TNF-α promotes cerebral pericyte remodeling in vitro, via a switch from α1 to α2 integrins
© Tigges et al.; licensee BioMed Central Ltd. 2013
Received: 5 December 2012
Accepted: 15 February 2013
Published: 1 March 2013
Skip to main content
© Tigges et al.; licensee BioMed Central Ltd. 2013
Received: 5 December 2012
Accepted: 15 February 2013
Published: 1 March 2013
There is increasing evidence to suggest that pericytes play a crucial role in regulating the remodeling state of blood vessels. As cerebral pericytes are embedded within the extracellular matrix (ECM) of the vascular basal lamina, it is important to understand how individual ECM components influence pericyte remodeling behavior, and how cytokines regulate these events.
The influence of different vascular ECM substrates on cerebral pericyte behavior was examined in assays of cell adhesion, migration, and proliferation. Pericyte expression of integrin receptors was examined by flow cytometry. The influence of cytokines on pericyte functions and integrin expression was also examined, and the role of specific integrins in mediating these effects was defined by function-blocking antibodies. Expression of pericyte integrins within remodeling cerebral blood vessels was analyzed using dual immunofluorescence (IF) of brain sections derived from the animal model of multiple sclerosis, experimental autoimmune encephalomyelitis (EAE).
Fibronectin and collagen I promoted pericyte proliferation and migration, but heparan sulfate proteoglycan (HSPG) had an inhibitory influence on pericyte behavior. Flow cytometry showed that cerebral pericytes express high levels of α5 integrin, and lower levels of α1, α2, and α6 integrins. The pro-inflammatory cytokine tumor necrosis factor (TNF)-α strongly promoted pericyte proliferation and migration, and concomitantly induced a switch in pericyte integrins, from α1 to α2 integrin, the opposite to the switch seen when pericytes differentiated. Inhibition studies showed that α2 integrin mediates pericyte adhesion to collagens, and significantly, function blockade of α2 integrin abrogated the pro-modeling influence of TNF-α. Dual-IF on brain tissue with the pericyte marker NG2 showed that while α1 integrin was expressed by pericytes in both stable and remodeling vessels, pericyte expression of α2 integrin was strongly induced in remodeling vessels in EAE brain.
Our results suggest a model in which ECM constituents exert an important influence on pericyte remodeling status. In this model, HSPG restricts pericyte remodeling in stable vessels, but during inflammation, TNF-α triggers a switch in pericyte integrins from α1 to α2, thereby stimulating pericyte proliferation and migration on collagen. These results thus define a fundamental molecular mechanism in which TNF-α stimulates pericyte remodeling in an α2 integrin-dependent manner.
Pericytes are vascular mural cells that lie in close proximity to endothelial cells of capillaries, arterioles, and venules [1, 2]. Pericytes are regularly positioned along cerebral microvessels, and ultrastructural studies have shown that they are located within the abluminal vascular basal lamina that surrounds vessels . Pericytes are crucial regulators of vascular development, stability, and remodeling , and increasing evidence suggests that they also regulate capillary blood flow [5, 6]. One area currently attracting great interest is the role of pericytes in vascular remodeling. The current view is that pericytes act as central regulators of angiogenesis, through their ability to stabilize or destabilize microvessels [7, 8]. According to this view, pericytes promote vessel stability by maintaining close adhesive contacts with both endothelial cells and the underlying ECM, thus locking the vascular components into place. At an early stage of the angiogenic program, pericytes undock from endothelial cells, migrating within the ECM-rich basal lamina . This leads to endothelial cells breaking connections, both with each other and with the underlying basal lamina, in order to migrate and proliferate, and to sprout new blood vessels. Upon completion of endothelial remodeling, pericytes migrate back to regain contact with endothelial cells, thereby stabilizing newly formed vessels. This important role for pericytes in vascular remodeling is best illustrated by the finding that mutant mice lacking platelet-derived growth factor beta (PDGF-B) or the PDGF-β receptor fail to show efficient pericyte coverage of blood vessels, resulting in perinatal lethality due to leaky dysfunctional blood vessels [10, 11].
The basal lamina of cerebral blood vessels comprises a number of different ECM proteins and proteoglycans, the precise make-up of which varies with the vessel-maturation state. The basal lamina of mature vessels is comprised of three major constituents: collagen IV, laminins, and heparan sulfate proteoglycan (HSPG) [12–14]. In addition, immature vessels of the developing CNS and those undergoing remodeling in the adult CNS also contain increased levels of fibronectin and vitronectin, which are downregulated upon vessel maturation [15, 16]. Broadly speaking, the ECM influences many aspects of cell behavior, including cell proliferation, migration, differentiation, and stabilization [17, 18]. These effects are mediated by the ECM receptors, integrins, which are expressed at the cell surface as αβ heterodimers, of which the β1 class is the major type [19, 20]. In a number of studies, we have highlighted a role for the remodeling protein fibronectin in driving cerebral angiogenesis after cerebral hypoxia or ischemia. In these models, angiogenic cerebral vessels show strong upregulation of fibronectin and the fibronectin receptor, α5β1 integrin [21, 22]. Furthermore, using endothelial-specific deletion of the α5 integrin, we showed previously that α5β1 integrin plays an important role in promoting endothelial cell proliferation at an early stage of the angiogenic response .
As pericytes lie within the basal lamina ECM of cerebral microvessels , it seems likely that pericytes also respond to environmental cues provided by the ECM. Because little is currently known about the influence of the ECM on pericyte behavior, the aim of this study was to address the following questions: 1) how is cerebral pericyte adhesion, proliferation, and migration influenced by the different ECM constituents present in the vascular basal lamina; 2) which integrins do pericytes express; 3) how do cytokines regulate pericyte remodeling state and expression of integrins; and 4) are any of the identified integrins required for pericyte remodeling?
The studies described were reviewed and approved by the Scripps Research Institute (TSRI) Institutional Animal Care and Use Committee. All cell cultures were obtained from C57Bl/6 mice, which were maintained under pathogen-free conditions in the closed breeding colony of TSRI.
Experimental autoimmune encephalomyelitis (EAE) was induced using a commercial protocol and materials (Hooke Laboratories, Lawrence, MA, USA). Briefly, C57Bl/6 female mice, 8 to 10 weeks old, were immunized with 100 μl of 1 mg/ml MOG33-35 peptide emulsified in complete Freund’s adjuvant (CFA) containing 2 mg/ml Mycobacterium tuberculosis by subcutaneous injection in both the base of the tail and upper back. In addition, on days 0 and 1, mice also received an intraperitoneal injection of 200 ng pertussis toxin. Control mice received CFA not containing the MOG peptide. This protocol leads to robust induction of clinical EAE on days 12 to 14 after immunization. Animals were monitored daily for clinical signs and scored as follows: 0, no symptoms; 1, flaccid tail; 2, paresis of hind limb; 3, paralysis of hind limbs; 4, quadriplegia; 5, death. At 21 days post-immunization, corresponding to the acute symptomatic stage of disease, mice were euthanized by intraperitoneal injection of sodium pentothal.
Pure cultures of mouse brain endothelial cells (BECs) or pericytes were prepared as previously described [24, 25]. Briefly, brains were removed from 8 week-old C57Bl/6 mice, minced, dissociated for 1 hour in papain and DNase I, centrifuged through 22% BSA to remove myelin, and endothelial cells cultured in endothelial cell growth media (ECGM), consisting of Hams F12 supplemented with 10% FBS, heparin, ascorbic acid, L-glutamine, penicillin/streptomycin (all from Sigma Chemical Co., St. Louis, MO, USA) and endothelial cell growth supplement (ECGS) (Upstate Cell Signaling Solutions, Lake Placid, NY, USA), on six-well plates coated with type I collagen (Sigma Chemical Co.). To obtain BECs, puromycin (4 μg/ml; Alexis GmbH, Grunberg, Germany) was included in the culture media on days 1 to 3 to remove contaminating cell types. Endothelial cell purity was >99% as determined by flow cytometry with CD31. For all experiments, BECs were used only for the first passage.
Pericytes were obtained using the same approach, except that the puromycin step was omitted. The pericyte cultures were grown in ECGM, with the medium changed every 3 days. On reaching confluency, cultures were harvested with trypsin and passaged. During the first two passages, pericyte cultures were grown in ECGM, but on the third passage, they were switched to pericyte medium (PCM; ScienCell Research Laboratories, Carlsbad, CA, USA) containing 2% FBS. In previous studies we found that, using this approach, cultures of pericytes become highly purified after the third passage, at which point these cultures are more than 99% pericytes as determined by expression of the pericyte marker NG2 and the PDGF-β receptor, and contain less than 1% of contaminating endothelial cells (CD31), astrocytes (glial fibrillary acidic protein; GFAP), or microglia (Mac-1), as determined by fluorescent immunocytochemistry . All studies were performed on pericytes at passages 4 to 8. Pericytes were expanded in PCM containing 2% FBS, but all functional assays were performed in serum-free DMEM containing N1 supplement, L-glutamine, and penicillin/streptomycin (all from Sigma Chemical Co.).
To investigate the influence of cytokines on pericyte behavior and expression of integrin subunits, pericytes were cultured on collagen I in the presence of 20 ng/ml basic fibroblast growth factor (bFGF; Invitrogen Corp., Carlsbad, CA, USA), 20 ng/ml platelet derived growth factor (PDGF-B) 2 ng/ml transforming growth factor (TGF)-β1 10 ng/ml tumor necrosis factor (TNF)-α, or 10 ng/ml vascular endothelial growth factor (VEGF) (all R&D Systems, Minneapolis, MN, USA). These concentrations were selected based on the findings of previous studies [26, 27]. The following monoclonal antibodies (BD Pharmingen, La Jolla, CA, USA) were used: monoclonal antibodies reactive for the integrin subunits α1 (clone Ha31/8), α2 (clone Ha1/29), α4 (clone MFR4.B), α5 (clone 5H10-27 (MFR5), α6 (clone GoH3), β1 (clone Ha2/5), and Mac-1 (clone M1/170); CD31 (clone MEC13.3); and isotype control antibodies: rat anti-KLH (A110-2) and hamster anti-TNP-KLH (G235-1). Other antibodies used in this study included Cy3-conjugated anti-GFAP (Sigma Chemical Co.) and rabbit anti-NG2 and anti-PDGF-β receptor antibodies (both kindly provided by Dr William Stallcup, Sanford-Burnham Medical Research Institute, La Jolla, CA, USA).
Integrin expression by BECs and pericytes (treated with different cytokines for 2 days) was examined as described previously . Briefly, cells were removed from the six-well culture plates, and cell-surface expression of the integrin subunits α1, α2, α4, α5, α6, or β1 was analyzed by flow cytometry using phycoerythrin (PE)-conjugated monoclonal antibodies (all BD Pharmingen). The fluorescent intensity of the labeled cells was analyzed with a flow cytometer (FACScan; Becton Dickinson, San Diego, CA, USA), with 10,000 events captured for each condition. In each experimental condition, the mean fluorescent intensity was compared with the control (no factor) condition, and expressed as the percentage change relative to control. Each experiment was repeated a minimum of four times.
Adhesion assays were performed as described previously . Briefly, substrates were prepared by coating the central area of glass coverslips in 24-well plates (Nunc; BD Biosciences, San Jose, CA, USA) with 25 μl of ECM solution (10 μg/ml of collagen I, collagen IV, fibronectin, HSPG, or laminin-1; all from Sigma Chemical Co.) for 2 hours at 37°C. Substrates were washed twice before addition of cells. Pericytes were prepared as described above, centrifuged, and re-suspended in N1 serum-free media, then 2,000 cells were applied to the substrates in a 25 μl drop and incubated at 37°C for 1, 4, or 8 hours. In function-blocking experiments, antibodies were included at 5 μg/ml. The assay was stopped by adding 1 ml of DMEM and washing off any loosely attached cells. Attached cells were fixed in 4% paraformaldehyde in PBS for 20 minutes, and stored in PBS. Adhesion was quantified under phase microscopy by counting all attached cells within five fields of view per condition. In each experiment, each condition/time-point was performed in duplicate.
Glass coverslips were coated with the ECM substrates as described above, and pericytes plated out in serum-free N1 medium. Cytokines and/or integrin-blocking antibodies were included at the time of plating. The following morning, pericytes were incubated for 3 hours with bromodeoxyuridine (BrdU; Invitrogen Corp.), fixed in acid/alcohol, and analyzed by immunofluorescence for BrdU incorporation, in accordance with the manufacturer’s instructions. BrdU-positive cells were expressed as the percentage of total cells (Hoechst staining).
Pericyte migration was quantified using the scratch assay. Pericytes were plated into ECM-coated wells of a 24-well plate, and cultured in PCM. Upon reaching confluence, vertical and horizontal scratches were made in the monolayer using a 1 ml pipette tip. The PCM and cell debris were removed, and replaced with serum-free N1 medium (cytokines and/or integrin-blocking antibodies were included at this time), and the scratch width was recorded for all samples. The new width of the scratch was recorded 16 hours later, and the distance of cell migration calculated. In each experiment, each condition was performed in duplicate.
Immunohistochemistry was performed as described previously , on 10 μm frozen sections of cold PBS-perfused brain, using monoclonal antibodies specific for the integrin subunits α1 (clone Ha31/8) and α2 (clone Ha1/29), and the pericyte marker NG2. Secondary antibodies used included anti-hamster Alexa Fluor 488 (Invitrogen Corp.) and anti-rabbit Cy3 (Jackson Immunoresearch, Baltimore, PA, USA). Images of brain sections were taken using a ×20 objective on a microscope (Imager M1.m; Zeiss, Thornwood, NY, USA). Three images were taken, and the number of integrin-positive vessels per field of view recorded for each section per subject.
All results represent the mean ± SEM of four experiments, except for immunohistochemistry, which was performed with three different animals per condition. The Student’s t-test was used to analyze the results of the proliferation assays and the immunohistochemistry, while the paired Student’s t-test was used to analyze the results for flow cytometry, and the cell-adhesion and migration assays. For all tests, P<0.05 was defined.
To investigate how pericyte behavior is influenced by the different ECM molecules present in the vascular basal lamina, pure populations of cerebral pericytes were cultured in serum-free N1 medium on glass coverslips coated with collagen I, collagen IV, fibronectin, HSPG, or laminin-1.
Next, the influence of different ECM components on pericyte proliferation and migration was investigated. Pericyte proliferation was examined by BrdU incorporation. Pericytes were cultured in serum-free conditions on the different substrates, then BrdU was added to cells for 3-hours, followed by BrdU immunofluorescence (IF) detection. Pericyte proliferation was promoted most strongly by fibronectin (Figure 1C), with a proliferation rate four-fold greater than any other substrate (21.1 ± 3.4% on fibronectin compared with 5.7 ± 0.8% on collagen I, P<0.01). Collagen IV also supported pericyte proliferation (5.7 ± 2.2%). However, compared with collagen I (5.7 ± 0.8%), both laminin-1 (2.6 ± 1.2%; P<0.05) and HSPG (1.6 ± 0.5%; P<0.01) significantly inhibited pericyte proliferation.
The influence of the ECM substrate on pericyte migration was investigated by using the scratch assay. Pericytes were first grown to confluence in serum-containing PCM and cultured on ECM-coated 24-well plates, then horizontal and vertical scratches were made to the monolayer to produce linear regions devoid of cells. The medium was then switched to serum-free N1 medium, and migration was measured over the next 16 hours. Pericyte migration was most effectively promoted by collagen I (0.28 ± 0.02 mm versus 0.04 ± 0.01 mm on uncoated plastic, P<0.001), followed by 50% lower levels on fibronectin (0.14 ± 0.02 mm, P<0.01) and laminin-1 (0.14 ± 0.03 mm, P<0.02) (Figure 1D and Additional file 1: Figure S1). Compared with uncoated plastic (baseline), cells on HSPG showed an anti-migratory trend, although this failed to reach statistical significance. Taken together, these results demonstrate that specific ECM substrates have markedly different effects on pericyte behavior. Consistent with its upregulation during cerebrovascular remodeling and its regenerative influence on other cell types [21, 22, 29], fibronectin supports pericyte remodeling by strongly promoting pericyte proliferation and migration. Both collagen I and IV also support pericyte proliferation, but the two collagens have differential effects on pericyte migration, with collagen I having a much stronger effect. Although laminin-1 is only a weak promoter of pericyte adhesion and proliferation, it does support migration. Most strikingly, HSPG appears to be non-permissive for all aspects of pericyte behavior, suggesting that, within intact blood vessels, HSPG restricts pericyte proliferation and migration, thus preventing excessive and unwanted vascular remodeling.
Because TNF-α promotes parallel increases in pericyte remodeling status and α2 integrin expression, we next examined whether α2β1 integrin mediates some of the change in pericyte behavior, by measuring pericyte migration and proliferation on collagen I under the influence of TNF-α, in the presence of integrin-blocking antibodies. TNF-α significantly promoted pericyte migration and proliferation compared with controls, and the pro-modeling effects of TNF-α were significantly blocked by antibodies against the β1 (migration reduced from 0.47 ± 0.04 mm to 0.24 ± 0.04 mm, P<0.02; proliferation reduced from 35.8 ± 4.7% to 17.6 ± 2.2%, P<0.02) or α2 (migration reduced from 0.47 ± 0.04 mm to 0.23 ± 0.01 mm, P<0.01; proliferation reduced from 35.8 ± 4.7% to 20.4 ± 3.4%, P<0.05) integrin subunits, but were not significantly affected by the anti-α1 integrin antibody (Figure 5D,E). This demonstrates that TNF-α stimulates pericyte migration and proliferation through a α2 integrin-dependent mechanism.
Pericytes have an extremely close relationship with the ECM components of the basal lamina of blood vessels [3, 7]. In this study, we took an in vitro approach to determine how different vascular ECM substrates influence pericyte adhesion, migration, and proliferation, and to define the integrin receptors that mediate these effects. We then examined the interplay between cytokines and ECM–integrin interactions in regulating pericyte behavior. Our studies showed that fibronectin and collagen I promote pericyte proliferation and migration, whereas the proteoglycan HSPG had an overall inhibitory influence on pericytes. Of the cytokines tested, TNF-α had the strongest pro-modeling influence, stimulating pericyte proliferation and migration, concomitantly triggering a marked switch in pericyte integrins, from α1 to α2 integrin, the exact opposite to that seen in differentiated pericytes. Inhibition studies showed that α2 integrin mediates pericyte adhesion to collagen I and IV, and function blockade of α2 integrin prevented the pro-modeling influence of TNF-α. To our knowledge, these are the first studies to demonstrate that ECM constituents are a major influence on pericyte remodeling. Specifically, they suggest a model in which HSPG restricts pericyte remodeling in stable vessels, but during inflammation, TNF-α triggers a switch in pericyte integrins, from α1 to α2, thereby promoting pericyte proliferation and migration on collagen. These studies thus identify a fundamental molecular mechanism that mediates pericyte transformation into an active remodeling phenotype.
Several factors that regulate pericyte behavior including PDGF-BB and VEGF, are known [10, 43], although surprisingly, the influence of ECM components has not been directly addressed. In this study, we found that fibronectin and collagen I drive cerebral pericytes towards a pro-modeling phenotype, which is in keeping with the influence of these ECM proteins on other cell types. Fibronectin is a strong promoter of endothelial cell proliferation and migration [29, 44], and is a strong promoter of vascular remodeling under different conditions including development, tumor-associated neovascularization, and hypoxia-induced cerebrovascular remodeling [21, 23, 45, 46]. Likewise, collagen I promotes angiogenic endothelial remodeling both in vitro and in vivo[47, 48]. Our finding that fibronectin and collagen I also stimulate pericyte remodeling suggests that endothelial cells and pericytes use common mechanisms to switch from a quiescent stable phenotype into an active remodeling one. In stark contrast, we found that the proteoglycan HSPG was non-permissive for all aspects of pericyte behavior, consistent with the finding that HSPG inhibits mesangial adhesion to fibronectin . Our data suggest that within stable cerebral blood vessels, HSPG might restrict pericyte proliferation and migration, thus preventing unwanted vascular remodeling. These results are consistent with the idea that the positive/negative balance of ECM cues may play an important role in determining vascular remodeling status.
Evidence suggests that TNF-α promotes vascular remodeling in vivo. Exogenous TNF-α was shown to promote angiogenic sprouting in the rat cornea and the chick chorioallantoic membrane . In a mouse model of airway inflammation, TNF-α and endothelial expression of TNF receptor 1 (TNF-R1) were increased, and inhibition of this pathway blocked remodeling . At the cellular level, TNF-α promotes endothelial cell proliferation , migration, and tube formation . In the current study, we found that TNF-α also promoted pericyte proliferation and migration, consistent with recent data that TNF-α stimulates cerebral pericyte migration and matrix metalloproteinase-9 production . Together, these observations support a fundamental role for TNF-α in mediating vascular remodeling.
A major finding to emerge from this study is that the pro-modeling influence of TNF-α correlated with a switch in pericyte expression of β1 integrins, from α1 to α2, whereas differentiating pericytes showed the opposite switch. A similar switch has been described on chondrocytes . So what might be the functional significance of this switch? Although α1 and α2 integrins show great similarity in their sequence homology , some clear functional differences between these two integrins have been reported. First, the ligand specificity of α1 and α2 integrins seems to be cell-type-specific. Glomerular epithelial cells (GECs) use α2β1 to attach to collagen, and use both α1β1 and α2β1 to attach to laminin, whereas renal mesangial cells use both α1β1 and α2β1 to attach to collagen, but use only α1β1 to adhere to laminin [53, 54]. Second, compared with α1β1, α2β1 integrin has much higher affinity for collagen I [36, 37], implying that α2 integrin expression confers on cells an increased adhesion and signaling capability on this substrate.
Several studies have highlighted an important role for α2β1 integrin in promoting cell proliferation and migration in other systems. PDGF-B-induced proliferation and migration of vascular smooth muscle cells is blocked by function-blocking anti-α2 antibodies or enhanced by the α2β1 integrin agonist aggretin [55, 56]. Furthermore, many studies have described an important role for α2β1 integrin in promoting the migration and/or metastatic spread of tumor cells, including melanoma , and carcinoma cells of the colon, prostate, liver, and mammary gland [58–61]. In another study, TNF-α conferred an invasive transformed phenotype on mammary epithelial cells that was accompanied by increased α2 integrin expression, and specific blockade of α2 integrins inhibited this transformation . This TNF-α-induced transformation bears a remarkable similarity to our own findings with pericytes, suggesting the presence of a common fundamental mechanism by which TNF-α stimulates cell migration through a α2 integrin-dependent mechanism.
An angiogenic role for α2β1 integrin in endothelial cells has been well described [63, 64]. A recent study by Stratman et al. defined an important role for pericytes in stimulating endothelial basement membrane formation and vessel maturation, but also demonstrated a requirement for α2 integrin in the early stages of tube formation, . Taken with our own findings, this is consistent with the notion that α2β1 integrin provides pro-angiogenic signals, both in endothelial cells and pericytes during the early stages of vessel remodeling. So how essential is α2 integrin for these events? Interestingly, although α2 integrin knockout (KO) mice are viable and fertile, they exhibit defective branching morphogenesis in mammary epithelial ducts . In future experiments, we plan to test whether α2 integrin plays a similar role in vessel sprouting by examining cerebrovascular remodeling in α2 integrin KO mice, both during EAE and in a mouse model of chronic mild hypoxia.
The aim of this study was to determine how ECM components present in the vascular basal lamina influence pericyte remodeling behavior, and how cytokines regulate these events. Fibronectin and collagen I promoted pericyte proliferation and migration, but the proteoglycan HSPG had an inhibitory influence on pericyte behavior. The pro-inflammatory cytokine TNF-α strongly promoted pericyte proliferation and migration, and concomitantly induced a switch in pericyte integrins, from α1 to α2 integrin, the opposite to that seen when pericytes differentiate. Inhibition studies showed that α2 integrin mediates pericyte adhesion to collagens, and that function blockade of α2 integrin inhibited the pro-modeling influence of TNF-α. Together, these results suggest a model in which ECM constituents influence pericyte remodeling status. In this model, HSPG restricts pericyte remodeling in stable vessels, but during inflammation, TNF-α triggers a switch in pericyte integrins, from α1 to α2, which stimulates pericyte proliferation and migration on collagen. These results thus define a fundamental molecular mechanism by which TNF-α stimulates pericyte remodeling in an α2 integrin-dependent manner.
This work was supported by the National Multiple Sclerosis Society; a Harry Weaver Neuroscience Scholar Award to RM (JF 2125A1/1), and a Post-Doctoral Fellowship to JVW (FG 1879-A-1), and by the NIH RO1 grant (NS060770). This is manuscript number 22037 from TRSI.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.