The Anatomy of Fake Nutrition News

On Headlines, Press Releases and SciComm

TL;DR - a study that measured metabolites (somewhat related to diet) in colorectal tumor tissue vs healthy controls wildly extrapolated their results to suggest that seed oils and ultraprocessed foods drive colorectal cancer risk. very basic, super preclinical research that doesn’t even assess diet and puts forward hypotheses about disease pathophysiology should not be the basis of broad dietary guidance to the public.

For those who didn’t follow my early blogging days (nutrevolve & kcklatt on BlogSpot), I used to spend time deconstructing weekly media nutrition hype. It was somewhat fun and instrumental for me as an early academic researcher/registered dietitian, but after a few years, it became like groundhog day - the same spin, from often similar actors, hyping the same type of immediately actionable ‘nutrition’ research while downplaying the same limitations over and over. The media ecosystem became so oversaturated with hyped up nutrition recommendations that it was impossible for the average consumer to discern what is based on quality science (e.g. new guidelines based on rigorous systematic review processes) vs what is more click-based revenue generation. This reality is still very much the case today.

This week, we got a particularly egregious example that I think is quite telling of all the problems that nutrition science communication faces - problems that will only get worse in the era of MAHA. I think this example is worth bringing back the old study breakdown-style blog post as a way to point out all of the ways the media, researchers, their institutions and scicommers can all come together to degrade expertise and push what is essentially misinformation.

From the media fanfare, including ScienceDaily and ScientificAmerican, you’d think a study was published showing that ultraprocessed foods (UPF) drive colorectal cancer, and that this is in part mediated by UPFs that are made with ‘seed oils’. You’d think that, but you’d be wrong.

The actual study was published in the journal Gut. The study was an observational analysis comparing metabolite and gene expression patterns found in colorectal tumor tissue vs healthy tissues. This is a pretty normal study design, particularly given how hard it is to get human tissues for analyses but it is at the very baseline of understanding what’s happening in a diseased tissue and whether any of it is targetable through a therapy. To be clear: there was nothing experimental in this study. We did not manipulate a variable, randomize to a therapy, etc to understand whether any of the study’s findings are actionable. We are just describing in a cross section of time what the lipids (lipidome) and gene expression (transcriptome) look like in tissue from individuals who already have cancer vs those who don’t - we can’t tell whether any differences observed contributed to cancer development, are driving cancer progression, and/or are just molecular noise resulting from the impact of disease on tissue physiology.

To be clear, this research team is not nutrition scientists - they did not try to understand the patients diets at any time point, and they did not measure validated markers of dietary intake/nutritional exposures. Even if they had, given the cross-sectional nature of the study design, we still couldn’t say much, but there was zero effort made to assess diet to be able to begin to link diet to the underlying tissue metabolite and gene expression. Why is everyone interpreting this as a nutrition study then…? Extrapolations on extrapolations (an all too common phenomena in nutrition).

The major analyses in the paper used liquid chromatography tandem mass spectrometry (LC-MS/MS) to assess the lipids found in tumors vs healthy tissue. This was done in an untargeted, discovery approach (trying to capture as much of the lipid species as their method can identify) as well as a targeted approach (measuring a pre-defined list of lipids). The term ‘lipids’ here may be a bit confusing given how often this term is used to describe both components of the diet as well as ‘blood lipids’. This study is measuring tissue lipids, which broadly refer to a number of molecular species with different chemical formulas that are lipid soluble (triglycerides, phosphatidylcholines, phosphatidylethanolamines, cardiolipins, etc) - these lipids are somewhat influenced by dietary intakes (many contain fatty acids that can be diet-derived) but also what the body synthesizes itself, what the tissue takes up from circulation, and the tissue’s specific metabolic program. Diet will contain some of these lipids but these are largely broken down during digestion and processed by many tissues (e.g. intestine, liver, adipose) before entering circulation to be taken up by the colon tissue. Blood lipids typically refers to the lipoproteins we measure that transport lipids between tissues.

Lipids at the tissue level are involved in a number of cellular processes - they influence membrane composition and the activity of the proteins that sit in them, some bind to proteins directly and coordinate cell signaling and gene expression, some are metabolized and their products excreted and signal to nearby tissues. For the purposes of this paper, it’s important to know that lipids are substrates for enzymes that produce products important for coordinating inflammation. It’s super scientifically valid to want to understand what’s happening to all the lipids in the context of tumors vs healthy controls, especially since lipids may be drivers of cancer proliferation and altered lipid metabolism may be targetable for cancer therapy. The authors found a number of lipids that were different between tumors and healthy tissue.

The most striking finding of their metabolite data is the alterations in specific metabolites of arachidonic acid (eicosanoids) that coordinate inflammation. Arachidonic acid is a fatty acid in tissues that can metabolized by different families of enzymes to produce both pro- and anti-inflammatory as well as pro-resolving compounds - the production of these compounds needs to be coordinated in order to maintain cellular homeostasis, and coordinate the response to things like injury. One major increased compound in tumor tissues was 5-HETE, produced from arachidonic acid by an enzyme called 5-LOX/ALOX5. 5-LOX activity on arachidonic acid facilitates the production of ‘pro-inflammatory’ molecules called leukotrienes, with tumor tissues showing an increase in several of these products including LTB4 and LTC4. As expected due to their increased metabolite concentrations, gene expression analyses showed that the enzymes synthesizing these products were also increased in tumor tissue, likely due to the infiltration of immune cells that express high levels of these enzyme into the tumor tissue niche. Other arachidonic acid metabolites that more negatively regulate inflammation, namely prostglandin E2 and D2, synthesized by another class of enzymes (cyclooxygenases), were reduced in tumor tissues; a cyclooxygenase product, 18-HEPE, of the omega 3 fatty acid, eicosapentaenoic acid (EPA), was also reduced, though broadly the differences in omega 3-derived mediators were minimal in tumor vs normal tissue relative to the more significant arachidonic acid mediated changes. The authors go on to validate their metabolite and gene expression observations in this cohort using several public and proprietary datasets, collectively painting the story that in colorectal cancer, there is dysregulation in the production of arachidonic acid-derived mediators that shifts the profile towards a more inflammatory, less anti-inflammatory and pro-resolution phenotype. There’s some additional cool single cell and spatial transcriptomic data worth digging into for those interested in just the science here.

At this stage, you might be wondering why all of the coverage seems to be about seed oils and ultraprocessed foods? The paper on its own really shouldn’t be used to make inferences about diet - again, it’s just assessing what the metabolome and transcriptome of tissue that is already cancerous vs not looks like. There is no evidence from this paper that any of these changes drove the development of cancer or facilitate its proliferation alone though there is a lot of interest in targeting inflammatory processes and several arms of arachidonic acid metabolism in colorectal cancer - this paper is just one piece of a growing puzzle that disturbed arachidonic acid metabolism may contribute to colorectal cancer pathogenesis. The best way to target this, be it diet or specific drugs, will require further intervention studies.

The link to diet, driven by comments from the authors themselves in the media, stems from the fact that the essential omega 6 fatty acid in the diet, linoleic acid, is required in part because it is elongated and desaturated to arachidonic acid. Arachidonic acid can also be found in the diet and is consumed at high amounts when high animal product diets are consumed, but these often gets left out of popular conversations. In tissues, the goal for nutrition scientists is typically to have an appropriate amount of omega 6’s and omega 3 fatty acids to ensure that their pro and anti-inflammatory as well as pro-resolving mediator products are both capable of being generated by tissues to maintain homeostasis. Historically, there was a large focus in nutrition on linoleic acid-rich oils because these lower circulating LDL-C , a risk factor for atherosclerosis. Over time, the independent health benefits of omega 3’s, especially the marine-derived omega 3s, EPA and DHA, began to be appreciated and there has been a large (but not very successful) public health push to increase intakes of these through low-mercury fish foods or appropriate fortification/supplementation.

One might think its quite intuitive that if arachidonic acid in membranes is being used to produce ‘bad’ pro-inflammatory substances in the context of cancerous tissue, occam’s razor is to lower its levels and one to do that is to reduce the intake of its precursor, linoleic acid. The simplistic idea that you reduce the dietary substrate to reduce the amount of product in a tissue might work in some circumstances but for linoleic acid → arachidonic acid, this is a highly highly regulated process. The elongation and desaturation to the right amount of arachidonic acid is so important to cells that it is limited, and we see that consumption beyond 1-2% of calories from linoleic acid results in decreased synthesis of arachidonic acid (there’s good isotope tracing evidence of this) - essentially, cells reduce arachidonic acid synthesis from linoleic when they have enough. Epidemiologically, you also see very low correlations between intakes of linoleic acid/biomarkers of linoleic acid with arachidonic acid levels. Simply eating more doesn’t increase synthesis and thus, just restricting linoleic acid won’t decrease arachidonic acid levels until you get to the point of deficiency, which would harm your healthy tissues. Thus, the recommendation to just avoid seed oils to limit linoleic acid intake and thus arachidonic acid levels in your membranes is not so straightforward even in the context of a healthy metabolism - whether this would even work in the context of a tumor tissue with its own dysregulated metabolism remains fully unexplored. Even if lowering linoleic acid did lower arachidonic acid levels in the tumor, you'd still need to further go on and see if lowering the amount of arachidonic in the tumor tissue meaningfully reduced inflammatory mediator signaling & altered clinical outcomes. We can't say any of that from this study or generally from what is known in all of nutritional biology.

A second theory is that eating more omega 6 linoleic acid reduces the endogenous synthesis of the omega 3 EPA and DHA fatty acids and thus causes a pro-inflammatory phenotype by reducing the synthesis of EPA and DHA derived anti-inflammatory mediators. While the competition between linoleic acid and the EPA/DHA precursor, alpha linolenic acid, is real, evidence from humans demonstrates that very little alpha linolenic acid gets converted to EPA and even less DHA, even when very large doses of ALA are given. This low conversion is likely to be due to high linoleic acid intakes that compete for and saturate enzymes required for its desaturation but it remains unclear whether lowering linoleic acid intakes to very low levels in humans would facilitate high EPA and DHA synthesis and improve inflammatory biomarkers (note: this current paper in Gut did not observe significant disturbances in omega 3 mediators but it remains possible that higher omega 3 tissue levels could be beneficial due to their ability to antagonize arachidonic acid metabolism - again, assuming the hypothesis these arachidonic acid inflammatory mediators are causal in the disease process is true) . To date, the trial literature focusing on omega 3 fatty acids and inflammation has used high levels (~1-3g/d), mimicking high fatty fish consuming populations, to show anti-inflammatory benefits; the literature in huamns has not showed that relying on endogenous EPA and DHA synthesis is enough to elicit this same anti-inflammatory & pro-resolving response. Only a single trial in humans has lowered linoleic acid to very low levels but this was in the context of co-supplementation with EPA & DHA - there was no added benefit to changing fatty acid tissue biomarkers from lowering linoleic relative to just supplementing with EPA and DHA alone. In general, the field is at the stage where there’s no clear benefit to reducing linoleic acid and the anti-inflammatory effects of omega 3’s appear best achieved when consumed in a preformed state, as endogenous synthesis is not reliable. It is reasonable to test in a trial whether high dose omega 3 supplementation is beneficial in colorectal cancer treatment, but entirely speculative and low plausibility that just reducing seed oils (and other sources of linoleic acid) is going to prevent or treat colorectal cancer.

In the context of the current paper, the paper’s lead authors telling the media that people should avoid seed oils because they’re pro-inflammatory, and the media assuming that high UPF intake = high seed oil intake = colorectal cancer are clearly indefensible. The investigation showed largely dysregulated arachidonic acid metabolism at a single time point in colorectal cancer tissue vs healthy tissue. There was no evidence put forth that these observations were due to a higher intake of linoleic or arachidonic acid, of seed oils, or of UPFs. It’s also concerning that the current paper is assessing tissue from individuals who already have colorectal cancer at one point in time, which at best could be used to suggest a treatment for colorectal cancer but the information being out to the public is suggesting limiting seed oils will prevent colorectal cancer, a big stretch. Actual studies assessing intake and biomarkers of linoleic acid intake typically show no relationship or an inverse relationship to colorectal cancer risk - far from supporting this extremely preemptive public nutrition advice from the authors.

I want to highlight a few additional notes about the authors’ approach:

1. Looking at tissue levels of a metabolite that theoretically can be derived from diet to make inferences about dietary intake is pretty problematic. As I stated above, the body transforms what’s eaten in the diet into the form that it likes and thus, there is often nothing close to a perfect correlation between dietary intake and the level of that compound in the body (even worse, the correlation varies between diet and the level across different tissues, because each tissue has its preferred amount/form for facilitating its own homeostasis). For essential fatty acids like linoleic acid, which the body can’t synthesize, there is a reasonable correlation between intakes and levels that get incorporated into healthy tissues (this is mostly validated for red blood cells and adipose tissue), such that we can measure the amount of linoleic acid (expressed as a % of the total faty acids) in these samples and rank groups based on whether they had higher or lower intakes of linoleic acid. If the authors wanted to look at whether diet was a potential contributor the metabolic phenotype they observed, they would have needed to measure linoleic acid levels in the red blood cell membranes to get a feel for what habitual intakes were, and whether these correlated with their other metabolite changes. Alas, this wasn’t done, likely because they don’t have washed red blood cells to do this.

2. Looking at tissue metabolite levels in the way that the authors have done makes making dietary inferences even more problematic. When we measure fatty acids in blood/red blood cells/tissues to try to link back to diet, we measure the total levels of the fatty acids in the cell or in phospholipid membranes - the linoleic acid that we eat gets distributed across these cellular lipid compartments, and the arachidonic acid that is synthesized from it sits in these lipid membranes as a metabolic reservoir capable of being metabolized to its variety of pro and anti inflammatory as well as pro resolving mediators. These measurements are typically done by gas chromatography methods after converting the fatty acids in lipid species to methyl esters. This is not the approach the authors took to measure fatty acids. Rather than get a feel for the total amount of linoleic or arachidonic acid in their tissues, which might indicate the amount in the diet or the total amount of arachidonic acid capable of being metabolized, the authors measured the relative amounts of specific phospholipid species, some of which contain linoleic and arachidonic acid. This measurement is done via liquid chromatography mass spectrometry, and generally does not allow us to make any inference about dietary intakes or whether total arachidonic acid levels were increased, which could better (though far from perfectly) facilitate inferences about whether lowering linoleic acid or increasing intakes of EPA/DHA might be beneficial. I would actually love to know whether the tumor tissues exhibit lower total EPA and DHA levels but confidently inferring that from this approach is limited (for any looking at the paper, figure 2A shows EPA and DHA as well as arachidonic acid, but these are the free fatty acid levels and don’t capture all of the phospholipid & sphingolipid bound as well as triglyceride forms).

   1. Note: even if the authors had done GC on the tissue to get total fatty acid levels, we’d still not readily be able to make inferences about diet given that its very likely the tumor’s metabolism is altered. We have validated data showing that increasing linoleic acid and EPA/DHA intake in the diet increase their levels in red blood cell membranes and tissues in the healthy state, but we would need to validate that this linear relationship holds true for diet and colorectal cancer tissue first - something we currently don’t have data on and would require a controlled feeding trial/intervention to do.

The last thing I want to highlight here is that the source of the rather pre-emptive and pretty uninformed dietary inferences being made from the study are coming from the authors themselves. They even got their university to put out a press release that seems to be the basis of what the media has relied on - a well known phenomena that university press offices are often the sources of exaggerations about their own science. The state of science communication and trust in science feels bleak when the incentives drive scientists to overhype their work to the point of not even accurately describing what they did or why, university press offices eat it up to sell prestige and garner social capital, and an ad-driven media ecosystem eats it up, mostly uncritically. And we wonder why no one trusts nutrition recommendations.

As a basic scientist and registered dietitian, I implore other basic scientists thinking about the translation of their work to reach out to their nutrition colleagues to assess whether their data has strong nutrition implications. Nutrition is a science and there’s decades of work assessing how dietary intake influences metabolism that is not taught or known to many of the folks coming from diverse fields of inquiry to the study of metabolism right now. The public does not need another supposedly well-intentioned person telling them what to eat based on extremely preliminary data.

-KCK