Human umbilical vein endothelial cells (HUVEC) were purchased from Promocell, grown, and expanded in endothelial growth media-2 (EGM-2, Promocell, Germany). Primary human saphenous vein pericytes (SVP) were kindly donated from Prof Madeddu (University of Bristol) and cultured in EGM-2. SVP were grown and expanded in flasks or 35 mm dishes pre-coated with 0.1% fibronectin (Merck, United States) and 0.05% gelatin (Merck, United States) solution. Cells were maintained at 37°C with 5% CO₂, passaged at 80% confluency and used up to passage 8 for experiments.

For gene expression experiments, cells were grown overnight in 6-well plates until confluency (90–100%) was reached. Confluent plates were used for all experiments to exclude the cell cycle as a potential cause of any patterns in gene expression that might be observed over the time course. Confluent monolayers were exposed to growth media supplemented with 50% FBS for 2 h, at 37°C with 5% CO₂, to synchronize the circadian clock between cells. The start of the 2 h serum shock was considered time ZT0 (Zeitgeber time 0), after which total RNA was collected every 4 h for 36 h (N = 3, n = 9). The results from the first 8 h were excluded from the analysis to avoid the immediate effect of the treatment with high serum concentration on the cells. RNA was isolated using Quick-RNA Miniprep (Zymo Research, United States). The cell lysis buffer was added directly to the dish at the designated time point and RNA was extracted following the manufacturer’s protocol. All samples were quantified using NanoDrop to determine RNA concentration and purity. Single-strand cDNA was synthesized using QuantiTect Reverse Transcription Kit (Qiagen, United Kingdom). Real-time quantitative PCR was conducted using the PowerUp SYBR Green Master Mix (Applied Biosystems, United States). Samples were analyzed on QuantStudio7 Flex Real-Time PCR System (Applied Biosystems, United States). The 2^−ΔΔCt method was used to calculate the relative fold gene expression for each sample. Briefly, the triplicate Ct values of the target genes and the β-actin (housekeeping gene) were averaged, the first Δ was calculated by subtracting the average Ct value of the target gene to the average Ct value of β-actin; the second Δ was calculated by subtracting the ΔCt of samples collected at time points to the ΔCt of the ZT0 sample. Finally, the fold gene expression was calculated by doing the 2 to the power of negative ∆∆Ct.

All primers listed in Table 1 were designed using human sequences on the Primer Design tool (NCBI-NIH) and tested for the correct amplicon length in the SVP cells by PCR using GoTaq Green Mastermix (Promega, United Kingdom).

For protein quantification, cells were grown on 13 mm diameter glass coverslips. Confluent cells were exposed to 50% FBS growth media for 2 h, at 37°C with 5% CO₂, to synchronize their circadian clock. This was considered time ZT0, after which cells were collected every 4 h and immediately fixed in 4% paraformaldehyde (PFA, ThermoFisher Scientific, United States) for 20 min at room temperature. Immunofluorescence staining was performed using PBS with 4% FBS as blocking agent and antibodies anti-ARNTL produced in rabbit (1:200, Merck, United States) or anti-NR1D1 produced in rabbit (1:200 μg/μl, Merck, United States), followed by AlexaFluor goat anti-rabbit 488 (1:200, ThermoFisher Scientific, United States) secondary antibodies. Phalloidin-iFluor 594 Reagent (Abcam,United Kingdom) was used to stain the actin filaments and DAPI (4′,6-diamidino-2-phenylindole, 1:1,000, Sigma) was used for nuclear staining. The stained slides were imaged using the 20x objective on the Nikon Eclipse. Images were processed using the NIS Elements software (Nikon) and analyzed using the software ImageJ. For each image, a two-step method was developed. First, individual nuclei were determined based on a user-defined threshold. These nuclear signals were converted in ROI (region of interest) sets containing all nuclei and Integrated Density (IntDen) of FITC signal was quantified within individual nuclei. For each time point, a technical duplicate was performed for each of the three biological replicates (N = 3, n = 6). Data were plotted, and a cosinor curve fitting was performed using the software GraphPad Prism v.8.0.

Promoter sequences of three clock genes (Bmal1, Per2 and Rev-erbα; Table 2) were identified from human genome database and purified by PCR from the DNA of a human donor. Plasmids were cloned using the Gateway Cloning technology (ThermoFisher, United States) and using the final plasmid backbone pLNT-GW-JDG, kindly donated by the Prof Waddington’s group, to produce bioluminescent reporters (Buckley et al., 2015). For the lentivirus production, HEK-293FT producer cells (Gibco, United States) were transfected with 3 plasmids: plasmid of interest (carrying Bmal1, Per2 or Rev-erbα promoter), pMD2. G VSVg env (Addgene, United States) and pCMVR8,74 packaging (Addgene, United States). Lentiviral particles were collected from the cell culture supernatants, concentrated by ultracentrifugation, and immediately stored at −80°C. Lentiviral particles titration was performed using p24 ELISA kit (Takara Bio, United States). SVP and HUVEC were transduced with 10 multiplicity of infection (MOI) and 5 MOI of lentiviral vectors, respectively. Fresh media was added after 24 h to replace the lentivirus-containing growth media.

For bioluminescence recordings, cells that were transduced with the bioluminescence reporters were grown overnight into 35 mm dishes until confluency (90–100%) was reached. Confluent cells were exposed to 50% FBS growth media for 2 h, at 37°C with 5% CO₂, to synchronize the circadian clock. Cells were washed twice with Dulbecco’s phosphate buffered saline (D-PBS) and the recording media was added to the dish. The recording media contained phenol-red free EGM-2 (Promocell, Germany) with 0.1 mM endotoxin-free Luciferin (Promega,United Kingdom) and 20 mM HEPES (pH 7.4). Culture dishes were sealed with high-vacuum grease and 40 mm glass coverslips (VWR, United Kingdom). Bioluminescence signals were recorded in the LumiCycle device (Actimetrics, United States) and baseline subtraction on running average measurements was performed by using the Lumicycle Analyzer software. A damped sin wave (non-linear regression) with a 24 h wavelength was superimposed on bioluminescence reports, using GraphPad Prism 8.0.

Lentiviral particles containing shRNA for BMAL1 knockdown were purchased from Merck (TRCN0000331078, Sequence: CCG​GGC​AAC​AGC​TAT​AGT​ATC​AAA​GCT​CGA​GCT​TTG​A-TAC​TAT​AGC​TGT​TGC​TTT​TTG). MISSION^® pLKO.1-puro Non-Mammalian shRNA Control Plasmid DNA was purchased from Merck (SHC002) and used to produce lentiviral particles from HEK-293FT producer cells, as described above. HUVEC and SVP were transduced with 10 MOI of lentivirus and stably transduced clones were selected using 4 μg/ml Puromycin dihydrochloride (Merck, United States). Viability/apoptosis were measured with the ApoTox Glo assay (Promega, United Kingdom), following manufacturer’s protocol. Briefly, 2 × 10⁴ cells were plated in a black clear-bottom 96well plate and viability/apoptosis was assessed 24, 48, 72 and 96 h after plating (for each time point N = 1; n = 4).

1.5 × 10⁴ HUVEC and 0.5 × 10⁴ shNEG/BMAL1 SVP were cultured separately or co-cultured in a ratio 3:1 in wells of a 96-well plate and allowed to attach overnight. The following day, co-cultures were synchronized with a single pulse of 100 nM dexamethasone, as previously described, and an aliquot of 5 µl of culture supernatant was collected every 4 h for downstream analysis. Lactate release and glucose intake were measured using Lactate-Glo (Promega, United Kingdom) and Glucose-Glo (Promega, United Kingdom) assays respectively, following manufacturer’s protocols. Supernatants were diluted 1:50 for lactate and 1:200 for glucose analysis.

For the Matrigel assay shNEG/shBMAL1 HUVEC and shNEG/shBMAL1 SVP cells were detached and counted. For the single culture of shNEG/shBMAL1 HUVEC, a total number of 3 × 10⁴ cells was used, while for the co-cultures, the ratio between HUVEC:SVP was 3:1 to obtain 3 × 10⁴ total number of cells. 30 µl of growth factor-reduced Matrigel (Corning, United States) was dispensed in each well of a 96-well plate and allowed to solidify. Three bright-field images were taken from each well after 8 h. Three biological replicates, and 3 technical replicates were performed for each experiment (N = 3, n = 9). Image analysis was performed using ImageJ software and applying the macro “Angiogenesis analyzer” (Carpentier et al., 2020) with additional integrated features from OCTAVA toolbox (Untracht et al., 2021).

Polyurethane (PU) scaffold was fabricated as previously described (Velliou et al., 2015; Totti et al., 2018; Gupta et al., 2019; Gupta et al., 2020) by thermal induced phase separation (TIPS) method. The generated scaffolds are highly porous, with interconnected pores of an average size of 120–140 μm. The scaffolds were cut in blocks of dimensions 2.5 mm × 2.5 mm x 5 mm. The scaffolds were sterilized with 70% ethanol overnight and exposure to dim UV-light for 40 min. Post-sterilization, the scaffolds were surface modified with fibronectin/gelatin via physisorption. Briefly, the scaffolds were immersed in 0.1% fibronectin and 0.05% gelatin solution, centrifuged for 20 min and placed in wells of a 48-well plates on top of a polydimethylsiloxane (PDMS) disks. PDMS is a hydrophobic material chosen for reducing the wash-out effect after cell seeding.

5 × 10⁵ HUVEC, 2.5 × 10⁵ SVP or a total of 5 × 10⁵ cells (HUVEC:SVP, 3:1) were suspended in 30 µl of EGM-2 and seeded on each scaffold. The scaffolds were pinned to the PDMS disks using entomology pins, and fresh media was added in each well of a 48-well plate. Scaffolds were incubated at 37°C with 5% CO₂. For the synchronization experiments, a single pulse of 100 nM dexamethasone was added to the 48-well plates for 2 h, to allow a proper absorption of the synchronizing agent into the scaffold thickness. The scaffolds were washed with PBS and incubated in normal growth media for 2,7,10 or 14 days (N = 3 per each condition). At each time point, the scaffolds were immersed in 4% PFA for 30 min and then washed 3 times with PBS.

The scaffolds were embedded in OCT compound (Agar Scientific, United Kingdom), snap-frozen in cold iso-pentane and kept at −80°C. Cryosections were obtained using the cryostat Zeiss Hyrax (20 μm thickness) on SuperFrost adhesion slides (FisherScientific, United States). Immunofluorescence staining was performed using 20% goat serum (Merck, United States) to block non-specific binding sites. Antibodies anti-CD31 produced in mouse (0.01 μg/μl, ThermoFisher, United States) and anti-NG2 produced in rabbit (0.01 μg/μl, EMD Millipore, United Kingdom) followed by AlexaFluor secondary antibodies produced in goat (1:200, ThermoFisher Scientific, United States) were used to stain endothelial cells and pericytes respectively. DAPI was used to counterstain the nuclei.

All graphs were plotted using GraphPad Prism 8.0 Software. Values are reported as mean ± SEM. For the analysis of circadian rhythms, Cosinor fit was performed on GraphPad Prism 8.0, using equation Y=(Mesor + Amplitude*cos (((2*pi)*(X-Acrophase))/Period)) where the initial values to be fit were calculated as follows: Mesor = average of all values; Amplitude = variance of all values; Acrophase = Ymax value. The p value of the fit was obtained by comparing the non-linear regression fit to a straight line. For all the other statistical analysis, a 2-way ANOVA with Greiner-Greenhouse correction and Sidak multiple comparison test was performed.

We exposed human primary pericytes to a 50% serum concentration for 2 h (serum shock) and measured the expression of circadian molecules using real-time PCR, immunofluorescence, and reporter cell lines.

Results revealed that the mRNA expression of Per2, Bmal1 and Rev-erbα present rhythmicity in human pericytes (Figures 1A–C; N = 3; mean ± SEM and 95% confidence interval shown) peaking approximately at 8, 12 and 18 h after the serum shock, respectively. As expected, Bmal1 and Per2 expression patterns were in antiphase, in line with their reciprocal roles in the circadian regulatory feedback loop (Figure 1D). Cosinor fit analysis was performed, and statistical analysis indicated that all the genes analyzed presented a rhythmic profile, with a period close to the expected 24 h (Per2 p ≤ 0.005, Period = 24.29 h; Bmal1 p ≤ 0.01, Period = 22.95 h; Rev-erbα p < 0.0001, Period = 21.82 h).

To confirm that the gene expression results translated into protein expression, we performed immunofluorescence staining at the same time points, and measured the nuclear intensity of each protein. Results showed that BMAL1 and REV-ERBα accumulation in the nuclei was rhythmic after serum shock (Figures 1E,F). Cosinor fit analysis and statistical analysis was performed to identify the rhythmic profile (N = 3; BMAL1 p ≤ 0.05; REV-ERBα p < 0.0001). The protein quantification data shows a delay of 8 h in relation to the RNA rhythms, in line with the delay expected for protein translation. In addition, we confirmed our findings with bioluminescence assays. We generated lentiviruses carrying the human promoters of Per2, Bmal1, and Rev-erbα upstream the luciferase reporter gene, and transduced primary pericytes. Upon synchronization, the promoter bioluminescence activity was assessed by Lumicycle with a 10 min interval between measurements. Recordings showed that human primary pericytes’ promoter activation of Per2, Bmal1 and Rev-erbα was rhythmic up to 4 days of culture (Figures 1G–I; N = 2; damped sin wave with fixed 24 h period fitted).

Taken together, these findings indicate that human vascular pericytes display a robust, rhythmic expression of clock genes measured at both gene and protein level, with a synchronicity that is in line with their known functionality.

Endothelial cells are key players in vascular homeostasis and many in vivo studies have identified a rhythmicity in their function (Anea et al., 2009; Wehrens et al., 2012; Shang et al., 2016; Crnko et al., 2018). However, the precise signals driving the synchronization of endothelial cell function are not completely understood, and in particular the role of the closely associated pericytes has not been evaluated.

Here, we assessed the clock gene expression in primary human endothelial cells using the same tools applied to pericytes. In our experimental set-up, we were unable to confirm the presence of a circadian clock in HUVECs. Gene expression analysis revealed a non-rhythmic expression of Bmal1 and Rev-erbα (Figures 2A,B; N = 3), as shown by the preferred fitted model being a straight line rather than a cosinor (p > 0.05). These results were confirmed in the bioluminescence reporter measures of promoter activity where none of the promoters showed rhythmicity (Figures 2C,D; N = 2, n = 6) We also tried synchronization with dexamethasone rather than serum shock, but this yielded comparable results (Supplementary Figure S1).

Next, we co-cultured endothelial cells (HUVEC) with pericytes (SVP) by seeding Bmal1:Luc transduced endothelial cells on top of a synchronized monolayer of pericytes. In this contact co-culture approach, Bmal1 promoter activity in the endothelial cells showed a 24 h oscillation up to 3 days of culture (Figure 2E; N = 3). To assess whether the synchronizing effect was due to contact-dependent factors or secreted molecules exchanged between the two cell types, we repeated the experiment in a non-contact co-culture. Bmal1:Luc transduced endothelial cells were seeded on transwell insert, placed on top of a monolayer of synchronized pericytes. We observed that the lack of contact prevented endothelial cells’ clock synchronization, suggesting that the synchronization is at least in part dependent upon contact-mediated mechanisms (Figure 2F; N = 3).

These results suggest that, in our experimental conditions, endothelial cells synchronization requires stimuli imparted by pericytes, co-cultured in close contact.

Endothelial cells are the main drivers of blood vessel sprouting (angiogenesis), however their function is finely regulated by perivascular cells, such as pericytes (Armulik et al., 2005; Campagnolo et al., 2016; Caporali et al., 2017; Eilken et al., 2017).

Here, we aimed at assessing the effect of the disruption of the vascular clock in endothelial cells and pericytes. We produced BMAL1 knockdown cells by transducing pericytes and endothelial cells with shBMAL1, followed by puromycin clonal selection and obtained 67 and 81% reduction of BMAL1 mRNA expression, respectively (Figures 3A,C). We then verified whether the absence of BMAL1 affected viability and apoptosis. In SVP, BMAL1 knockdown impaired viability compared to controls after 48 h (0.77 ± 0.05; p < 0.05) and 72 h (0.68 ± 0.04; p < 0.01), linked with a significant increase in apoptosis at all time points (Figure 3B; 24 h: 1.61 ± 0.10, p < 0.05; 48 h: 1.24 ± 0.05, p < 0.05; 72 h: 1.72 ± 0.05, p < 0.001; 96 h: 2.14 ± 0.15, p < 0.05). In HUVEC, BMAL1 knockdown did not affect the overall viability of the cells over time but increased apoptosis at all time points (Figure 3D; 24 h: 2.09 ± 0.12, p < 0.001; 48 h: 1.62 ± 0.8, p < 0.01; 72 h: 1.67 ± 0.06, p < 0.001; 96 h: 1.53 ± 0.08, p < 0.01).

Next, we assessed the effect of the loss of BMAL1 on angiogenesis by co-culturing combinations of HUVEC ± shBMAL1 with or without SVP ± shBMAL1, to form tube-like structures in a Matrigel assay. Results showed that BMAL1 knockdown in endothelial cells cultured alone or with pericytes did not affect tube formation, however, when both HUVEC and SVP’s clocks were disrupted at the same time, a reduction in the thickness of the branches and the total coverage was observed (Figures 3E,F). Interestingly, the total branching length was not affected in any of the tested conditions (Figure 3G) suggesting that a disrupted clock may only affect the maturation and morphology of the vascular-like structures in vitro, rather than the overall angiogenic capacity.

In summary, these results indicated that pericytes are disproportionately affected by circadian disruption and that only the concurrent disruption of clock in both endothelial cells and pericytes determines defective angiogenesis, by reducing vessel maturation.

The metabolic by-product lactate is a known intracellular signaling molecule influencing angiogenesis, and is rapidly shuttled between cells, favoring endothelial progenitor cells recruitment, endothelial cells migration and the release of vascular endothelial growth factor (VEGF) and transforming growth factor-beta (TGFβ) (Brooks, 2009; Baltazar et al., 2020; Brooks, 2020). Lactate has been implicated in the regulation of the circadian clock (Rutter et al., 2001; Krishnaiah et al., 2017) and is readily up-taken and utilized by endothelial cells (Harjes et al., 2012).

We quantified the amount of lactate released in HUVEC/SVP co-cultures after synchronization and detected a steady accumulation of lactate over time in the supernatant (Figure 4A). When BMAL1 was knocked down only in SVP, we observed a decreasing slope in lactate release in the co-culture (Figure 4A; HUVEC + SVP shNEG linear regression slope = 124.7 ± 4.9; HUVEC + SVP shBMAL1 linear regression slope = 82.63 ± 2.6), associated with a lower concentration of lactate in the supernatant at all time points (mean difference 163 ± 81 µM). This is remarkable, especially considering that endothelial cells are the major producers of lactate within the co-culture (Supplementary Figure S2).

Of note, single culture of pericytes produced and released lactate, even if in lower concentrations as compared to endothelial cells. Disruption of the circadian clock in SVP did not dramatically impact the overall lactate accumulation overtime (Figure 4B); however shNEG pericytes produced lactate with an oscillatory pattern with a period of approximately 8 h (Figure 4C), while shBMAL1 pericytes did not display an oscillatory pattern (Figure 4D). These changes were specific to the lactate, as glucose remained constant throughout (Figures 4C,D).

In conclusion, we demonstrated that clock disruption in pericytes affects the rhythmicity of lactate production and influences the overall accumulation of lactate in co-cultures. This may constitute a mediator influencing endothelial cells’ clock synchronization and angiogenesis.

Vascularization of tissue engineering constructs is critical to the development of complex 3D structures, however the role of the circadian clock in this process has never been investigated.

HUVEC and SVP were seeded separately in a polyurethane scaffold coated with fibronectin and gelatin, which provided a structured three-dimensional microenvironment with extracellular matrix (ECM) mimicry, offering a biomimetic biochemical composition for the cells to interact with the ECM and each other (Totti et al., 2018). We observed that both HUVEC (CD31⁺ cells) and SVP (NG2⁺ cells) were able to penetrate, adhere and grow inside the scaffold (Figures 5A,B).

We then seeded both cell types together in co-culture and observed marked changes in the morphology of cell distribution within the scaffold. In particular, vascular cells became more organized and formed vessel-like structures by lining the scaffold pores (Figure 5C).

To assess whether synchronization of the circadian clocks improved cell migration within the scaffold, we synchronized the cells within the scaffolds with a single pulse of dexamethasone and thereafter cultured the scaffolds for 2, 7, 10 and 14 days post-synchronization. We calculated the relative frequency distribution of cells relative to the distance from the edge of the scaffold. In both synchronized (synch) and non-synchronized (non-synch) conditions we observed a predominant presence of cells in the range 0–50 pixels from the edge, at every time point (Figures 5D,E, N = 3).

On the other hand, we noticed an increase in vessel-like structures associated with a better/more systematic organization of cells around the pores in the synchronized clock condition, at every time point of culture (Figure 6A). We quantified the average number of endothelial cells and pericytes in each pore and identified a significant increase of endothelial cells that associated with the scaffold pores at day 7, and of pericytes at day 14 and an overall increase in the number of cells per pore at day 7 (Figures 6B–D). Additionally, we calculated the pore coverage, as the percentage of the pore surface covered with cells, identifying a trend of increased coverage throughout the culture in synchronized samples, compared with non-synchronized samples (Figure 6E). Specifically, the mean percentage of pore coverage in the synchronized condition was always higher (53.3% ± 8.4) when compared with the non-synchronized condition (36.2% ± 10.4), with a significant difference at 10 days of culture (p < 0.05).

Taken together these results indicate that circadian clock synchronization affects the organization of vascular cells in complex vascular structures within a 3D scaffold.