose) (Figure S1D). With the marked effect of olive oil itself, one would not expect an additional effect of PAHSAs. The beneficial effects of olive oil may result from signaling effects of the constituent molecules, some of which are known to stimulate GLP-1 release (Rocca et al., 2001) and from delayed glucose absorption. In our work, we rigorously tested many vehicles including olive oil and DMSO to study PAHSA biology and eliminated olive oil due to its independent biological effects. PEG400/Tween80 gavage had no effect when 1 g/kg glucose was administered (Figure S1D), as in Yore et al. (2014), although it can affect glycemic excursion with higher glucose doses (unpublished data).

For one OGTT, Pflimlin et al. used the PEG400/Tween80 vehicle (Figure 3D in Pflimlin et al., 2018). However, it was used with the lard diet on which the mice had not developed glucose intolerance. There is no LFD control for the lard diet, but the glucose values in HFD vehicle mice in Figures 3A and 3D in Pflimlin et al. (2018) appear similar to the values for LFD vehicle in their Figure 2A. Hence, it is difficult to interpret these data.

High-fat diets.

Pflimlin et al. studied PAHSA effects in mice on several HFDs. Ideally, if PAHSAs are beneficial, the effects would be present on more than one diet, but not necessarily with a single PAHSA dose in insulin-sensitive mice. Although we showed beneficial effects of constant PAHSA infusion by subcutaneous minipump for 9–13 days (Syed et al., 2018) or daily gavage for 12 days (unpublished data) in chow-fed mice, we have not published effects of a single dose of PAHSAs in young, lean chow-fed mice with normal glucose tolerance and insulin sensitivity. To that point, in the study by Pflimlin et al., there is no evidence of glucose intolerance in their vehicle-treated HFD mice, and data indicating insulin resistance are insufficient, consisting only of elevated insulin levels 4.5 hr after food removal with no statistical difference reported for insulin tolerance test (ITT) (Figure 3B in Pflimlin et al., 2018). In fact, Figure 2B in Pflimlin et al. indicates that glucose tolerance is not different between vehicle-treated LFD and HFD mice using the hydrogenated vegetable diet, in contrast to the glucose intolerance and insulin resistance seen with this same HFD in Yore et al. (2014) and Syed et al. (2018). In addition, glucose tolerance in mice on the high-fat coconut oil diet (Figure S2A in Pflimlin et al., 2018) is not different from that in mice on the corresponding LFD. The olive oil vehicle may be improving the glucose tolerance, but nevertheless we would not expect an additive effect of PAHSAs and olive oil. In one experiment, Pflimlin et al. gavaged mice daily with PAHSAs for 6 days and saw no effects on OGTT (Figure 3E in Pflimlin et al., 2018). This is likely due to the fact that both vehicle- and PAHSA-treated mice received daily gavage with olive oil, which has been shown to improve glucose homeostasis and GLP1 secretion even with short-term treatment (Rocca et al., 2001). In addition, olive oil was used as the vehicle in the OGTT, which would mask the effects of PAHSAs.

Glucose doses for OGTT.

Pflimlin et al. used 2 g glucose/kg body weight and Yore et al. used 1 g/kg. This affects excursion on OGTT, which is critical for detecting improvement with therapeutic agents. We tested both glucose concentrations and found that PAHSA effects in our vehicle (PEG400/Tween80) were more pronounced with 1 g/kg.

Glucose-stimulated insulin and GLP-1 secretion.

Pflimlin et al. reported no effect of PAHSAs on glucose-stimulated insulin or GLP-1 secretion in HFD-fed mice. This may be a result of the olive oil vehicle stimulating GLP-1 secretion (Rocca et al., 2001), resulting in increased insulin secretion so there is no further effect of PAHSAs. However, it is worth noting that we did not report an effect of PAHSAs on GLP-1 or insulin secretion in HFD-fed mice with either acute or chronic treatments. We reported the effect of a single PAHSA dose on GLP-1 and insulin secretion only in aged, glucose-intolerant chow-fed mice (Yore et al., 2014) and the effect of continuous PAHSA infusion by mini pump in ~4-month-old chow-fed mice (Syed et al., 2018). So the data of Pflimlin et al. in HFD-fed mice do not differ from our data. Our results indicate that the beneficial metabolic effects of PAHSAs in HFD mice result primarily from improved insulin sensitivity (Syed et al., 2018).

PAHSA concentrations achieved with treatment.

We designed our studies to achieve a 1.5- to 3-fold elevation in serum PAHSA levels in chow-fed mice since this is the range of changes in PAHSA levels in tissues or serum with altered nutritional and metabolic states (Yore et al., 2014). Since HFD mice have lower PAHSA levels, we sought to restore levels or increase them modestly above chow levels (Syed et al., 2018). However, in their 6-day PAHSA treatment study, Pflimlin et al. achieved massive elevations in serum PAHSA levels (5-PAHSA, 477-fold; 9-PAHSA, >87-fold compared to olive oil alone; Table S2 in Pflimlin et al. and Table S2 in this letter). These massive elevations could engage signaling pathways that are not usually affected by PAHSAs, saturate receptors preventing responses to other ligands, or have other off-target effects. The possibility of saturating receptors for PAHSAs was mentioned by Pflimlin et al., although they thought it was unlikely. In addition, the accuracy of the PAHSA levels is questionable. See the liquid chromatography-mass spectrometry (LC-MS) section below.

In vitro experiments.

Numerous methodological issues are also likely to contribute to the different results in cultured cell systems (Table S1). For example, we found that PAHSAs augment insulin-stimulated glucose transport in 3T3-L1 adipocytes and freshly isolated mouse adipocytes using multiple insulin concentrations (Vijayakumar et al., 2017; Yore et al., 2014). In contrast, Pflimlin et al. did not find PAHSA effects on glucose transport in human SVF-derived adipocytes or a genetically modified L6 myocyte cell line. Critically, Pflimlin et al. studied the effects of PAHSAs on glucose transport in human cells only with maximal insulin concentrations, conditions in which it would be impossible to detect the insulin-sensitizing effects of PAHSAs. In the study using human adipocytes, the lowest insulin concentration used by Pflimlin et al. (1 nM) stimulated glucose transport to the same degree as the highest insulin dose (100 nM). Therefore, under the conditions used by Pflimlin et al., 1 nM insulin is maximally stimulating in their human adipocyte culture system. Thus, no conclusions can be drawn from their in vitro studies regarding whether PAHSAs augment insulin sensitivity. Differences in the incubation time with PAHSAs could also contribute to the different results, although we have seen effects with incubation times ranging from 2.5 hr (Vijayakumar et al., 2017) to 6 days (Yore et al., 2014) (Table S1).

Regarding the inability of Pflimlin et al. to detect augmentation of glucose-stimulated insulin secretion (GSIS) by PAHSAs in pancreatic islets, a recent paper corroborated our results using a completely different methodology, i.e., microfluidics (Bandak et al., 2018). That paper showed that in human islets from both healthy donors and type 2 diabetic donors and in murine islets, 5-PAHSA increased the total amount of GSIS and also improved the dynamic insulin release profiles as indicated by more pronounced insulin oscillations. The absence of PAHSA effects on GSIS in Pflimlin et al. may result from substantial differences in the study protocols compared to Yore et al. and Syed et al. (Table S1). For example, we used 100 human islets/incubation condition while Pflimlin et al. used only 3 islets/condition. Also, no glucose was present in the preincubation buffer in our studies, whereas glucose was present in the studies by Pflimlin et al. This could affect the subsequent response to glucose. For studies of GLP-1 secretion, different cell lines were used that have a different constellation of fatty acid receptors. Assay conditions are also critical since PAHSAs are extremely hydrophobic and need to be maintained at ≥37°C to remain in solution.

General issues.

Other important differences (see Table S1) include (1) the number of mice used for studies of acute PAHSA effects (12–14 per group in our studies; 6 per group in Pflimlin et al.); (2) different mouse suppliers, which can result in different severity of insulin resistance even among mice from the same strain (Ussar et al., 2015); and (3) different treatment duration (for chronic treatment, 4.5–5 months in our studies; 6 days in Pflimlin et al.). Our data also show greater PAHSA effects in a mixed genetic back-ground (Syed et al., 2018), indicating the effects are not limited to one strain.

LC-MS.

The LC-MS protocol used by Pflimlin et al. to measure PAHSA levels did not recapitulate any of the protocols published by three independent groups (Brezinova et al., 2018; Kolar et al., 2018; Kuda et al., 2018; Ma et al., 2015; Syed et al., 2018; Zhang et al., 2016; Yore et al., 2014) (Table S2) or more recently by an additional group (Zhu et al., 2018). Furthermore, no information is provided comparing their protocol to any of the reported procedures to validate their results. As a result, when Pflimlin et al. report that endogenous PAHSA levels are below the limit of quantitation in serum of mice fed the hydrogenated vegetable diet with which we detect quantifiable serum levels, it most likely reflects limitations in their methodology. We (Yore et al., 2014; Vijayakumar et al., 2017; Syed et al., 2018) and others (Brezinova et al., 2018) also detect PAHSA levels in serum of chow-fed mice and some isomers are higher than in HFD-fed mice. In contrast, Pflimlin et al. found levels to be below the limit of detection in mice on several LFDs, including the chow diet we used. We are also puzzled by the Pflimlin et al. observation that administration of equivalent doses of 5-PAHSA or 9-PAHSA results in markedly higher 5-PAHSA levels in comparison to 9-PAHSA levels in serum (Table S2 in Pflimlin et al.). For example, oral gavage of an equivalent dose of 5- or 9-PAHSA resulted in 5-PAHSA concentration of 320 nM but a 9-PAHSA concentration of only 21 nM. Other conditions reported in this manuscript also result in a similarly disproportionate increase in 5-PAHSA over 9-PAHSA for equivalent doses. We have not performed these experiments under these conditions, but when we deliver equivalent doses of 5- and 9-PAHSA, we observe similar fold increases in serum levels of each PAHSA (Syed et al., 2017, 2018). Previous reports from our group (Kolar et al., 2018; Yore et al., 2014; Zhang et al., 2016) and others (Brezinova et al., 2018) noted the possibility of a ceramide contaminant overlapping with 5-PAHSA, which could lead to artificially high estimates of 5-PAHSA. Therefore, we are concerned that the LC-MS protocols employed by Pflimlin et al. are overestimating 5-PAHSA and we strongly caution other groups not to use their protocol to measure serum or tissue PAHSA levels.

In summary, Pflimlin et al. challenge our conclusion that PAHSAs improve glucose control in mice. Many important methodological issues contribute to the different results and several of these were discussed by Pflimlin et al. Most importantly, we demonstrate here that olive oil, a bioactive nutrient used by Pflimlin et al. as a vehicle, markedly improves glucose tolerance in HFD-fed mice (Figures S1B and S1C), which would mask effects of PAHSAs since constituents of olive oil signal through some of the same pathways as PAHSAs. Since vehicles used to formulate therapeutic agents are critical for compound bioavailability, efficacy, and safety (Strickley, 2004), the choice of vehicles can determine whether a compound has therapeutic potential. Furthermore, it is important that other investigators not use the LC-MS protocol published by Pflimlin et al. since it has not been validated and it is likely to detect a contaminant(s).