Introduction
International guidelines recommend treating patients with atherosclerotic disease with high-dose statins [
20,
41]. For every 40 mg/dL reduction in low-cholesterol diet (LDL) cholesterol by statins the relative risk for major adverse cardiovascular events is curtailed by about 20% [
17]. Intravascular imaging studies of atherosclerotic coronaries documented plaque regression with a reduction in plaque volume and macrophage content alongside an increase in fibrous cap thickness following statin treatment [
24].
Statins lower LDL-cholesterol levels by inhibiting hydroxy-methylglutaryl-coenzyme A reductase-dependent cholesterol biosynthesis and inducing LDL receptor expression in the liver. LDL-cholesterol lowering also reduces C-reactive protein. These and other findings have fueled the debate as to whether the beneficial effects of statins are primarily attributed to lipid lowering or to pleiotropic, i.e. anti-inflammatory and directly vasoprotective, effects [
23,
40]. Multiple in vitro studies documented lipid-independent effects of statins on macrophages and endothelial cells, and statin treatment of Apolipoprotein E (Apoe)-deficient mice limited atherogenesis and monocyte recruitment without affecting cholesterol levels [
3,
21,
51,
69]. In contrast, APOE*3Leiden.Cholesteryl ester transfer protein (CETP) mice, a translational mouse model with a humanized lipoprotein metabolism [
70], respond to oral statin treatment with (V)LDL-cholesterol lowering, and showed attenuated plaque formation when fed a Western diet [
25]. The authors proposed impaired monocyte recruitment into atherosclerotic lesions and suppressed inflammation as underlying mechanisms [
26,
64].
While recruited monocytes give rise to plaque macrophages, we showed that local macrophage proliferation dominates cell renewal in established lesions in atherosclerotic mice [
49]. Macrophage proliferation in atherosclerotic lesions, particularly in foam cell rich areas, has been described before [
19,
45,
50], but its relevance to plaque development was not known. When monocyte production and recruitment are attenuated experimentally, new onset atherogenesis is limited, while interventions in established disease fail to slow plaque progression as lesional macrophages continue to proliferate [
29]. Although plaque regression is the ultimate goal in cardiovascular preventive medicine, controversy still surrounds the mechanisms that control the decline in plaque macrophages. Reduction in monocyte influx [
13,
42], requirement of differentiation of infiltrating monocytes into reparative macrophages [
43], and macrophage emigration [
16,
30] were reported to determine plaque regression. In this study, we examined the relative contribution of monocyte infiltration, macrophage proliferation, death and egress in APOE*3Leiden.CETP mice, which model human-like lipid changes in response to oral statin treatment. We show that monocyte-to-macrophage differentiation is a relatively rare event, both in plaque progression and regression, and that the decline in macrophage numbers in regressing plaques mainly results from the suppression of cholesterol-driven local proliferation. Notably, plaque lipid contents and serum cholesterol levels in patients undergoing carotid endarterectomy positively correlated with local macrophage proliferation, supporting the rationale for targeting macrophage proliferation therapeutically.
Methods
Animals and diet
Female APOE*3-Leiden.CETP mice were purchased from TNO (Leiden, Netherlands). We used female mice as they are more susceptible to cholesterol-containing diets by having higher plasma cholesterol levels relative to APOE*3-Leiden.CETP males, with established dose-dependent responses to statins [
63,
70]. Correspondingly, we purchased female Apoe
−/− mice (B6.129P2-Apoe
tm1Unc) from The Jackson Laboratory (Bar Harbor, ME, USA). 8-week-old mice were fed a high-cholesterol diet (HCD, 1.25% w/w cholesterol, D12108 mod., Ssniff GmBH, Soest, Germany) ad libitum for 12 weeks to accelerate atherogenesis, followed by a low-cholesterol diet (LCD, 0.05% w/w cholesterol, semi-synthetic diet T, TNO, Leiden) for 4 weeks to lower proatherogenic plasma cholesterol to levels that allow for normalization through intervention. Next, mice were randomly assigned to up to three groups: 1. continued LCD (control group), 2. LCD supplemented with 0.01% w/w atorvastatin (Pfizer) corresponding to 10 mg/kg body weight per day (statin group), and 3. diet T without cholesterol (free group) for another 4 weeks. To translate the dosing used in our mouse studies (10 mg/kg/d) to human dosing, the following simplified calculation based on body surface area, as accepted by the FDA, can be used as a guide: mouse dose/12.3 × human body weight [
36]. Accordingly, 10 mg/kg/day atorvastatin in mice correspond to 65 mg/day for an 80 kg human.
Female Ldlr
–/– mice (B6.129S7-Ldlr
tm1Her/J) were lethally irradiated (10 Gy) and reconstituted with a 1:1 mixture of bone marrow cells from CD45.1 C57Bl/6 (B6.SJL-Ptprc
a Pepc
b/BoyJ) and CD45.2 Msr1
–/– (B6.Cg-Msr1
tm1Csk/J) or CD45.2 CD36
–/– (B6.129S1-CD36
tm1Mfe/J) mice, as previously described [
49]. Mice were purchased from The Jackson Laboratory (Bar Harbor, ME, USA). Following 6 weeks of reconstitution consuming chow diet, mice were fed HCD (1.25% w/w cholesterol, D12108 mod., Ssniff GmBH, Soest, Germany) for 12 weeks to induce atherosclerosis.
Mice were housed under specific pathogen-free conditions
Patients
Twenty-three patients scheduled for elective carotid endarterectomy because of significant carotid artery stenosis gave written informed consent to having their blood and endarterectomy specimens analyzed as approved by the Institutional Review Board of the University Hospital of Freiburg. Blood and tissue samples were collected within 12 h after the last intake of statin drugs if applicable. Detailed patient characteristics are listed in supplemental table 1. Liver tissue biopsies were sampled from three patients during bariatric surgery that had been on oral atorvastatin 40–80 mg/day. Detailed patient characteristics are listed in supplemental table 2.
In vivo cell labeling
Four weeks into the high-cholesterol diet, 1-μm large yellow-green fluorescent beads (Fluoresbrite YG plain microspheres, Polysciences Inc., Eppelheim, Germany), diluted in 1:4 sterile PBS, were injected intravenously into APOE*3-Leiden.CETP and Apoe−/− mice, respectively, for in vivo cell labeling and tracking of myeloid cells. Fluorescent bead accumulation in the plaque was quantified by immunofluorescence histology. 1 mg bromodeoxyuridine (BrdU) (BD Bioscience, San Jose, CA, USA) was injected intravenously 2 h prior to killing to label proliferating macrophages. BrdU incorporation into CD68+ macrophages in the plaque was quantified by immunofluorescence histology.
Partial body irradiation and bone marrow transfer
Female UBC-GFP mice (C57BL/6-Tg(UBC-GFP)30Scha/J) were purchased from The Jackson Laboratory (Bar Harbor, ME, USA) as donors for bone marrow cell transplantation. APOE*3Leiden.CETP mice with established atherosclerosis following 11 weeks of HCD were lethally irradiated with 10 Gy (RS-2000 Pro Biological System; Rad Source, Buford, GA, USA) while shielding the torso with a lead casing (Belly Shield, Braintree Scientific Inc, Braintree MA, USA, placed on a ¼-inch-thick lead plate) to protect the thoracoabdominal aorta from radiation. 1 × 107 GFP+ bone marrow cells were intravenously transplanted into partially irradiated mice.
Modified LDL uptake by bone marrow-derived Mϕ 2 × 105 bone marrow cells were isolated from CD45.1 C57Bl/6 wild-type (WT), CD45.2 Msr1–/– and CD45.2 CD36–/– mice which were used to generate mixed irradiation bone marrow chimeras in Ldlr–/–, as described above. Bone marrow cells differentiating into bone marrow-derived Mϕ (BMDM) in the presence of 30 ng/ml colony stimulating factor 1 (Peprotech, Rocky Hill, NJ, USA) over 5 days were stimulated with DiI-medium oxidized LDL (10 μg/ml), DiI-acetylated LDL (1 μg/ml) (Kalen Biomedical, Montgomery Village, MD, USA) or PBS for 4 h before being detached from 24-well plates using trypsin/EDTA. Cells were stained with fixable viability dye (Thermofisher Scientific, Waltham, MA, USA) and anti-F4/80 (Biolegend, San Diego, CA, USA) for flow cytometric quantification of percent viable BMDM and mean fluorescence intensity (MFI) of DiI-labeled modified LDL into the respective cells.
Histology
Murine aortic roots and arches were embedded in Optimal cutting temperature (OCT) Tissue Tek (Sakura Finetek, Tokyo, Japan) and cut into serial cryostat sections (5 μm) starting at the level of the aortic valve (for root sections). Sections were stained with Oil-red O (Sigma-Aldrich, St. Louis, MO, USA), Masson Trichrome (Sigma-Aldrich, St. Louis, MO, USA), TUNEL (DeadEnd™ Fluorometric TUNEL System, Promega, Mannheim, Germany), anti-CD68 (clone FA-11, BioRad AbD Serotec, Puchheim, Germany), anti-BrdU (GTX128091, GeneTex, Irvine, CA, USA and ab1893, Abcam, Cambridge, UK), anti-GFP (ab290, Abcam, Cambridge, UK). Human carotid endarterectomy specimens were embedded in OCT, frozen, sectioned, permeabilized with 0.1% Triton X-100, and stained with with anti-Ki67 (ab15580, Abcam, Cambridge, UK), anti-CD68 (clone PG-M1, Agilent, Santa Clara, CA, USA), and Hoechst 33342 (Thermofisher Scientific, Waltham, MA, USA). Adjacent slides of the same plaques were stained for Oil-red O (Sigma-Aldrich, St. Louis, MO, USA). Secondary antibodies included rabbit–anti rat biotin conjugated (BA-4001) followed by ImmPACT AMEC Red Substrate (Vector Laboratories, Burlingame, CA, USA), rabbit–anti rat TRITC (PA1-28570, Thermofisher Scientific, Waltham, MA, USA), rabbit–anti rat AF647 (ab169349, Abcam, Cambridge, UK), goat–anti-rabbit (BA-1000) followed by fluorescein avidin DCS (A-2011, Vector Laboratories, Burlingame, CA, USA), donkey-anti-sheep TRITC (ab6897, Abcam, Cambridge, UK), alpaca-anti-rabbit AF488 (ChromoTek, Planegg-Martinsried, Germany), and DAPI Mounting Medium (Carl Roth, Karlsruhe, Germany) according to the manufacturers’ instructions. Images were recorded with the Axioplan 2 and Apotome 2 imaging light-/fluorescence microscope with an AxioCam camera (Carl Zeiss MicroImaging GmbH, Göttingen, Germany) and a confocal Leica TCS SP8 X microscope (Leica Microsystems, Wetzlar, Germany). Images were analyzed with Image Pro Premiere 9.2 (Media Cybernetics, Silver Springs, USA) and Zeiss Zen lite (Carl Zeiss MicroImaging GmbH, Göttingen, Germany).
Flow cytometry
Murine aortic cells were retrieved through enzymatic digestion with collagenase I, collagenase XI, hyaluronidase, DNAse I and HEPES solution (Sigma-Aldrich, St. Louis, MO, USA) in a thermocycler for 45 min at 750 rpm and 37 °C. Murine blood samples were lysed in RBC lysis buffer (Biolegend, San Diego, CA, USA). Isolated cells from the blood and aorta were counted using a Neubauer chamber (Marienfeld, Lauda-Königshofen, Germany). Cells were stained with specific fluorescent antibodies as indicated (Supplemental Table 3). Ly6Chigh monocytes were identified as CD45+ CD11b+, Lin− (Lin = CD3, CD19, NK1.1, Ly6G), Ly6Chigh, CD115+, F4/80low. Macrophages were identified as CD45+ CD11b+, Lin−, Ly6Clow, F4/80high. Intracellular staining with anti-Ki67 and anti-active Caspase 3, BD Cytoxfix/Cytoperm (#554,722, BD Biosciences, San Diego, CA, USA), BD Perm/Wash (#554,723) and BD Permeabilization Buffer Plus (#561,651) was conducted according to the manufacturer’s instructions. Data were collected on a BD Facs Canto II (BD Bioscience, San Diego, CA, USA) and analyzed with FlowJo (Treestar, Ashland, OR, USA).
Real-time polymerase chain reaction (PCR)
RNA was extracted from murine aortas using Qiazol and RNeasy Mini Kit (Qiagen, Valencia, CA, USA) according to the manufacturer’s instructions. Quantitative TaqMan-PCR was performed using a Bio-Rad CFX96 Touch Real-Time PCR System and TaqMan probes Mm00443258_m1 (Tnfa), Mm01336189_m1 (IL1ß), Mm00446190_m1 (IL6), Mm00439614_m1 (IL10), Mm01178820_m1 (Tgfb1), Mm01320970_m1 (Vcam1), Mm00441242_m1 (Ccl2). Data were statistically analyzed using the 2−▵Ct method.
Lipid and enzyme-linked immunosorbent assays
Murine plasma cholesterol and triglyceride levels were measured using cholesterol and triglycerides FS 10′ Multi-purpose kits (DiaSys Diagnostic Systems GmbH, Holzheim, Germany) according to the manufacturer’s instructions. Lipoprotein profiling in murine plasma samples was conducted by LipoSEARCH services (Skylight Biotech Inc., Akita, Japan). Serum Amyloid A levels were measured using a SAA Mouse ELISA Kit (Thermo Fisher Scientific, Waltham, MA, USA), and ApoB levels were measured using the Mouse ApoB ELISA Kit (Abcam, Cambridge, UK) according to the manufacturers’ instructions.
Mass spectrometry
Plasma and tissue samples were weighed, snap frozen and sent to ImaBiotech SAS (Lille, France) for the quantification of atorvastatin concentration by mass spectrometry. In brief, matrix-assisted laser desorption ionization–Fourier transform ion cyclotron resonance (MALDI–FTICR) was used for imaging mass spectrometry analysis of atorvastatin in murine frozen liver samples sectioned onto indium–tin–oxide glass slides and covered with DAN MALDI matrix (10 mg/ml) in acetonitrile:water 1:1 (v/v). Recordings were made using MALDI–FTICR in negative ionization mode and CASI (continuous accumulation of selected ions) mode centered on m/z 564.2917 ± 50 Da mass range with a laser set at 300 shots, 1 kHz to follow atorvastatin in the sections at 200 µm spatial resolution with a SolariX mass spectrometer (Bruker Daltonics, Bremen, Germany). FTMS Control 2.0 and FlexImaging 4.1 software packages (Bruker Daltonics, Bremen, Germany) were used to control the mass spectrometer and set imaging parameters. For liquid tomography–tandem mass spectrometry (LC–MS/MS)-based quantification of atorvastatin, murine tissue samples were extracted with methanol, and murine plasma and human samples were extracted with acetonitrile, and spiked with the internal standard atorvastatin-d5, 1.5 nM. The LC–MS/MS consisted of an UHPLC Ultimate 3000 coupled with TSQ Quantiva (Thermo Scientific, Courtaboeuf, France) equipped with a Cortecs C18 2.7 µm; 2.1 × 30 mm column (Waters, Saint-Quentin-en-Yvelines, France).
Statistics
Results are presented as mean ± SEM. Differences between two groups were analyzed with unpaired Student’s T test or Mann–Whitney test as indicated in the figure legend. To assess differences between more than two groups, one-way ANOVA with Holm–Sidak’s multiple comparisons testing or Kruskal–Wallis with Dunn’s multiple comparisons testing were applied. p values ≤ 0.05 denote significant changes. Pearson’s correlation coefficient was used to test for correlation.
Discussion
We set out to investigate the cellular processes that determine plaque regression during statin treatment. Plaque progression depends on macrophage accumulation, and, conversely, macrophages vanish during plaque regression [
16,
60]. Multiple processes determine macrophage accumulation in the plaque, including monocyte recruitment and differentiation, macrophage proliferation, death and egress [
32]. Macrophages of prenatal origin reside in the adventitia and self-sustain through proliferation into adulthood largely independent of monocyte recruitment [
15,
65]. Whether these embryonically derived macrophages in the adventitia directly contribute to macrophages accumulating in intimal lesions is unknown. A small population of intima resident macrophages shows limited proliferation capacity and is marginalized during plaque progression [
68] as monocytes infiltrate giving rise to proliferating macrophages [
49].
Previous experimental studies in atherosclerotic mice showed that both macrophage accumulation and plaque growth are limited by statin treatment. Most of these studies were conducted in Apoe
–/– and Ldlr
–/– mice, which do not respond to statins with a significant reduction in serum LDL-cholesterol at clinically relevant doses [
6,
10,
34,
35]. This was interpreted as a proof of pleiotropic drug effects. Protective mechanisms proposed included suppression of endothelial cell adhesion molecule expression, stimulation of endothelial nitric oxide synthase and inhibition of leukocyte integrin LFA1 [
2,
40,
66]. As a result of these molecular effects, monocyte recruitment to atherosclerotic lesions would be reduced, a finding we observed in our study as well. It was thought that impaired monocyte infiltration limited macrophage accumulation in plaques following statin treatment, but this has not been tested. Our model of irradiating mice with established atherosclerosis and aortic shielding followed by GFP
+ bone marrow transplantation, allowed us to quantify the contribution of blood monocytes to the plaque macrophage pool with and without cholesterol lowering. We calculated that during 4 weeks of plaque progression, only 11% of macrophages derived from newly recruited monocytes. These numbers matched our previous estimate of 13% monocyte contribution to the macrophage pool, based on parabiosis and BrdU incorporation in Apoe
–/– mice [
49]. Thus, inhibition of monocyte recruitment, as observed with atorvastatin, could not fully explain the large reduction in macrophage numbers by 40–50% in atherosclerotic lesions following atorvastatin treatment.
We, therefore, investigated cellular processes localized within the plaque, i.e. macrophage proliferation, death or egress. Indeed, macrophage proliferation, as determined by intracellular Ki67 staining and BrdU incorporation, was reduced by almost 50% by atorvastatin treatment. A previous study reported that lipophilic simvastatin, incorporated into a synthetic HDL particle, invaded the plaque, locally inhibiting macrophage proliferation and decreasing macrophage numbers in Apoe
–/– mice, without affecting serum cholesterol levels [
14,
59]. Multiple in vitro studies also documented direct anti-inflammatory and anti-proliferative effects of statins on macrophages in culture [
5,
52,
62]. When we treated APOE*3-Leiden.CETP mice with oral atorvastatin, achieving plasma drug levels comparable to those in patients treated with 40–80 mg atorvastatin per day, the drug was not detected in atherosclerotic arteries, unlike the nanoparticle approach. In line, atorvastatin was also hardly detectable in human carotid artery plaques. The absence of drug accumulation in plaques does not formally exclude the possibility of local pharmacological effects. However, two additional findings in our study argue against relevant direct pleiotropic statin effects on macrophage proliferation within the plaque. First, diet induced plasma cholesterol lowering, alone, to levels achieved with oral atorvastatin treatment in APOE*3-Leiden.CETP mice yielded similar results with regard to plaque regression, dampening of systemic and local inflammation, and inhibition of macrophage proliferation in the plaque. Second, when we treated atherosclerotic Apoe
–/– mice with oral atorvastatin, plasma cholesterol levels remained elevated, and macrophage counts and proliferation rates were unaffected. This finding is in accordance with a previous observation we made when inhibition of monocyte production, lesion infiltration and differentiation in Apoe
–/– mice with established atherosclerosis failed to slow plaque progression and accumulation of macrophages, which continued to proliferate in situ [
29]. When we deleted modified lipoprotein uptake-mediating scavenger receptors (Msr1 or CD36) in macrophages accumulating in murine atherosclerotic aortas next to scavenger receptor expressing macrophages, their proliferation was relatively suppressed. These data support our hypothesis that the uptake of cholesterol-rich modified lipoproteins stimulates macrophage proliferation in atherosclerotic lesions, directly, but the intracellular signaling pathways remain to be determined. Notably, intracellular cholesterol was also found to stimulate hematopoietic stem and progenitor cell proliferation in atherosclerotic mice [
33]. Oxidized LDL provokes colony-stimulating factor 1 (Csf1) secretion by endothelial cells, for example [
44]. A recent study described increased macrophage survival and proliferation in the plaque in response to Csf1 production by endothelial cells and vascular smooth muscle cells, in particular [
57]. This may represent an additional, indirect mechanism by which statins, via systemic reduction in cholesterol-rich LDL particle numbers, inhibit lesional macrophage proliferation.
In support of our lipid uptake hypothesis, a recent experimental study comparing gene expression profiles of lipid-rich and lipid-poor macrophages from atherosclerotic murine aortas reported (in the supplement) that proliferation marker Ki67 expression was almost doubled in lipid-rich cells [
22]. According to a recent meta-analysis of leukocyte diversity in atherosclerotic mouse aortas, based on single-cell RNA sequencing, these lipid-rich foamy macrophages correspond to the Trem2 (triggering receptor expressed on myeloid cells-2) macrophage subset specialized in lipid metabolism, distinct from inflammatory, resident-like and interferon-inducible macrophage subsets [
71]. These macrophage populations partially overlap with subsets identified in human plaques [
12,
18]. Ingenuity pathway analysis of a single cell dataset of aortic macrophages isolated from Apoe
–/– mice described enrichment of proliferative, survival and motility genes in one of four macrophage clusters, as opposed to inflammation, apoptosis and phagocytosis related genes in the other clusters [
31]. While these transcriptional data substantiate the presence of proliferating macrophages in atherosclerotic aortas, they do not inform on whether all subsets of lesional macrophages are equally prone to undergo cell cycling in situ. The plasticity of lesional macrophages is the focus of ongoing research. Macrophage-like cells that express macrophage markers such as CD68 and Lgals3, but not leukocyte marker CD45, may arise from vascular smooth muscle cells [
54], which clonally expand in the plaque [
9]. They are reported to account for 16–30% of macrophage marker-positive cells in the plaque [
1,
54]. In our study in APOE*3-Leiden.CETP mice, lowering of cholesterol and atherogenic ApoB-lipoprotein levels reduced proliferation of CD68 expressing cells in general in the plaque, and this may affect macrophages of both leukocyte and vascular smooth muscle cell origin.
The relevance of macrophage egress for plaque regression is controversially debated with several papers arguing against egress [
28,
29,
42,
67] and a number of papers that argue for egress [
16,
30,
55]. The discrepancies may arise from differences in the models used to induce normolipidemia and timing related to bead transfer. In our models of drug- and diet-induced cholesterol lowering in APOE*3-Leiden.CETP mice, and of statin treatment of Apoe
–/– mice over 4 weeks, we found no indication for significant macrophage egress from established plaques, in line with the aforementioned papers.
Translating our experimental findings to humans, we showed that the levels of serum cholesterol lowering achieved in patients undergoing carotid endarterectomy is inversely correlated with the frequency of plaque macrophage proliferation. The more lipids one detects in the plaque, the higher the proportion of proliferating macrophages in situ. Studying the effects of statins, the most widely used and potent drugs in cardiovascular secondary prevention, our work documents the importance of local macrophage proliferation for plaque progression and lipid therapy-based regression. While oral atorvastatin inhibits macrophage proliferation indirectly via systemic cholesterol and ApoB-lipoprotein reductions, direct targeting of macrophage proliferation may emerge as a potent add-on therapy to support plaque regression. Anti-proliferative drugs such as paclitaxel and methotrexate or mTOR-inhibitor rapamycin, incorporated into a lipid nanoparticle LDE resembling low-density lipoproteins or biomimetic nanoparticles, induced plaque regression and loss of plaque macrophages in murine and rabbit models of atherosclerosis with limited systemic toxicity [
4,
7,
11]. More recently, LDE-paclitaxel was injected into eight patients with aortic atherosclerosis six times every 3 weeks, and computer tomography images of the atherosclerotic aortas were compared before and after 1–2 months of treatment with those obtained in untreated patients. Remarkably, half of the treated patients showed a reduction in plaque size without significant changes to their blood cell counts, while at the same time all the nine untreated patients showed mild disease progression [
56]. A number of clinical trials on treating inflammation in atherosclerosis have recently been published or are ongoing. While colchicine and canakinumab appear to protect from atherosclerotic complications post myocardial infarction [
37,
38,
47,
48,
61], low-dose methotrexate failed to do so in the latest CIRT trial [
46]. Of note, at low doses and with folic acid supplementation, methotrexate does not interfere with DNA synthesis, but it may suppress inflammation via the release of adenosine [
8], although no signs of modulating inflammation were seen in the CIRT trial. Whether interleukin-1β blockade and colchicine treatment influence plaque macrophage proliferation remains to be determined. We are finally entering an era where our conceptual understanding of the role of inflammation in atherosclerosis derived from numerous preclinical studies translates into clinics. Our study fits into this picture by identifying macrophage proliferation as a relevant and modifiable determinant of inflammatory cell accumulation in atherosclerotic lesions to be targeted therapeutically in support of plaque regression.