Our gut microbiomes – the many bacteria, viruses and other microbes living in our digestive tracts – play important roles in our health and risk for disease in ways that are only beginning to be recognized.
University of California San Diego researchers and collaborators recently demonstrated in older men that the makeup of a person’s gut microbiome is linked to their levels of active vitamin D, a hormone important for bone health and immunity.
The study, published November 26, 2020 in Nature Communications, also revealed a new understanding of vitamin D and how it’s typically measured.
Vitamin D can take several different forms, but standard blood tests detect only one, an inactive precursor that can be stored by the body.
To use vitamin D, the body must metabolize the precursor into an active form.
“We were surprised to find that microbiome diversity – the variety of bacteria types in a person’s gut – was closely associated with active vitamin D, but not the precursor form,” said senior author Deborah Kado, MD, director of the Osteoporosis Clinic at UC San Diego Health.
“Greater gut microbiome diversity is thought to be associated with better health in general.”
Kado led the study for the National Institute on Aging-funded Osteoporotic Fractures in Men (MrOS) Study Research Group, a large, multi-site effort that started in 2000. She teamed up with Rob Knight, Ph.D., professor and director of the Center for Microbiome Innovation at UC San Diego, and co-first authors Robert L. Thomas, MD, Ph.D., fellow in the Division of Endocrinology at UC San Diego School of Medicine, and Serene Lingjing Jiang, graduate student in the Biostatistics Program at Herbert Wertheim School of Public Health and Human Longevity Sciences.
Multiple studies have suggested that people with low vitamin D levels are at higher risk for cancer, heart disease, worse COVID-19 infections and other diseases.
Yet the largest randomized clinical trial to date, with more than 25,000 adults, concluded that taking vitamin D supplements has no effect on health outcomes, including heart disease, cancer or even bone health.
“Our study suggests that might be because these studies measured only the precursor form of vitamin D, rather than active hormone,” said Kado, who is also professor at UC San Diego School of Medicine and Herbert Wertheim School of Public Health. “Measures of vitamin D formation and breakdown may be better indicators of underlying health issues, and who might best respond to vitamin D supplementation.”
The team analyzed stool and blood samples contributed by 567 men participating in MrOS.
The participants live in six cities around the United States, their mean age was 84 and most reported being in good or excellent health. The researchers used a technique called 16s rRNA sequencing to identify and quantify the types of bacteria in each stool sample based on unique genetic identifiers.
They used a method known as LC-MSMS to quantify vitamin D metabolites (the precursor, active hormone and the breakdown product) in each participant’s blood serum.
In addition to discovering a link between active vitamin D and overall microbiome diversity, the researchers also noted that 12 particular types of bacteria appeared more often in the gut microbiomes of men with lots of active vitamin D. Most of those 12 bacteria produce butyrate, a beneficial fatty acid that helps maintain gut lining health.
“Gut microbiomes are really complex and vary a lot from person to person,” Jiang said. “When we do find associations, they aren’t usually as distinct as we found here.”
Because they live in different regions of the U.S., the men in the study are exposed to differing amounts of sunlight, a source of vitamin D. As expected, men who lived in San Diego, California got the most sun, and they also had the most precursor form of vitamin D.
But the team unexpectedly found no correlations between where men lived and their levels of active vitamin D hormone.
“It seems like it doesn’t matter how much vitamin D you get through sunlight or supplementation, nor how much your body can store,” Kado said. “It matters how well your body is able to metabolize that into active vitamin D, and maybe that’s what clinical trials need to measure in order to get a more accurate picture of the vitamin’s role in health.”
“We often find in medicine that more is not necessarily better,” Thomas added. “So in this case, maybe it’s not how much vitamin D you supplement with, but how you encourage your body to use it.”
Kado pointed out that the study relied on a single snapshot in time of the microbes and vitamin D found in participants’ blood and stool, and those factors can fluctuate over time depending on a person’s environment, diet, sleep habits, medications and more.
According to the team, more studies are needed to better understand the part bacteria play in vitamin D metabolism, and to determine whether intervening at the microbiome level could be used to augment current treatments to improve bone and possibly other health outcomes.
Nearly 15 million people in the United States are living with an autoimmune disease, and this number increases annually (1). In autoimmunity, the immune system recognizes, targets, and causes damage to normal tissues such as skin, kidney, pancreas, the nervous system, joints etc. Vitamin D deficiency has long been associated with systemic autoimmune disease and is suspected to play a role in disease pathogenesis.
Although vitamin D is well-known for its role in calcium homeostasis it also has numerous direct and indirect regulatory effects on the immune system that briefly include promoting regulatory T cells (Tregs), inhibiting differentiation of Th1 and Th17 cells, impairing development and function of B cells, and reducing monocyte activation [reviewed in (2, 3)].
Given its predominantly immunosuppressive effects, vitamin D could be of therapeutic benefit. In fact, many preclinical studies in multiple sclerosis (MS) and colitis models (fewer in arthritis and lupus) have demonstrated benefit to oral or intraperitoneal administration of vitamin D (3).
However, unequivocal benefit has not been achieved in clinical studies, suggesting that the relationship between vitamin D and autoimmunity is more complicated than originally believed. It remains unclear whether vitamin D can act through mechanisms alternative to immunosuppression to impact autoimmunity.
The human microbiome is “the ecological community of commensal, symbiotic, and pathogenic microorganisms” that survive on/in our bodies (4). It consists of 12 different bacterial phyla, with 93.5% classified as Bacteroidetes, Proteobacteria, Firmicutes, Actinobacteria, or Euryarchaeota phyla (5, 6).
Intestinal microbes help us digest foods into compounds and nutrients that can be absorbed and utilized by the body. In the last 10 years, it has become evident that the gut microbiome plays an important role in shaping the immune system, and contributing to health and disease (7–9). The microbiome is of particular interest in autoimmunity due to “molecular mimicry,” the concept that foreign microbial peptides might share structure and sequence similarities with self-antigens and are thus capable of initiating immune cell auto-reactivity.
In this review we explore the interaction between microbiome and autoimmunity and the ways in which vitamin D might influence this interaction to facilitate autoimmune disease.
Gut Microbiome Influences Immune Responses and Autoimmune Disease
Evidence of dysbiosis, alterations in gut flora composition, in autoimmunity is becoming increasingly concrete. However, exactly how the microbiota and the immune system interact directly or indirectly to promote disease remains unknown.
Despite variation in dysbiosis across autoimmune diseases, there is evidence to suggest that specific bacteria differentially promote or inhibit immune responses, collectively implying that there may be a greater polymicrobial influence on inflammatory states.
The gut barrier is a physical and functional barrier between host cells and the external environment composed of outer and inner mucus layers, intestinal epithelial cells, immune cells of the lamina propria, and gut-associated lymphoid tissues (GALT). A mucus layer, produced by goblet cells, physically prevents bacteria from coming into contact with the host. If the mucus layer is breached, a single layer of intestinal epithelial cells acts as the next line of defense.
This layer is composed of specialized epithelial cells including enterocytes, Paneth cells, goblet cells, and microfold cells each providing a unique mechanism of protection ranging from phagocytosis to secretion of antimicrobial peptides and IgA [reviewed by (10)] (Figure 1).
The proposed function of secretory IgA produced by plasma cells ranges widely to include binding bacteria to prevent host interaction, promoting downregulation of inflammatory epitopes or toxin neutralization, and coating bacteria for subsequent interaction with the host immune system in peyer’s patches [reviewed in (11)].
The intestinal epithelial cell layer maintains its impermeability to pathogens and toxins via intact tight junctions. Disruption to any component of the physical or functional gut barrier raises host susceptibility to pathogen invasion and subsequent interaction with the host immune system.
Intestinal Bacteria Influence Inflammation and Immune Responses
Intestinal immune cells are primarily found within the lamina propria and GALT. GALTs are comprised of B cell rich lymphoid follicles reminiscent of lymph nodes, and are associated with overlying specialized follicle-associated epithelium containing M cells that facilitate antigen entry from the intestinal lumen. In some regions of the gut, dendritic cells extend their dendrites through the epithelial layer to sample antigen from the lumen (12). Through these mechanisms, resident dendritic cells and T cells acquire access to luminal antigens and promote B cell differentiation and class switch recombination into IgA-producing cells. Plasmablasts home to the intestinal lamina propria where they differentiate into plasma cells (13).
The resident bacterial community is critical for proper immune function. It has been shown that antibiotic mediated depletion of gut microbiota disrupts this relationship and impairs normal innate immune responses such as type I and type II IFN responses by macrophages (14). The presence of bacteria within intestinal macrophages specifically induces IL-1β without an effect on IL-6, which drives differentiation of Th17 cells (15).
Commensal bacteria coexist within the intestines mediating effects on the immune system in a balancing act that maintains homeostasis [reviewed by (16)]. Whereas, Bacteroides fragilis has an inhibitory effect on Th17 cells, segmented filamentous bacteria (SFB) have a well-documented ability to promote a Th17 response (17–19).
This response is dependent on SFB adherence to intestinal epithelial cells (20) via cell wall glycopolymers which are common to gram-positive bacteria (21); however adherent gram-negative bacteria are also capable of inducing Th17 responses (20).
Furthermore, coinfection with SFB and Listeria monocytogenes generated Th17 and Th1 cells, respectively, demonstrating an important concept that individual bacteria can elicit specific immune cell responses (22). Colonic Tregs are also capable of undergoing expansion in response to certain bacterial species.
For example, a cocktail of Clostridial strains isolated from healthy human fecal samples reduced features of TNBS-associated colitis and allergic diarrhea models via Treg upregulation (23, 24).
In the intestines, B cells primarily localize to the lamina propria (LP). LP B cells were found to express Rag2 and DNA polymerase, characteristics of pro-B cells, which suggests that B cell development may occur in the gut (25). Interestingly, colonized germ free mice (by weaning with specific pathogen free mice for 7 days) had significantly increased Rag1 and Rag2, and increased percentages of pro-B cells (CD19+ B220-low CD43+) in the bone marrow, spleen, and LP compared to their germ free littermates (25).
Furthermore, germ free mice display reduced numbers of IgA+ plasma cells that increased in response to colonization [(26, 27), and reviewed in (13)]. Thus, gut microbiota is associated with and may potentially serve as an antigen source for immature B cell development in the gut.
The interaction between microbiota and B cells also influence one another to maintain homeostasis. In arthritis-induced mice, gut colonization stimulated IL-1β and IL-6 production which promote the development and function of splenic and mesenteric lymph node IL-10-producing B cells (28).
Furthermore, colonic bacterial lysate or specific species such as B. fragilis, have also been shown to stimulate IL-10-producing B cells, which are capable of suppressing T cell mediated inflammation and colitis (29–31). These interactions, by promoting immunoregulatory activity, contribute to our ability to live symbiotically with bacteria.
The Bacteroidetes phylum is the largest phylum of gram-negative bacteria and has a reputation of promoting health. Within this phylum, the Bacteroides genus is the most prevalent in the gut (32). Polysaccharide A (PsA), a component of the B. fragilis cell wall, has been widely studied. PsA induces IL-10 production by intestinal T cells, possibly via ligation of TLR2 on plasmacytoid dendritic cells (33).
Induction of regulatory T cells (Treg) was shown to be dependent on IL-10-producing B cells, and protected against herpes encephalitis (29). Corresponding with this immunoregulatory response, PsA inhibits Th17 cell expansion, while a modified B. fragilis that lacks PsA loses the ability to induce IL-10 production and becomes proinflammatory (34, 35).
As discussed, the literature supports that microbiota promote both humoral immunity (B cell development and proinflammatory T cell responses) as well as immune regulation (regulatory B and T cells). In MS, multiple studies have shown that disease improves with B cell depletion (rituximab, anti-CD20), yet is exacerbated by neutralization of a B cell growth factor (atacicept, TACI-Ig) [reviewed in (36)].
It can be inferred from these data that there are disease-promoting and disease-fighting B cell subsets, and that it is possible that these specific cell subsets are differentially influenced by microbial influences.
Metabolites produced by intestinal bacteria (e.g., short chain fatty acids, lipids, vitamins) also play an important role in immune modulation [reviewed by (37, 38)]. Short chain fatty acids (SCFA; e.g., butyrate, acetate, propionate) are byproducts of dietary fiber fermentation in the large intestine.
In general, the Bacteroidetes phylum primarily produce acetate and propionate, whereas the Firmicutes phylum mainly produce butyrate (39), though this is a simplification. Butyrate and propionate, but not acetate, were shown to promote extrathymic Treg differentiation (40, 41).
Additionally, butyrate leads to a downregulation of LPS-induced proinflammatory cytokine production (i.e., NO, IL-6, IL-12) by intestinal macrophages (42), further supporting butyrate as an anti-inflammatory metabolite. There is also increasing data suggesting that SCFAs help maintain blood-brain barrier integrity (43) which is believed to contribute to neurologic conditions including MS that are being increasingly associated with the gut [reviewed in (38)].
Secondary bile acids (i.e., deoxycholic acid and lithocholic acid) are converted from primary bile acids by colonic bacteria. Activation of bile acid activated receptors by secondary bile acids triggers an anti-inflammatory response characterized by increases in Tgfb, Il10, and Foxp3 gene expression, and suppression of NF-kB mediated expression of proinflammatory cytokines (Il6, Tnfa, Il1b, and Ifng) (44, 45).
Finally, a variety of metabolites including secondary bile acids, fatty acids, and secondary metabolites, act intracellularly to regulate transcription or act on metabolite sensing-G protein-coupled receptors to regulate inflammatory leukocytes, Tregs, and/or modulate the intestinal barrier [reviewed by (46)].
Vitamin D and Immune Defense in the Gut
Vitamin D is well-known for its role in calcium homeostasis and bone growth, but is also well-studied for its anti-inflammatory properties [reviewed by (2, 3)].
Briefly, vitamin D classically acts through the vitamin D receptor (VDR) to regulate gene transcription. Within the immune system, vitamin D inhibits Th17 and Th1 responses, promotes Tregs, impairs B cell development and function, and stimulates antimicrobial peptides from immune cells. In this section we will focus on how vitamin D specifically affects microbiota composition and the gut barrier.
Influence of Vitamin D on Gut Microbiome Composition
It has recently been shown that the composition of the gut microbiome can be altered by vitamin D status/exposure (108, 109). Rodent studies demonstrate that vitamin D deficiency by dietary restriction, lack of CYP27B1, or lack of VDR promote increases in the Bacteriodetes (109–112) and Proteobacteria phyla (109, 110, 112).
Furthermore, a recent GWAS identified two VDR polymorphisms as significant contributors to microbiota variation within a combined cohort of 2029 individuals from the general German population and patients with specific disease entities (e.g., autoimmune disease, metabolic syndrome, sarcoidosis) (113).
In this study, human VDR polymorphisms consistently influenced the genus Parabacterioides (phylum: Bacterioidetes), and subsequent evaluation of VDR−/− mice showed a corresponding increased abundance of Parabacteroides compared to WT mice (113).
Human studies have reported significant associations between vitamin D and microbiome composition.
In a cross-sectional study of healthy individuals, vitamin D intake was negatively associated with abundance of Prevotella and strongly positively associated with Bacteroides, both of the phylum Bacteroidetes (114). In contrast, Luthold et al. found that healthy individuals with higher reported vitamin D intake had greater fecal abundance of Prevotella, and reduced Haemophilus (phylum: Proteobacteria) and Veillonella (phylum: Firmicutes) (108).
In the same study, bacterial enrichment differed in individuals with higher serum 25(OH)D, as they displayed greater abundance of Megaphaera (phylum: Firmicutes), yet maintained a reduction in Veillonella and Haemophilus (108). In a study that utilized endoscopy and colonoscopy biopsies in addition to stool samples, it was found that 8-weeks of vitamin D3 supplementation resulted in increased species richness in the gastric antrum, decreased Proteobacteria (specifically gammaproteobacteria) in the upper GI tract (gastric corpus and gastric antrum), and increased Bacteroidetes (gastric corpus and descending duodenum) (115).
Of note, microbial composition of the lower GI tract and stools did not differ between pre- and post-vitamin D3 treatment (115), suggesting that stool sample analysis may not be the appropriate means to study the effect of vitamin D3 on microbial communities.
In support of this, an observational study did not find an association between habitual vitamin D intake and relative abundance of fecal bacterial genera (116). Whether vitamin D influences microbial composition along the GI tract vs. in stool is of particular importance and raises caution regarding the location of fecal/stool sample collection for future gut microbiome studies.
Furthermore, methodological differences in the assessment of vitamin D “dose” [e.g., sun exposure, reported dietary, and nutritional supplement vitamin D intake, serum 25(OH)D] can lead to inconsistent results between studies.
Surprisingly, little is known about the direct effects of vitamin D on bacteria. This review identified a single study which demonstrated that vitamin D inhibited the growth of specific mycobacterial species in vitro (117). If this finding is corroborated, antimicrobial effects of vitamin D would be consistent with known immunoregulatory properties.
If these findings are not corroborated, then it is likely that microbiota are mediated indirectly by vitamin D’s immunologic properties [reviewed in (3)].
In contrast, there is data to support that bacteria actually influence vitamin D metabolism as some bacteria express enzymes involved in hydroxylation of steroids and thus are capable of processing and activating vitamin D in a manner similar to humans (118).
Bacterial CYP105A1 (Streptomyces griseolus) converts vitamin D3 into 1,25(OH)2D3, in two independent hydroxylation reactions, representing the bacterial functional equivalent of the combined activity of vitamin D metabolic enzymes CYP2R1, CYP27A1, and CYP27B1 (119).
Additional review of a microbial genome database for CYP27A1 and CYP27B1 revealed homologous protein from Ruminococcus torques (Phylum: Firmicutes) Mycobacterium tuberculosis, respectively (120). Capitalizing on these microbial enzymes, there is even a patent (U.S. Patent 5474923) for a process by which hydroxylated vitamin D derivatives are obtained by incubating vitamin D with Nocardia, Streptomyces, Sphinogmonas, and Amycolata. Additional studies are needed to understand the relationship between vitamin D and gut bacteria, and the role of bacteria in maintaining adequate vitamin D levels. Additionally, other factors responsible for modulating this relationship, such as FGF23 which decreases vitamin D in germ free mice, are important to investigate and understand how they influence this process (121).
Vitamin D Supplementation and Changes in Autoimmune Disease Microbiome
Autoimmune disease and vitamin D deficiency are known co-morbidities, such that vitamin D supplementation for autoimmunity is a common practice. Thus, far we have discussed the importance of the microbiome in autoimmunity, as well as the ability for vitamin D to impact microbiome composition along the GI tract.
However, not much is known about how vitamin D supplementation (or deficiency) impacts the microbiome of autoimmune patients. Only a few studies summarized below have addressed this question.
In a 4-week long vitamin D intervention for vitamin D deficient CD patients in remission, there were reduced bacterial taxa and changes in bacterial abundance following supplementation, without an effect in healthy controls (122). Megasphaera and Lactobacillus were enriched at week 4, but still comprised a relatively low abundance overall (122).
A study of active and inactive UC patients found overall microbiota diversity to be unchanged following 8 weeks of vitamin D supplementation, but a significant increase in abundance of Enterobacteriaceae (phylum: Proteobacteria) in UC patients (123). Healthy control mice fed a high vitamin D diet (10,000 IU/kg) displayed reduced species diversity and were enriched with Paludibacter (phylum: Bacterioidetes) and Sutterella (phylum: Proteobacteria), the latter of which was also enriched in DSS colitis mice (124).
Interestingly, DSS colitis mice fed this high vitamin D diet displayed a worsened colitis phenotype compared to moderate (2,280 IU/kg) or no vitamin D (0 IU/kg), suggesting that excessive vitamin D intake could promote a disease-exacerbating microbial community consistent with colitis (124).
Finally, a small study assessed fecal bacterial communities of 2 untreated and 5 glatiramer acetate-treated MS patients, and 8 healthy controls after vitamin D3 supplementation (79). Although there were only 2 untreated MS patients, these patients demonstrated an increase in Faecalibacterium (phylum: Firmicutes), Akkermansia (phylum: Verrucomicrobia), and Coprococcus (phylum: Firmicutes) following vitamin D supplementation compared to healthy controls and treated MS patients (79). Interestingly Faecalibacterium and Akkermansia have been reported in the IBD/colitis literature to be protective against disease (85, 125).
Vitamin D Supports Intestinal and Immune Cell Defenses at the Gut-Immune Interface
The intestinal epithelium is in constant interaction with the external environment. Adequate barrier integrity and antimicrobial function at epithelial surfaces are critical in maintaining homeostasis and preventing invasion or overcolonization of particular microbial species. A healthy intestinal epithelium and intact mucus layer are critical to protect against invasion by pathogenic organisms, and vitamin D helps to maintain this barrier function.
Data supporting a role for vitamin D in maintaining tight junctions stem from studies of VDR−/− mice that demonstrate susceptibility to invasive bacteria and LPS as measured by a reduction in transepithelial resistance.
In contrast, vitamin D supplementation in the setting of functional VDR strengthens the epithelial barrier by reducing paracellular permeability of polarized epithelial cells (126–128). Multiple studies found that vitamin D3/VDR signaling modulates tight junction protein quantity and distribution.
For example, there is reduced expression of ZO-1, occludin, and claudin-1 in DSS-treated Caco-2 cell culture that is at least partially rescued by the addition of 1,25(OH)2D3 (129), and SW480 cells treated with 1,25(OH)2D3 enhanced ZO-1, claudin-1, and E-cadherin protein expression (128).
Similarly, tight junction mRNA transcripts and proteins were reduced in epithelial cell lines exposed to bacteria or LPS, and rescued with 1,25(OH)2D3 supporting a role for bacterial disruption of the barrier (127). In contrast, claudin-2, known to be a “leaky” tight junction protein, was found to be upregulated upon vitamin D3 supplementation (128) and downregulated in VDR−/− mice (130).
As a “leaky” protein that allows movement of ions into the intestinal lumen, claudin-2 expression in the setting of functional vitamin D deficiency may contribute to colitis pathology. Furthermore, deletion of VDR from intestinal or colonic epithelial cells led to profound intestinal epithelial cell apoptosis (131).
Vitamin D upregulates antimicrobial peptide mRNA and protein expression including cathelicidin (132), defensins (133), and lysozyme (112) in vitro, and Ang4 in vivo (134). Antimicrobial peptides, primarily secreted by Paneth cells in the gut, are important mediators of microbiome composition as shown by in vivo studies demonstrating, for example, increased bacterial translocation following Paneth cell ablation and increased susceptibility to colitis or pathogen infection [reviewed by (135)].
Cathelicidins are secreted at surfaces interacting with the external environment where they are capable of forming transmembrane pores in the bacterial cell wall and also have antiviral and antifungal properties [reviewed by (136)].
Defensins are secreted by epithelial cells, Paneth cells, and immune cells, and are important components of the innate immune response in the gut. Loss of VDR expression by intestinal epithelial cells led to abnormal Paneth cells, reduced lysozyme mRNA expression, impaired autophagy, and increases in E. coli and B. fragilis (112). Finally, vitamin D deficiency has been associated with reduced colonic expression of Ang4 and a 50-fold increase in colonic bacterial infiltration in mice (134).
How May Vitamin D Deficiency Affect Intestinal Bacteria and Orchestrate Autoimmunity?
Based on the evidence presented above, we suggest that vitamin D deficiency may affect the microbiome and the immune system hereby contributing to autoimmune disease (Figure 2) as follows:
- Vitamin D deficiency or supplementation changes the microbiome, and manipulation of bacterial abundance or composition impacts disease manifestation.
- Lack of vitamin D signaling due to dietary deficiency or genetic impairment of VDR expression/activity can impair physical and functional barrier integrity (Figure 1). This allows bacteria to interact with the host leading to stimulation or inhibition of immune responses.
- Our natural, innate immunologic defenses may be compromised in the setting of vitamin D deficiency.
As evidenced by in vitro and in vivo studies reviewed elsewhere (3), vitamin D acts directly on immune cells to promote an anti-inflammatory state, and the balance between proinflammatory and anti-inflammatory activity is disrupted in vitamin D deficiency in favor of the former.
Despite the numerous ways in which vitamin D can affect the immune system, vitamin D deficiency alone is insufficient to initiate autoimmunity. However, through its effects on bacterial communities, epithelial integrity, or immune function, vitamin D has the capacity to exacerbate other predispositions, such as genetic polymorphisms (e.g., in VDR, metabolic enzymes, intestinal barrier function, immune function), dietary and environmental factors.
As depicted in Figure 2, the cyclic influences between these factors obscure cause and effect relationships. While a genetic predisposition(s) is required, it is possibly the simultaneous convergence of non-genetic factors upon a genetically susceptible individual that results in autoimmunity.
reference link : https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6985452/
More information: Robert L. Thomas et al. Vitamin D metabolites and the gut microbiome in older men, Nature Communications (2020). DOI: 10.1038/s41467-020-19793-8