Non-alcoholic fatty liver disease (NAFLD) and atherosclerosis are well recognized as comorbid conditions, particularly in individuals with metabolic syndrome. Accumulating clinical evidence suggests that NAFLD, in fact, contributes to the development of atherosclerosis. A recent meta-analysis of 26 studies (total of 85 395 participants) determined a significant independent association between NAFLD and subclinical atherosclerosis, identified by carotid artery intima-media thickness, arterial stiffness, coronary artery calcification and endothelial dysfunction [1▪▪]. Further independent association between NAFLD and noncalcified coronary artery plaques was provided through a cross-sectional study of 5121 individuals with no prior history of coronary artery disease [2▪▪]. The biological mechanisms responsible for this association are complex, involving hepatic insulin resistance, altered hepatocyte lipoprotein metabolism and dyslipidemia, and chronic hepatocyte inflammation – all in response to hepatic exposure to high concentrations of fatty acids. Adding to this complexity is the likelihood that liver-resident cell types other than hepatocytes, including stellate cells, macrophages and sinusoidal endothelial cells, which are similarly exposed to excess lipid during NAFLD, also contribute to plaque development. Here, we discuss recent evidence implicating non-parenchymal hepatic cell responses to lipid excess in the progression of NAFLD and in the concomitant creation of a proatherogenic environment.
HEPATIC MACROPHAGES AND KUPFFER CELLS
NAFLD progression is partly a consequence of lipotoxicity, which we define as fatty acid-induced cell stress. This process occurs when fatty acid uptake (particularly saturated species) and de novo synthesis exceed the ability of cells to oxidize, export or store them safely as triglycerides. The conditions and processes involved in hepatocyte lipotoxicity are quite well understood, and have been reviewed extensively in recent years [3–5]. Hepatocyte fatty acid excess causes endoplasmic reticulum and oxidative stress, triggering response pathways that lead to impaired insulin signaling, inflammation and apoptosis, which promote disease progression from benign steatosis to nonalcoholic steatohepatitis (NASH).
Upon injury, hepatocytes release the chemokine CCL2 into the circulation, which elicits the recruitment of monocytes to the liver through activation of CC chemokine receptor 2 (CCR2). The contribution of this axis to NAFLD pathogenesis is supported by the finding that NAFLD patients exhibited increased hepatic and serum CCL2 concentrations, the latter of which was associated with increased severity of hepatic inflammation [6▪]. Furthermore, in mouse models of diet-induced steatohepatitis, pharmacological inhibition of CCR2 decreased hepatic accumulation of monocytes [6▪] and monocyte-derived macrophages . Lipotoxic hepatocytes have also been shown to release extracellular vesicles containing the macrophage chemokine CXCL10, in a JNK-dependent and mixed lineage kinase 3 (MLK3)-dependent manner, thereby inducing macrophage chemotaxis . Moreover, pharmacological inhibition of MLK3 reduced macrophage chemotaxis in vitro, and decreased serum CXCL10 and hepatic macrophage infiltration in mice with diet-induced steatohepatitis . Another mechanism implicated in the development of steatohepatitis is hepatocyte pyroptosis, a form of programmed necrosis. Specifically, the pyroptosis protein Gasdermin D (GSDMD) was observed to be elevated in livers of NAFLD and NASH patients. In mice with diet-induced steatohepatitis, genetic ablation of GSDMD decreased hepatic interleukin-1β (IL-1β), tumor necrosis factor α (TNFα) and CCL2, as well as hepatic infiltration of macrophages . Together, these findings denote hepatocellular injury as a key stimulus for the recruitment of inflammatory cells in the progression of NAFLD to steatohepatitis.
Upon recruitment to the liver, myeloid cells infiltrate hepatic tissue through cell–cell adhesion to liver sinusoidal endothelial cells (LSEC). In obese mice, LSEC showed increased expression of cell adhesion molecules, such as VCAM1 and ICAM1, and monocytes extracted from these mice exhibited increased adhesion to LSEC in vitro. This suggests that lipid excess increases cell–cell adhesion, which promotes hepatic infiltration of myeloid cells. Within the fat-laden liver, monocyte-derived macrophages and Kupffer cells are exposed to various lipid species, which can have distinct effects on macrophage phenotype. Kupffer cells exposed to palmitate in vitro were polarized to a proinflammatory M1 phenotype, characterized by increased TNFα and IL-6 expression, whereas exposure to polyunsaturated fatty acids elicited an anti-inflammatory M2 profile, shown by elevated expression of MRC2 and IL-10. Together with the knowledge that NASH patients exhibit increased hepatic palmitate content , these findings suggest that accumulation of palmitate is critical to hepatic inflammation. Consistent with this, exposure of macrophages to excess palmitate in vitro caused intracellular accumulation of palmitate crystals, resulting in lysosomal dysfunction, and subsequent NLRP3 inflammasome activation and IL-1β release . Moreover, hepatic macrophages exposed to cholesterol crystals alone [15▪] or derived from lipid-laden hepatocytes  exhibited NLRP3 activation and IL-1β secretion. The involvement of the NLRP3 inflammasome in liver inflammation has been further corroborated by evidence that inhibition of NLRP3 in genetically and diet-induced obese mice with steatohepatitis reduced plasma IL-1β, CCL2 and IL-6, and reversed hepatic inflammation and fibrosis [15▪].
In lipotoxic conditions characteristic of NAFLD, hepatic macrophages secrete various proinflammatory cytokines, such as IL-1β, IL-6 and TNFα, all of which may enter the circulation and have been directly implicated in the progression of atherosclerosis . However, there may be additional mechanisms through which hepatic inflammatory cells contribute to atherogenesis. Exposure of hepatocytes to TNFα was reported to induce the expression of PCSK9 . Given that PCSK9 is a regulator of hepatic LDL clearance and plasma LDL-C levels, this suggests that hepatic TNFα may disrupt LDL metabolism and contribute to the elevated LDL-C observed in atherosclerosis. A recent study also showed that TNFα increased apoB secretion in mouse hepatocytes in vitro and promoted hepatic VLDL secretion in mice , indicating a link between hepatic TNFα and elevated plasma triglycerides. Moreover, Kupffer cells were identified as the predominant source of plasma CETP, a protein which mediates the exchange of cholesteryl esters for triglycerides between lipoproteins and is associated with dyslipidemia . In obesity-associated NAFLD, hepatic Kupffer cell content and plasma CETP are increased, concomitant with atherogenic dyslipidemia characterized by small dense LDL, elevated triglycerides and reduced HDL . These findings suggest that activation of liver macrophages in NAFLD may alter lipoprotein metabolism and secretion to promote increased plasma LDL-C and triglycerides, and reduced HDL-C, contributing to an atherogenic lipid profile during NAFLD progression.
HEPATIC STELLATE CELLS
Unlike hepatocyte lipotoxicity, our knowledge of hepatic stellate cell (HSC) responses to fatty acid excess is limited, despite the activation and proliferation of these cells in advanced NAFLD, their chronic exposure to lipid excess and their known contribution to fibrosis. We recently determined the sensitivity of human primary activated HSC to high concentrations of saturated and unsaturated fatty acids. Exposure to either high palmitate or high oleate alone induced cell stress, but through different mechanisms. Palmitate stimulated transient expression of the endoplasmic reticulum stress-induced apoptotic factor, CHOP, whereas oleate decreased CHOP expression and increased expression of TXNIP. TXNIP (thioredoxin-interacting protein) can be induced by activation of either PERK or IRE1 during endoplasmic reticulum stress, and can activate the inflammasome under these conditions . Additional evidence from human and rat HSC lines suggests that palmitate, possibly through the generation of the metabolite dihydroceramide , promotes HSC activation and fibrotic activity through XBP-1-mediated induction of autophagy , and inflammasome-mediated hedgehog signaling .
Accumulating evidence of a robust inflammatory response in HSC upon exposure to high fatty acids supports the possibility that these cells could generate inflammatory cytokines that enter the circulation. In direct support of this, Shoji et al. demonstrated that plasma IL-34, derived from liver fibroblasts, is dramatically increased in patients with NAFLD which has progressed to fibrosis. IL-34 has recently been identified as a significant predictor of cardiovascular mortality, and is proposed to contribute to atherosclerosis progression by promoting the release of other proinflammatory cytokines including IL-1β, IL-6 and TNF-α . Similarly, CCL5 and CCL20, both potent chemokines, are increased in serum from individuals with NAFLD/NASH, and originate from both hepatocytes and activated HSC [27,28]. Most recently, circulating IL-6 was independently associated with subclinical atherosclerosis in an NAFLD subgroup of the Multi-Ethnic Study of Atherosclerosis (MESA) cohort [29▪▪]. In fact, IL-6 concentrations stratified NAFLD patients according to their coronary plaque burden. Although this study did not further subdivide NAFLD patients according to severity of liver disease, IL-6 production occurs in a variety of cell types within the liver, including fibroblasts , raising the possibility that increased hepatic IL-6 production directly impacts plaque development. Further work is warranted to determine whether activated HSC-derived proinflammatory cytokines are key mediators of NAFLD-related atherosclerosis.
LIVER SINUSOIDAL ENDOTHELIAL CELLS
LSEC are the most abundant non-parenchymal cells in the liver. Similar to HSC, little is known of the contributions of LSEC lipotoxicity to disease progression, despite the chronic exposure of these cells to excess lipid in NAFLD. LSEC are highly specialized and unique from vascular endothelial cells as they lack a basement membrane and have a multitude of fenestrae that regulate transport of macromolecules, including lipids and lipoproteins, across the sinusoid. Under normal conditions, LSEC maintain homeostatic regulation over hepatic vascular tone, primarily through the production of nitric oxide. An important cross-talk also occurs between LSEC and HSC, which serves to maintain HSC quiescence under homeostatic conditions . However, in the presence of hepatocyte lipotoxicity and injury, LSEC lose their regulatory functions, rapidly become dysfunctional and undergo LSEC capillarization, a process characterized by the development of a basement membrane and loss of fenestrations. This is then followed by angiogenesis [31–33]. In light of recent findings that LSEC dysfunction precedes hepatic inflammation and fibrosis, it is important to recognize that communication between LSEC and other hepatic cell types likely plays a significant role in NAFLD progression.
Accumulation of fatty acids and cholesterol within hepatocytes causes hepatocyte ballooning, which leads to sinusoidal compression, increased intrahepatic vascular resistance (IHVR) and increased shear stress, thereby acting as a mechanical stressor on LSEC to promote dysfunction and capillarization . LSEC dysfunction, elicited by continuous vascular stress, is mainly characterized by decreased nitric oxide bioavailability either through impaired eNOS production, or through its reaction with superoxides . It has also been postulated that impaired LSEC nitric oxide production may be due to eNOS phosphorylation induced by insulin resistance, resulting from hepatocyte lipotoxicity ; however, the specific mechanism underlying this effect remains controversial .
LSEC oxidative stress, as a consequence of LSEC lipotoxicity, may be an important driver of NAFLD progression. Saturated fatty acids, such as palmitate, have been shown to activate toll-like receptor-4 (TLR4) in LSEC . In this setting, TLR4 activation induces NOX1 and subsequent generation of superoxide, which reacts with nitric oxide to form peroxynitrate, thereby reducing nitric oxide bioavailability [37▪]. Oxidative stress can also modulate LSEC cyclooxygenase (COX) activity and downstream prostanoid production to elicit vasoconstriction, which further increases IHVR and thus exacerbates LSEC stress in a vicious cycle . Moreover, COX activation within LSEC can directly contribute to the generation of superoxides, which further depletes nitric oxide bioavailability . As discussed in the next paragraph, LSEC-derived reactive oxygen species (ROS) can enter the circulation, possibly contributing to a proatherogenic environment, in addition to promoting NAFLD progression.
Activated HSC are pivotal drivers of hepatic fibrosis, but are increasingly recognized for their role in promoting LSEC capillarization and angiogenesis leading to ROS generation. An interesting series of feed-forward loops involving HSC activation and LSEC appear to drive NAFLD progression. HSC activation induces LSEC capillarization via hedgehog signaling , and has also been shown to induce LSEC vascular endothelial growth factor expression in a hedgehog-dependent manner to further promote hepatic angiogenesis . Increased LSEC angiogenesis downstream of HSC activation can exacerbate hepatic fibrosis, which further increases IHVR and vascular stress, leading to further stimulation of LSEC ROS production [42,43]. The putative link between LSEC ROS generation and NAFLD progression is supported by the finding that serum reactive oxygen metabolites are significantly higher in patients with advanced NAFLD/NASH . This observation is in line with the concept that superoxides generated by LSEC can enter systemic circulation. Remarkably, serum markers of oxidative stress and increased carotid artery intima-media thickness have been independently associated with NASH . This raises the intriguing possibility that LSEC oxidative stress and superoxide generation are also mediators of atherosclerosis development during NAFLD.
Our targeted review suggests that non-parenchymal hepatic cell types mount some similar responses to lipid excess, and share some common pathogenic mechanisms that likely contribute to the coordinated progression of both NAFLD and atherosclerosis (Fig. 1). In particular, monocyte-derived macrophages, Kupffer cells and HSC initiate robust inflammatory responses upon exposure to high saturated fatty acids, whereas the LSEC lipotoxic response involves ROS generation. Consistent with this, inflammatory cytokines produced by macrophages, Kupffer cells and HSC have been implicated in the progression of NAFLD and atherosclerosis, whereas LSEC-derived ROS may independently contribute to the progression of both diseases. Further investigation will be required to determine the extent to which non-parenchymal cell-derived inflammatory cytokines and ROS play a role in both diseases, and whether liver-targeted therapies for NAFLD can modulate the disease promoting behavior of these cell types.
We thank Richard Zhang for proofreading the article.
Financial support and sponsorship
Work in the laboratory of Nica Borradaile has been supported by the Natural Sciences and Engineering Research Council of Canada (RGPIN-2017–04646), Canadian Institutes of Health Research (201604PJT363245) and Agriculture and Agri-Food Canada/Mushrooms Canada (GF20465). K.M.P. is supported by an Ontario Graduate Scholarship. R.B.W. is supported by a Canada Graduate Scholarship.
Conflicts of interest
There are no conflicts of interest.
REFERENCES AND RECOMMENDED READING
Papers of particular interest, published within the annual period of review, have been highlighted as:
- ▪ of special interest
- ▪▪ of outstanding interest
1▪▪. Zhou YY, Zhou XD, Wu SJ, et al. Nonalcoholic fatty liver disease contributes to subclinical atherosclerosis: a systematic review and meta-analysis. Hepatol Commun 2018; 2:376–392.
Meta-analysis revealed remarkably higher, independently associated risk of subclinical atherosclerosis in NAFLD patients in light of four different indices – carotid artery intima-media thickness, arterial stiffness, coronary artery calcification and endothelial dysfunction.
2▪▪. Lee SB, Park GM, Lee JY, et al. Association between nonalcoholic fatty liver disease and subclinical coronary atherosclerosis: an observational cohort study. J Hepatol 2018; 68:1018–1024.
NAFLD was identified as an independent risk factor for noncalcified plaque in a large cross-sectional study of asymptomatic individuals, suggesting increased cardiovascular risk.
3. Lee J, Ozcan U. Unfolded protein response signaling and metabolic diseases. J Biol Chem 2014; 289:1203–1211.
4. Marra F, Svegliati-Baroni G. Lipotoxicity and the gut-liver axis in NASH pathogenesis. J Hepatol 2018; 68:280–295.
5. Schuster S, Cabrera D, Arrese M, Feldstein AE. Triggering and resolution of inflammation
in NASH. Nat Rev Gastroenterol Hepatol 2018; 15:349–364.
6▪. Parker R, Weston CJ, Miao Z, et al. CC chemokine receptor 2 promotes recruitment of myeloid cells associated with insulin resistance in nonalcoholic fatty liver disease. Am J Physiol Gastrointest Liver Physiol 2018; 314:G483–G493.
Increased presence of CCR2-expressing myeloid cells was observed in liver from NASH patients. Inhibition of CCR2 in mice decreased hepatic inflammation and improved experimental liver disease.
7. Krenkel O, Puengel T, Govaere O, et al. Therapeutic inhibition of inflammatory monocyte recruitment reduces steatohepatitis and liver fibrosis. Hepatology 2018; 67:1270–1283.
8. Ibrahim SH, Hirsova P, Tomita K, et al. Mixed lineage kinase 3 mediates release of C-X-C motif ligand 10-bearing chemotactic extracellular vesicles from lipotoxic hepatocytes. Hepatology 2016; 63:731–744.
9. Tomita K, Kohli R, MacLaurin BL, et al. Mixed-lineage kinase 3 pharmacological inhibition attenuates murine nonalcoholic steatohepatitis. JCI Insight 2017; 2:e94488.
10. Xu B, Jiang M, Chu Y, et al. Gasdermin D plays a key role as a pyroptosis executor of nonalcoholic steatohepatitis in humans and mice. J Hepatol 2017; 68:773–782.
11. Miyachi Y, Tsuchiya K, Komiya C, et al. Roles for cell-cell adhesion and contact in obesity-induced hepatic myeloid cell accumulation and glucose intolerance. Cell Rep 2017; 18:2766–2779.
12. Luo W, Xu Q, Wang Q, et al. Effect of modulation of PPAR-gamma activity on Kupffer cells M1/M2 polarization in the development of nonalcoholic fatty liver disease. Sci Rep 2017; 7:44612.
13. Yamada K, Mizukoshi E, Sunagozaka H, et al. Characteristics of hepatic fatty acid
compositions in patients with nonalcoholic steatohepatitis. Liver Int 2015; 35:582–590.
14. Karasawa T, Kawashima A, Usui-Kawanishi F, et al. Saturated fatty acids undergo intracellular crystallization and activate the NLRP3 inflammasome in macrophages. Arterioscler Thromb Vasc Biol 2018; 38:744–756.
15▪. Mridha AR, Wree A, Robertson AAB, et al. NLRP3 inflammasome blockade reduces liver inflammation
and fibrosis in experimental NASH in mice. J Hepatol 2017; 66:1037–1046.
Selective inhibition of NLRP3 improved NAFLD disease and fibrosis in obese mice, potentially because of the blockade of cholesterol crystal-mediated NLRP3 inflammasome activation in myeloid cells.
16. Ioannou GN, Subramanian S, Chait A, et al. Cholesterol crystallization within hepatocyte lipid droplets and its role in murine NASH. J Lipid Res 2017; 58:1067–1079.
17. Ramji DP, Davies TS. Cytokines in atherosclerosis: key players in all stages of disease and promising therapeutic targets. Cytokine Growth Factor Rev 2015; 26:673–685.
18. Ruscica M, Ricci C, Macchi C, et al. Suppressor of cytokine signaling-3 (SOCS-3) induces proprotein convertase subtilisin kexin type 9 (PCSK9) expression in hepatic HepG2 cell line. J Biol Chem 2016; 291:3508–3519.
19. Song Y, Zhao M, Cheng X, et al. CREBH mediates metabolic inflammation
to hepatic VLDL overproduction and hyperlipoproteinemia. J Mol Med (Berl) 2017; 95:839–849.
20. Wang Y, van der Tuin S, Tjeerdema N, et al. Plasma cholesteryl ester transfer protein is predominantly derived from Kupffer cells. Hepatology 2015; 62:1710–1722.
21. Hetherington AM, Sawyez CG, Zilberman E, et al. Differential effects of palmitate and oleate in activated human hepatic stellate cells and epithelial hepatoma cells. Cell Physiol Biochem 2016; 39:1648–1662.
22. Lee AY, Lee JW, Kim JE, et al. Dihydroceramide is a key metabolite that regulates autophagy and promotes fibrosis in hepatic steatosis model. Biochem Biophys Res Commun 2017; 494:460–469.
23. Kim RS, Hasegawa D, Goossens N, et al. The XBP1 arm of the unfolded protein response induces fibrogenic activity in hepatic stellate cells through autophagy. Sci Rep 2016; 6:39342.
24. Duan NN, Liu XJ, Wu J. Palmitic acid elicits hepatic stellate cell
activation through inflammasomes and hedgehog signaling. Life Sci 2017; 176:42–53.
25. Shoji H, Yoshio S, Mano Y, et al. Interleukin-34 as a fibroblast-derived marker of liver fibrosis in patients with nonalcoholic fatty liver disease. Sci Rep 2016; 6:28814.
26. Tao R, Fan Q, Zhang H, et al. Prognostic significance of interleukin-34 (IL-34) in patients with chronic heart failure with or without renal insufficiency. J Am Heart Assoc 2017; 6:e004911.
27. Li BH, He FP, Yang X, et al. Steatosis induced CCL5 contributes to early-stage liver fibrosis in nonalcoholic fatty liver disease progress. Transl Res 2017; 180:103–117. e104.
28. Chu X, Jin Q, Chen H, et al. CCL20 is up-regulated in nonalcoholic fatty liver disease fibrosis and is produced by hepatic stellate cells in response to fatty acid
loading. J Transl Med 2018; 16:108.
29▪▪. Simon TG, Trejo MEP, McClelland R, et al. Circulating interleukin-6 is a biomarker for coronary atherosclerosis in nonalcoholic fatty liver disease: results from the Multi-Ethnic Study of Atherosclerosis. Int J Cardiol 2018; 259:198–204.
Circulating IL-6 was found to independently associated with the prevalence and severity of subclinical atherosclerosis in an NALFD subgroup of the Multi-Ethnic Study of Atherosclerosis cohort.
30. Schmidt-Arras D, Rose-John S. IL-6 pathway in the liver: from physiopathology to therapy. J Hepatol 2016; 64:1403–1415.
31. Natarajan V, Harris EN, Kidambi S. SECs (sinusoidal endothelial cells), liver microenvironment, and fibrosis. Biomed Res Int 2017; 2017:4097205.
32. DeLeve LD, Maretti-Mira AC. Liver sinusoidal endothelial cell
: an update. Semin Liver Dis 2017; 37:377–387.
33. Poisson J, Lemoinne S, Boulanger C, et al. Liver sinusoidal endothelial cells: physiology and role in liver diseases. J Hepatol 2017; 66:212–227.
34. Baffy G. Origins of portal hypertension in nonalcoholic fatty liver disease. Dig Dis Sci 2018; 63:563–576.
35. Persico M, Masarone M, Damato A, et al. Non alcoholic fatty liver disease and eNOS dysfunction in humans’. BMC Gastroenterol 2017; 17:35.
36. Sutter AG, Palanisamy AP, Lench JH, et al. Dietary saturated fat promotes development of hepatic inflammation
through toll-like receptor 4 in mice. J Cell Biochem 2016; 117:1613–1621.
37▪. Matsumoto M, Zhang J, Zhang X, et al. The NOX1 isoform of NADPH oxidase is involved in dysfunction of liver sinusoids in nonalcoholic fatty liver disease. Free Radic Biol Med 2018; 115:412–420.
Increased expression of NOX1 was observed in liver of NASH patients and of mice fed a high-fat and high cholesterol diet. NOX1 was much higher in cell fractions enriched for LSEC than for hepatocytes, and exposure of primary cultured LSEC to palmitate induced NOX1 expression.
38. Lin L, Cai M, Deng S, et al. Amelioration of cirrhotic portal hypertension by targeted cyclooxygenase-1 siRNA delivery to liver sinusoidal endothelium with polyethylenimine grafted hyaluronic acid. Nanomedicine 2017; 13:2329–2339.
39. Gonzalez-Paredes FJ, Hernandez Mesa G, Morales Arraez D, et al. Contribution of cyclooxygenase end products and oxidative stress to intrahepatic endothelial dysfunction in early non-alcoholic fatty liver disease. PLoS One 2016; 11:e0156650.
40. Zhang F, Hao M, Jin H, et al. Canonical hedgehog signalling regulates hepatic stellate cell
-mediated angiogenesis in liver fibrosis. Br J Pharmacol 2017; 174:409–423.
41. Zhao S, Zhang Z, Yao Z, et al. Tetramethylpyrazine attenuates sinusoidal angiogenesis via inhibition of hedgehog signaling in liver fibrosis. IUBMB Life 2017; 69:115–127.
42. Marrone G, Shah VH, Gracia-Sancho J. Sinusoidal communication in liver fibrosis and regeneration. J Hepatol 2016; 65:608–617.
43. Greuter T, Shah VH. Hepatic sinusoids in liver injury, inflammation
, and fibrosis: new pathophysiological insights. J Gastroenterol 2016; 51:511–519.
44. Shimomura Y, Takaki A, Wada N, et al. The serum oxidative/antioxidative stress balance becomes dysregulated in patients with nonalcoholic steatohepatitis associated with hepatocellular carcinoma. Intern Med 2017; 56:243–251.
45. Leach NV, Dronca E, Vesa SC, et al. Serum homocysteine levels, oxidative stress and cardiovascular risk in nonalcoholic steatohepatitis. Eur J Intern Med 2014; 25:762–767.
* Kia M. Peters and Rachel B. Wilson contributed equally to this work.