PCSK9 (Proprotein Convestase Subtilisin Kexin Type 9) is the ninth member of the proprotein convertase (PC) family 1 . Genetic studies have demonstrated that PCSK9 is a major determinant of cholesterol homeostasis 2 3 . Gain-of-function (GOF) PCSK9 mutations are associated to autosomal dominant hypercholesterolemia and premature atherosclerosis 4 . In contrast, loss-of-function (LOF) PCSK9 mutations lead to low concentrations of plasma low-density lipoprotein cholesterol (LDL-C) and confer protection against cardiovascular disease 5 . PCSK9 is secreted by the liver and acts as a natural inhibitor of the LDL receptor (LDLR) pathway, by targeting the receptor to the lysosomal pathway for degradation 3 . Current clinical trials with monoclonal anti-PCSK9 antibodies and SiRNA show that PCSK9 neutralisation is a promising way to achieve low levels of LDL-C in combination with statins 6 7 .

Several studies have focussed on the metabolic determinants of plasma PCSK9 concentration. Circulating PCSK9 concentrations are associated with LDL-C, plasma concentrations of triglycerides (TG), glucose and insulin in non-diabetic cohorts 8 9 10 11 . The association between plasma PCSK9 and LDL-C, although the most reproducible, is weak as illustrated by the Dallas Heart Study where variations in fasting plasma PCSK9 only accounted for approximately 7% of the variations in LDL-C 11 . At the molecular level, PCSK9 is under the control of the sterol regulatory element binding protein-2 (SREBP-2) 12 pathway and as such it is downregulated by cholesterol 13 . We showed that lipogenic transcription factor SREBP-1c can also regulate PCSK9, acting via the same response element as SREBP-2 in the promoter of PCSK9 14 , suggesting that SREBP-1c might not be dominant under non interventional conditions. One diet that induces SREBP-1 activity is the Fructose enriched diet 15 .

We previously showed that hepatic PCSK9 expression is subjected to nutritional regulation, being decreased upon fasting and increased following re-feeding with a high carbohydrate diet in rodents 14 . Insulin increases hepatic PCSK9 expression both in vitro in hepatocytes and in vivo in mice 14 . Conversely, PCSK9 is repressed by glucagon in rat liver 16 . Accordingly, fasting, but not a ketogenic diet, reduces plasma PCSK9 concentrations in healthy volunteers, with a ≈ 20–35% decrease after 18 h 17 18 . However, so far there are only two reports that describe a dietary modulation of PCSK9 in human. The Mediterranean diet 19 and n-6 PUFAs 20 have been shown to decrease plasma PCSK9 concentrations by ≈ 12% and 13%, respectively.

Here, we assessed the variations of plasma PCSK9 concentrations following various diets that affect differently LDL-C and plasma TG. The aim was to put to test the relationship between PCSK9 and these lipid parameters, as well as the hypothesis that plasma PCSK9 always reflects liver cholesterol synthesis. We used acute fat loads, as well as several short-term dietary interventions (either high-fat or high-fructose diets).

Recent studies have focused on the association between circulating PCSK9 concentrations and metabolic parameters in human 8 9 10 11 . Plasma PCSK9 has been found to be consistently associated with LDL-C, and less robustly with TG, fasting plasma glucose and HOMA-IR. The major finding of our studies is that plasma concentrations of PCSK9 were induced in response to short-term HFruc diets by 27 to 93% in healthy volunteers (Figure 2B-D). Circulating PCSK9 levels were associated with both whole-body, hepatic insulin resistance, liver steatosis and VLDL-TG (Figure 4 and Table 1). In accordance with a regulation of PCSK9 by carbohydrate intake, we previously demonstrated that high-carbohydrate refeeding in mice 14 increases hepatic PCSK9 mRNA and protein levels. Our study does not explain whether changes in plasma PCSK9 upon a high-fructose diet are causative of the variations in VLDL-TG and what could be the molecular mechanisms involved, in particular whether PCSK9 acts upon hepatic VLDL production. In humans, using lipoprotein kinetics with stable isotopes, we observed an increase of VLDL production in 2 family members with PCSK9 GOF variant S127R 33 but it is unclear whether this is related to this specific variant or to a general trait of PCSK9 GOF variants. In mice, we showed that PCSK9 overexpression is accompanied with hypertriglyceridemia due to VLDL overproduction. However this phenotype was restricted to fasted mice, and was not observed in fed mice 34 . Interestingly, we showed that PCSK9 is normally decreased during fasting 14 . We hypothesized that VLDL production was increased due to a lack of re-uptake of nascent VLDL by the LDLR (as described by Twisk J. et al. 35 ). Indeed, fasting seemed to increase the effect of PCSK9 on the LDLR degradation and these mice had virtually no LDLR in their liver compared with fed mice that overexpressed PCSK9. Fructose inhibits hepatic lipid oxidation and favors VLDL-TG-synthesis 36 and it cannot be excluded that PCSK9 was associated with nascent VLDL particles produced by the liver. However, whether PCSK9 is physically associated with lipoproteins remains a controversial issue 8 37 .

Plasma PCSK9 follows a diurnal rhythm that parallels fluctuations of lathosterol to cholesterol ratio 18 . Cholesterol synthesis is driven by SREBP-2 that translocates to the nucleus in response to lower cholesterol content of the endoplasmic reticulum membrane and activates HMG-Coa reductase. SREBP2 also up-regulates PCSK9 12 39 and LDLR expression 39 . In order to estimate how the present diets affect cholesterol synthesis, we measured serum ratios of lathosterol to cholesterol as a surrogate marker of cholesterol synthesis 40 . There was an increase upon HF diets, as previously described 41 (Figure 3A). This might relate to the non-significant trend towards an increase of plasma PCSK9 we observed. It is possible that the trend would turn out to be significant with more subjects. However, upon HFruc diet, there was no change in cholesterol synthesis (Figure 3B), but plasma PCSK9 increased significantly, suggesting that SREBP-2 pathway is not responsible for these changes. We showed that SREBP-1c is able to drive the expression of PCSK9 and that SREBP2 and SREBP1c share the same response element on the promoter of PCSK9 14 42 . Recent studies in hamsters also support the implication of SREBP-1c in PCSK9 regulation 43 . Diurnal fluctuations of PCSK9 that parallel cholesterol synthesis suggest that SREBP2 pathway is dominant over SREBP-1c activation under non-interventional conditions. Our finding that plasma PCSK9 is increased by a high fructose diet but that cholesterol synthesis is not affected (Figure 3B) suggests that PCSK9 regulation is not dependent upon SREBP-2 under this specific diet. Because SREBP-1c is induced by a high fructose diet 15 , it might be responsible for the increase of plasma PCSK9.

Our study underlines the disconnection that might take place between PCSK9 and LDL-C level under specific nutritional conditions. Indeed, the increase of LDL-C (reported in 23 ) under a HF diet was not linked to an increase of PCSK9. Conversely, the increase of PCSK9 under HFruc diet was not associated to an increase of LDL-C (Figure 2B-D). It is surprising that the large increase of circulating PCSK9 seen under fructose (up to 93% in healthy volunteers in HFruc2) was not associated to an increase of LDL-C (Figure 2D). Further studies are needed to unravel the molecular mechanisms involved in this disconnection. It is also unclear why the two studies led to such different magnitude of increase in PCSK9 (23% for the 7 day-long vs 93% for the 6 day-long diet). Subjects had on average similar basal concentrations of PCSK9. It is possible that a peak of concentration occurs at day 6 or before. However, HFruc2 induced a higher hypertriglyceridemia than HFruc1 (+33% in HFruc1 Healthy Patients vs +107% in HFruc2 25 ), suggesting a better efficacy of the diet.

Several elements suggested a potential association between PCSK9 and postprandial lipidaemia, in majority represented by chylomicrons and their remnants. First PCSK9 might influence chylomicron clearance by degrading the LDLR, although there is conflicting data in Familial Hypercholesterolemia patients on the role of the LDLR in chylomicrons clearance 44 45 . Second, we showed that PCSK9-deficiency is associated with reduced postprandial hyperlipidaemia in mice challenged with an olive oil bolus, due to decreased apoB output and a modification of chylomicron size, number and catabolism 28 . Here, we failed to detect any variation in plasma PCSK9 concentrations following the acute oral fat load in healthy volunteers (Figure 1B). In addition, we found that 2 subjects with PCSK9 LOF mutation responded in a similar fashion than controls. It is possible that an olive oil load, similar to what we did in PCSK9 knockout mice 28 , would have changed the outcome of the investigation in these 2 subjects. However, these subjects cannot be considered as entirely deficient for PCSK9 because it is unclear how much wild type PCSK9 is present in the cells of these individuals and because some wild type protein is still being secreted for one of them. The R104CV114A variant is not cleaved and not secreted. The variant exerts a dominant negative effect over the wild type protein 22 . Carriers of PCSK9 R104CV114A have different concentrations of plasma PCSK9 despite being both heterozygous for the mono-allelic double mutation. For one of them PCSK9 was virtually absent from the blood, while for the other carrier concentrations were around 100 ng/ml. We hypothesized that this variability is due to the dominant effect of the variant 22 . Because the variant is not secreted, we assume that plasma PCSK9 in these subjects is the wild type protein. If plasma PCSK9 had a role in postprandial lipemia, these two subjects would have had a different response to the oral fat load. Of course, some limitations to our study are to be taken into account, as discussed below. All together, our data suggest that plasma PCSK9 is not associated to postprandial hyperlipidaemia in human.

Recent studies suggest that PCSK9 may interfere with glucose homeostasis, since: i) insulin increases PCSK9 expression in vitro in hepatocytes and in vivo in mice and rats 14 16 ; ii) the expression of PCSK9 is altered in rodent models of diabetes 46 ; and iii) circulating PCSK9 concentrations were found to be correlated with the level of insulin sensitivity assessed by the HOMA-R index both in adults 11 and in children and adolescents 9 . A recent phenotyping of PCSK9-deficient mice revealed that they were hypoinsulinemic, hyperglycaemic and glucose intolerant 47 . Our own investigations in mice with a different genetic background didn’t point out any obvious abnormality in terms of glucose homeostasis and pancreatic beta cell function 48 . High fructose intake leads to hypertriglyceridemia and hepatic insulin resistance and obesity 49 . Whether the molecule of fructose itself is responsible for these deleterious effects is not established because high sucrose diets leads to similar defects 49 .

We show here that PCSK9 was only positively associated with both whole-body and hepatic insulin resistance in healthy volunteers (including OffT2D) when they fed a short-term HFruc diet, but not under basal conditions (Figure 4 and Table 1). Additionally, we found that PCSK9 is associated with liver steatosis upon HFruc diet, without any correlation with IMCL. Previous characterizations of these subjects 23 24 showed that hepatic steatosis was not accompanied with hepatic insulin resistance when induced by the HF, HFHP diets but that it was under the HFruc diet. Such a positive association between circulating PCSK9 levels and liver TG content assessed by proton magnetic spectroscopy was previously described in the cohort of the Dallas Heart Study, although the level of the correlation coefficient was weak (r = 0.13) 11 . In accordance with a potential link between PCSK9 and liver steatosis, we recently described a positive association between PCSK9 and gamma-glutamyl transferase levels, a marker of hepatic steatosis, in type 2 diabetic patients 50 .

Finally, our study had certain limitations. Although our metabolic phenotyping was exhaustive, the number of subjects is small and potentially limited the ability to detect weak correlations. In addition, the duration of each diet is short (≤ 7 days) and additional studies with extended periods of dietary intervention need to be performed to confirm our observations. The variation in the extent of fructose-induced PCSK9 expression between the HFruc1 (+23%) and HFruc2 (+93%) diets is surprising since both diets were similar in term of fructose and energy intake. The only difference was the addition of maltodextrin in HFruc2 diet. In parallel with a higher increase of PCSK9, the hypertriglyceridemia was also more robust in HFruc2 diet. While some of our preliminary observations in a small number of healthy volunteers (n=6) suggest that high glucose diet also increase circulating PCSK9 levels (+47%, p=0.17) (data not shown), it would be interesting to confirm this observation in a larger number of subjects. Although our findings suggest that plasma PCSK9 concentrations do not parallel cholesterol synthesis under a high fructose diet, these studies were not designed to explore these aspects. In particular diurnal rhythms of cholesterol synthesis and plasma PCSK9 concentrations were not determined in these patients. Concerning the effect of PCSK9-deficiency on post-prandial lipid profile, we only investigated two subjects from the same family with the same PCSK9 LOF and that were not from the same gender. It cannot be excluded that different conclusions would emerge from a study with more subjects or with subjects with a different LOF mutation.

In summary, we demonstrated that circulating PCSK9 levels are significantly increased following a short-term high-fructose diet. Under these specific nutritional conditions, PCSK9 concentrations were positively correlated with insulin resistance, liver steatosis and VLDL-TG concentrations.

BC has served on the advisory panel for Sanofi-Regeneron and Amgen. Kim-Anne Lê is presently an employee of Nestec Ltd. The others authors declare that they have no competing interests.

The author’s responsibilities were as follows : BC and PC were responsible for the design of the study, analyzed the results and prepared an initial draft of the manuscript which was reviewed and modified by all authors; CL, MLB and CLM performed the PCSK9 dosage; YZ, BC and MK recruited the subjects and performed the acute oral lipid load; MB, KAL, FT and LT designed and performed the clinical trial with the HF, HFHP and HFruc diets; CB and RK measured IHCL and IMCL by ¹H-MRS and analyzed the data, BGD was responsible for the statistical analysis. All authors read and approved the final manuscript.