Scaffold-mediated lentiviral transduction for functionaltissue engineering of cartilageJonathan M. Brungera,b, Nguyen P. T. Huynha,c, Caitlin M. Guenthera,b, Pablo Perez-Pinerab, Franklin T. Moutosa,Johannah Sanchez-Adamsa, Charles A. Gersbacha,b,1,2, and Farshid Guilaka,b,c,1,2

Departments of aOrthopaedic Surgery and cCell Biology, Duke University Medical Center, Durham, NC 27710; and bDepartment of Biomedical Engineering,Duke University, Durham, NC 27708

Edited by Shu Chien, University of California, San Diego, La Jolla, CA, and approved January 15, 2014 (received for review November 20, 2013)

The ability to develop tissue constructs with matrix compositionand biomechanical properties that promote rapid tissue repair orregeneration remains an enduring challenge in musculoskeletalengineering. Current approaches require extensive cell manipula-tion ex vivo, using exogenous growth factors to drive tissue-specificdifferentiation, matrix accumulation, and mechanical properties, thuslimiting their potential clinical utility. The ability to induce and main-tain differentiation of stem cells in situ could bypass these steps andenhance the success of engineering approaches for tissue regenera-tion. The goal of this study was to generate a self-contained bioactivescaffold capable of mediating stem cell differentiation and formationof a cartilaginous extracellular matrix (ECM) using a lentivirus-basedmethod. We first showed that poly-L-lysine could immobilize lentivi-rus to poly(e-caprolactone) films and facilitate human mesenchymalstem cell (hMSC) transduction. We then demonstrated that scaffold-mediated gene delivery of transforming growth factor β3 (TGF-β3),using a 3D woven poly(e-caprolactone) scaffold, induced robust carti-laginous ECM formation by hMSCs. Chondrogenesis induced by scaf-fold-mediated gene delivery was as effective as traditional differen-tiation protocols involving medium supplementation with TGF-β3, asassessed by gene expression, biochemical, and biomechanical analy-ses. Using lentiviral vectors immobilized on a biomechanically func-tional scaffold, we have developed a system to achieve sustainedtransgene expression and ECM formation by hMSCs. This methodopens new avenues in the development of bioactive implants thatcircumvent the need for ex vivo tissue generation by enabling thelong-term goal of in situ tissue engineering.

biomaterials | regenerative medicine | genetic engineering |gene therapy | chondrocyte

The development of tissue constructs with matrix compositionand biomechanical properties suitable for tissue replacement

remains a significant goal of musculoskeletal tissue engineering.The canonical paradigm for engineering biological tissue sub-stitutes generally involves three critical components (1, 2): (i)cells capable of populating the tissue substitute and generatingappropriate neotissue constituents (3); (ii) biomimetic scaffoldsthat cells can infiltrate and that can support extracellular matrix(ECM) accumulation and organization (4); and (iii) environ-mental signals such as growth factors (5), small molecules, orphysical factors (6) that can direct cell differentiation toward theappropriate lineage of choice. In traditional tissue engineeringapproaches, these elements are generally applied ex vivo to in-duce tissue-specific lineage commitment and matrix formation.Although promising strategies for developing bone, cartilage,tendon, and muscle tissue substitutes based on this paradigmexist (7–9), these approaches require extensive in vitro manipu-lation of engineered replacements. Furthermore, upon implan-tation in vivo, the ability to further specify or modify cell fateis lost.The ability to induce stem cell differentiation in situ in the

absence of exogenous growth factors could significantly enhancetissue regeneration efforts by eliminating the need for in vitroculture of tissue replacements. Patient-derived cells could be

harvested and seeded onto scaffolds in a single surgery, or pro-genitor cells may populate the scaffold in vivo following implan-tation. This approach requires a biomaterial that can providefunctional tissue properties while neotissue is formed in responseto persistent cues that guide in situ ECM development. In thisregard, previous approaches have developed engineered constructsthat elute small molecules or locally deliver proteins capable ofinitiating cell differentiation (10) or migration (11). Furthermore,efforts to enhance cell differentiation have involved functionaliza-tion of synthetic scaffolds with ECM proteins or peptides recog-nized by cell surface receptors (12–14). More recently, techniquesfor incorporating gene vectors, such as naked plasmid DNA ornonintegrating adenoviruses and adeno-associated viruses, intotissue replacement constructs have been used (15–19). Theseefforts rely on the ability of transient signals, limited by proteinhalf-life or loss of transgene expression from nonintegrating vec-tors, to direct cell lineage commitment and drive ECM deposition.Moreover, whereas many of the substrates used in these studiesenable localized gene delivery, the mechanical properties do notrecapitulate those of native tissues. Thus, implantation of suchconstructs must be delayed until after ex vivo development ofneotissue to allow the implant to function in the context of nor-mal physiological loads.The goal of this study was to develop a bioactive scaffold ca-

pable of mediating cell differentiation and formation of an ECMsuitable for replacement of musculoskeletal tissues with mechanicalproperties that mimic those of native tissues. Lentiviral vectors were

Significance

Engineered replacements for musculoskeletal tissues generallyrequire extensive ex vivo manipulation of stem cells to achievecontrolled differentiation and phenotypic stability. The abilityto control cell differentiation using cell-instructive scaffoldsthat have biomechanical properties approximating those ofnative tissue would represent a transformative advance infunctional tissue engineering. The goal of this study was todevelop a bioactive scaffold capable of mediating cell differ-entiation and formation of an extracellular matrix with thebiochemical composition and mechanical features that mimicnative tissue properties. By combining innovative gene de-livery strategies with advanced biomaterial design, we dem-onstrate the feasibility of generating constructs capable ofrestoring biological and mechanical function.

Author contributions: J.M.B., C.A.G., and F.G. designed research; J.M.B., N.P.T.H., C.M.G.,P.P.-P., F.T.M., and J.S.-A. performed research; J.M.B. and N.P.T.H. analyzed data; and J.M.B.,C.A.G., and F.G. wrote the paper.

Conflict of interest statement: F.T.M. and F.G. are paid employees of Cytex Therapeutics.

This article is a PNAS Direct Submission.1C.A.G. and F.G. contributed equally to this work.2To whom correspondence may be addressed. E-mail: charles.gersbach@duke.edu or guilak@duke.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1321744111/-/DCSupplemental.

E798–E806 | PNAS | Published online February 18, 2014 www.pnas.org/cgi/doi/10.1073/pnas.1321744111

Dow

nloa

ded

by g

uest

on

Mar

ch 2

4, 2

020

used due to their capacity to efficiently transduce target cells andinduce sustained expression of transgenes relevant to cell dif-ferentiation and ECM formation in the context of cartilage tissueengineering. Moreover, as poly-L-lysine (PLL)-coated biomaterialseffectively immobilize other enveloped viruses and facilitate effi-cient transduction of cells (20), PLL was used to functionalize poly(e-caprolactone) (PCL)-derived biomaterials with lentivirus (LV).In this study, we first tested whether PLL could effectively im-mobilize LV to PCL films and lead to efficient transduction ofhuman mesenchymal stem cells (hMSCs). We then tested whetherscaffold-mediated gene delivery with a biomechanically relevant,3D orthogonally woven PCL scaffold could influence ECM for-mation by hMSCs in chondrogenic cultures.

ResultsLV Immobilized to PLL-Coated PCL Films Efficiently Transduces MSCs.The feasibility of LV immobilization on a 3D woven PCL scaf-fold was first tested on 2D PCL films. The goal was to immo-bilize a LV driving expression of eGFP (LVG) onto the PCLfilms and to assess transduction efficiency. Films were incubatedin a solution of PLL (0.002%) followed by incubation in LVG.Human MSCs were cultured for 3 or 5 days on the films beforeharvest for flow cytometry. PLL treatment of the PCL facilitatedhighly efficient cellular transduction, as 93.95 ± 2.28% of thecells were eGFP+ (Fig. 1A). In the absence of PLL, fewer than1% of cells were eGFP+, indicating that PLL treatment of PCLwas necessary and sufficient for mediating the biomaterial–len-tivirus interactions and subsequent transduction of hMSCs.

LV Immobilization Enables Scaffold-Mediated Transduction.Based onthe proof-of-principle from the 2D film model, this approach wasextended to test the effects of LV immobilization on a highlystructured, woven PCL scaffold (Fig. 1B). Cylindrically shapedscaffolds [6 mm diameter (ø) and 700 μm thick] were soaked in0.002% PLL solution overnight. The PLL adsorbed to the PCLscaffold and did not significantly alter the surface topography orhydrophobicity of the scaffold (Fig. S1 A–D). Concentrated LVGsupernatant was pipetted onto the scaffold and incubated for 1 hbefore seeding 500,000 hMSCs uniformly onto each of thescaffolds. After 2 wk in culture, cells were isolated using a pro-nase/collagenase digestion procedure. Flow cytometry showedthat 81.60 ± 4.36% of cells were eGFP+ (Fig. 1 C and D).

Scaffold-Mediated Transduction Results in Cartilage-Specific MatrixAccumulation. We postulated that scaffold-mediated transductionof LV delivering the appropriate transgene could modulate hMSCdifferentiation and extracellular matrix deposition. To test thishypothesis, LV driving the expression of transforming growthfactor β3 (TGF-β3) (LVT) under the control of a constitutive EF-1α promoter was generated (Fig. 2A). Early experiments showedthat TGF-β3 induced chondrogenesis in this system, whereas otherpotential transgenes (e.g., BMP-7, SOX9) were less effective atinitiating chondrogenic differentiation of hMSCs (Fig. S2). ThedsRed fluorescent protein was cotranscriptionally expressed viaan internal ribosomal entry site from the same vector. In initialstudies, cells were either pretransduced in monolayer with TGF-β3–carrying LV (pLVT) before seeding onto the 3D woven PCLscaffold or the scaffold-mediated transduction technique wasused to generate bioactive scaffolds with immobilized TGF-β3LV (iLVT). Either method resulted in production of TGF-β3at concentrations greater than 17 ng/mL (measured via ELISA)from 1 to 3 wks after seeding scaffolds (Fig. 2B), whereas con-structs comprised of nontransduced (NT) cells produced less than120 pg/mL. Quantification of sulfated glycosaminoglycan (sGAG)content via the dimethylmethylene blue (DMMB) assay confirmedthe hypothesis that the pLVT and iLVT treatments resulted insimilar chondrogenic induction after 2 wks (Fig. 2C; pLVT, 12.96μg sGAG/μg DNA vs. iLVT, 12.87 μg sGAG/μg DNA, P = 0.95),whereas NT constructs did not produce significant levels of sGAG(3.76 μg sGAG/μg DNA, P < 0.0002 vs. pLVT and iLVT). Therewas no sGAG detectable by Safranin-O staining of constructscomposed of nontransduced cells (Fig. 2D), whereas moderate toextensive sGAG deposition throughout the pores of the wovenPCL scaffold was observed in the pLVT and iLVT groups (Fig. 2E and F). Furthermore, fluorescence microscopy of hMSCs over-expressing TGF-β3 and dsRed demonstrated that iLVT-transducedcells adopted a spherical morphology, a phenotypic characteristic ofchondrocytes (Fig. 2I). Fluorescence was not observed in non-transduced cells (Fig. 2G), and cells engineered with GFP alone(iLVG) retained an elongated, fibroblast-like morphology (Fig. 2H).We next compared the chondrogenesis induced by scaffold-

mediated transduction to that induced by supplementation ofmedium with recombinant human TGF-β3 (rhTGF-β3). Con-structs were induced to differentiate via iLVT transduction oraddition of TGF-β3 to medium and were then cultured with half-medium changes every 3 days. Scaffold-mediated transductionresulted in overexpression of TGF-β3, measured by quantitativeRT-PCR (qRT-PCR) at days 14 and 28 (D14 and D28) afterchondrogenic induction (Fig. 3A; 397- and 560-fold over non-transduced controls, respectively). Total TGF-β3 levels in culturemedium for iLVT-transduced constructs ranged from 9.6 ng/mLto 21 ng/mL. Although the majority of the protein was bound inthe latent form, 1.1–1.8 ng/mL of free TGF-β3 was detected overthe course of the study (Fig. 3 B and C).Chondrogenic differentiation induced by scaffold-mediated

transduction was further examined by comparing the gene expres-sion profile of constructs from each group. Quantitative RT-PCR

Fig. 1. (A) Representative histogram showing green fluorescence intensityof nontransduced cells (black line) or cells after LVG immobilization to a 2DPCL film and subsequent transduction (green line). Flow cytometry wasperformed 5 days posttransduction. (B) Scanning electron micrograph showingthe architecture of the 3D orthogonal woven PCL scaffold. (Scale bar, 500μm.) (C) Representative histogram showing fluorescence intensity distribu-tion of cells isolated from 3D woven PCL scaffolds 14 days after seedingscaffolds. The nontransduced population is represented by the black curve,and the green curve corresponds to a representative iLVG sample. An av-erage of 81.6 ± 4.36% of cells were eGFP+ (n = 3). (D) Confocal microscopyimage showing eGFP expression of hMSCs transduced via LV immobilizationto the 3D woven PCL scaffold 14 days after seeding. (Scale bar, 100 μm.)

Brunger et al. PNAS | Published online February 18, 2014 | E799

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

4, 2

020

showed an up-regulation of aggrecan (ACAN) and the α1 chainof type II collagen (COL2A1), two markers of chondrogenic dif-ferentiation, in both the iLVT and the rhTGF-β3 groups com-pared with control constructs cultured in the absence of TGF-β3

(Fig. 4 A and B). ACAN expression levels were comparable be-tween iLVT and rhTGF-β3 groups (P = 0.21), but COL2A1 ex-pression levels were higher in the iLVT group at D14 than in therhTGF-β3 group (P < 0.02). Gene expression was also measured

Fig. 2. (A) Schematic of the viral genome used to produce the LVT vector. LTR, long terminal repeat; Ψ, psi packaging signal sequence; RRE, Rev responseelement; cPPT/CTS, central polypurine tract/central termination sequence; EF1α, internal promoter from the Elongation Factor 1α gene; TGF-β3, coding se-quence for human TGF-β3; IRES, internal ribosome entry site; dsRed, coding sequence for the red fluorescent protein from Discosoma sp.; WPRE, woodchuckhepatitis posttranscriptional regulatory element; U3PPT, U3 polypurine tract. (B) An ELISA was performed to compare the level of TGF-β3 production betweenconstructs containing cells pretransduced with the TGF-β3 lentivirus (pLVT) and cells transduced via immobilization of the virus to the 3D woven scaffold(iLVT). Levels were comparable between the two groups (P > 0.05). (C) Sulfated glycosaminoglycan (sGAG) per double-stranded DNA measured via thedimethylmethylene blue assay in nontransduced (NT), pLVT, or iLVT constructs. Bars represent the mean ± SEM (n = 4). (D–F) Safranin-O/fast green/hema-toxylin histology from NT (D), pLVT (E), and iLVT (F) constructs harvested 14 days after chondrogenic induction. (Scale bar, 500 μm.) (G–I) Fluorescence mi-croscopy from NT (G), iLVG (H), or iLVT (I) constructs after 28 days in chondrogenic culture. No fluorescence was observed in NT constructs, whereas cells withiniLVG constructs fluoresced and retained an elongated fibroblast-like morphology spanning along PCL fibers. Cells in the iLVT constructs fluoresced due to theIRES-dsRed component of the TGF-β3 expression cassette. These cells adopted a rounded morphology, indicated by white arrows, suggestive of a chondrocytephenotype. (Scale bar, 100 μm.)

Fig. 3. (A) Quantitative RT-PCR assessing expression of the TGF-β3 transcript from samples in the iLVT group at D14 and D28. Bars represent the mean foldchange in expression ± SEM (n = 4) compared with matched, nontransduced controls and normalized by the r18S reference gene. (B and C) ELISA resultsquantifying the levels of (B) total and (C) free TGF-β3 protein in culture media at various time points after seeding scaffolds in the iLVT group. Total levelsrepresent the sum of protein bound in the latent complex as well as protein free to bind the ALK5 receptor and initiate signaling. Bars represent the mean ±SEM (n = 4).

E800 | www.pnas.org/cgi/doi/10.1073/pnas.1321744111 Brunger et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

4, 2

020

for the α1 chains of types 1 (COL1A1) and 10 collagen (CO-L10A1), markers of a fibrotic and a hypertrophic phenotype, re-spectively (Fig. 5 A and B). COL1A1 and COL10A1 expressionwas up-regulated by both iLVT and rhTGF-β3 treatment at D14,with iLVT levels higher than rhTGF-β3 levels (P < 0.05). Theexpression of COL1A1 decreased to control levels by D28 for bothgroups, whereas COL10A1 remained elevated in both groupscompared with controls. The up-regulation of COL10A1 is con-sistent with previous reports of hypertrophy in hMSCs undergoingchondrogenic differentiation (6, 21, 22). At D28, up-regulation ofthe chondrogenic markers continued and was similar in bothgroups, indicating comparable levels of stable chondrogenic dif-ferentiation after iLVT-mediated differentiation or standard TGF-β3–supplemented differentiation protocols.Measurements of total collagen and sGAG content demon-

strated that the up-regulation of markers of differentiation led tothe development of an extracellular matrix consistent with chon-drogenesis (Fig. 4 C and D). Trends comparing the quantity ofsGAG normalized to double-stranded DNA (Fig. 4C) were equiv-alent to those comparing absolute values (Fig. S3). The rhTGF-β3and iLVT groups showed comparable levels of sGAG/DNA at D14(15.36 ± 1.77 μg/μg and 12.90 ± 0.88 μg/μg, respectively, P = 0.21),although the NT group without TGF-β3 showed significantly lowerlevels (4.67 ± 0.25 μg/μg, P < 0.0002). Sulfated GAG levels in-creased significantly by D28 (P < 0.00001) in the rhTGF-β3 and

iLVT groups, but were not statistically different from one another(35.97 ± 1.97 μg/μg and 36.07 ± 1.79 μg/μg, respectively, P = 0.96).Culture time did not influence sGAG levels in the NT group (P =0.96), which remained lower than those in the rhTGF-β3 and iLVTgroups (P < 0.00001). Similar trends held for total collagen levelsnormalized to DNA content (Fig. 4D), although a significant dif-ference at D14 between the rhTGF-β3 and iLVT groups was ob-served (P= 0.045). The significance of this difference was lost whentotal collagen content of constructs was not normalized to constructDNA content (P = 0.053, Fig. S2).Histological and immunohistochemical assays supported the

gene expression and biochemical evidence of chondrogenic dif-ferentiation in both the iLVT and the rhTGF-β3 groups. Safranin-O and type II collagen labeling showed limited ECM accumula-tion at D14, but extensive assembly of GAG and collagen II ECMconstituents was observed throughout the pores of the 3D wovenscaffold in the rhTGF-β3 and iLVT groups at D28 (Fig. 4 E–L).Although type I and type X collagens were detected in D28 sam-ples, they appeared less prominent than either type II collagen orGAG (Fig. 5 C–F). Limited ECM deposition was observed in theNT group.Confined compression testing was used to evaluate the me-

chanical properties of the tissue replacements generated in cul-ture, with the goal of attaining a compressive modulus similar tothat of native articular cartilage (23). The aggregate modulus

Fig. 4. (A and B) Relative expression levels of aggrecan (ACAN) and the α1 chain of type II collagen (COL2A1) comparing the hMSC response to rhTGF-β3– oriLVT-mediated differentiation at D14 (A) or D28 (B) after chondrogenic induction. Bars represent the mean fold change in expression ± SEM (n = 4) comparedwith matched, nontransduced controls and as normalized by the r18S reference gene. (C and D) Quantification of the development of cartilaginous ECMcomponents in the nontransduced (NT), rhTGF-β3 (rhT), and immobilized lentiviral TGF (iLVT) groups. Sulfated glycosaminoglycan content and total collagencontent were normalized to DNA content. Bars represent means ± SEM (n = 6). Groups not sharing the same letter or symbol are statistically different (P <0.05). (E–H) Safranin-O/fast green/hematoxylin staining of D28 rhTGF-β3 (E and F) and iLVT (G and H) constructs. [Scale bars, 500 μm (E and G) and 100 μm(F and H).] (I–L) Immunohistochemistry probing for type II collagen in D28 rhTGF-β3 (I and J) and iLVT (K and L) constructs. [Scale bars, 500 μm (I and K) and100 μm (J and L).]

Brunger et al. PNAS | Published online February 18, 2014 | E801

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

4, 2

020

(HA) of the scaffolds directly after cell seeding (D0) was 0.26 ±0.024 MPa. The HA of the cell-seeded constructs was not influ-enced by the addition of TGF-β3 to culture conditions (Fig. 6A),as no significant differences were detected among treatments(P > 0.18); however, HA did increase for each group comparedwith D0 (P < 0.003). The hydraulic permeability (k) was influ-enced by experimental treatment (Fig. 6B), with samples in therhTGF-β3 condition more rapidly approaching values similar toarticular cartilage than NT samples (P = 0.013). This differencewas not maintained to D28, as values for k were not significantlydifferent in any of the three culture groups. Permeability valuesdecreased in all groups in comparison with D0 samples (P < 0.037).

DiscussionThe findings of this study show that scaffold-mediated trans-duction of hMSCs with lentiviral vectors driving expression ofTGF-β3 leads to potent differentiation toward a chondrogeniclineage and accumulation of a cartilage-like ECM. Histologicaland biochemical measures confirmed the hypothesis that theiLVT treatment induced a chondrocyte-like phenotype, as dis-played by hMSCs adopting a spherical morphology and de-positing an ECM rich in type II collagen and glycosaminoglycan,major components of articular cartilage. By nearly all measuresof gene regulation, protein content, and biomechanical prop-erties, the level of chondrogenic differentiation achieved atD28 by immobilizing LVT was indistinguishable from that usingrhTGF-β3. The observed increase in sGAG and collagen content

is consistent with previous reports from our group and otherswhen using hMSCs for chondrogenesis (24–26).In these studies, LV was used to drive constitutive over-

expression of TGF-β3 to induce chondrogenesis. An extensivebody of literature has shown that TGF-β3 is a potent regulator ofhMSC differentiation to the chondrogenic lineage (27–32). Var-ious other transgenes and gene products, including IGF1, SOX9,and multiple members of the TGF-β superfamily, including TGF-β1, BMP-2, BMP-4, and BMP-7, and combinations thereof, havebeen used to induce chondrogenesis in progenitor cells (21, 33–39). In these studies, the induction of chondrogenesis with viralvectors generally has been performed in the context of monolayertransduction and chondrogenesis in pellet cultures. Althoughthese studies demonstrate the feasibility of virus-based transductionmethods to drive MSC differentiation, the ability to transduce cellsin situ via scaffold-mediated gene delivery opens new avenues forgenerating implant-ready, bioactive tissue engineering scaffolds ca-pable of restoring function to a joint immediately upon implantationwhile also driving the development of a mature, viable tissue re-placement using the host as the bioreactor.Moreover, the results from these studies indicate that the

scaffold-mediated transduction technique may provide a meansof transducing cells in vivo simply by seeding them intraoperativelyonto a cell-instructive, biomechanically functional scaffold thatdirects cell lineage commitment and ECM development in a con-trolled and persistent manner. Using this approach would abro-gate the need for ex vivo expansion of cells used in tissue engi-neering as well as the ex vivo conditioning of the maturing tissuesubstitute by allowing the host’s own cells to recapitulate func-tional tissues after infiltrating an acellular bioactive scaffold. Ourfindings suggest that scaffold-mediated gene delivery could pro-vide an efficient means of in situ transduction, owing to the abilityof PLL to immobilize vectors to the biomaterial, as well as theefficiency of transduction of the VSV-G pseudotyped lentivirus. In

Fig. 5. (A and B) Relative expression levels of the type I chains of collagens 1(COL1A1) and 10 (COL10A1), comparing the hMSC response to rhTGF-β3– oriLVT-mediated differentiation at D14 (A) or D28 (B) after chondrogenic in-duction. Bars represent the mean fold change in expression ± SEM (n = 4)compared with matched, nontransduced controls and as normalized by ther18S reference gene. Groups not sharing the same letter or symbol are sta-tistically different (P < 0.05). (C and D) Immunohistochemistry probing type Icollagen in D28 rhTGF-β3 (C) and iLVT (D) samples. (E and F) Immunohisto-chemistry probing type X collagen in D28 rhTGF-β3 (E) and iLVT (F) samples.(Scale bars, 100 μm.)

Fig. 6. (A and B) Aggregate modulus (HA) and hydraulic permeability (k) ofnontransduced (NT), rhTGF-β3 (rhT), and immobilized lentiviral TGF (iLVT)constructs cultured for 14 days or 28 days in chondrogenic induction me-dium. Dotted lines represent D0 values.

E802 | www.pnas.org/cgi/doi/10.1073/pnas.1321744111 Brunger et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

4, 2

020

addition, we have previously demonstrated that passive adsorptionof ECM onto highly cationic surfaces can direct both viral im-mobilization and cell adhesion (20), opening the possibility offunctionalizing polymeric scaffolds with gene delivery vehicles aswell as ECM components or synthetic peptides capable of en-hancing cell differentiation and tissue formation (12, 14). How-ever, important questions, such as whether endogenous cells thatinfiltrate the scaffold at the site of implantation can be driven todifferentiate toward the chondrocyte lineage, remain unanswered.Our findings extend previous work showing biomaterial-

mediated gene delivery using viral vectors (20, 40–49). The majorityof this work demonstrated the ability to specify the spatial lo-cation of delivered vectors, rather than the induction of tissue-specific differentiation. An exception to this was the work byPhillips et al., in which fibroblasts seeded on a collagen scaffoldcontaining PLL-immobilized γ-retrovirus were transdifferentiatedto the osteogenic lineage (43). An important advance of thecurrent study was the use of PLL to immobilize gene deliveryvectors to a highly structured, 3D woven PCL scaffold. Althoughother methods, such as 3D printing, electrospinning, and solventcasting/salt leaching are available for fabricating polymer-basedscaffolds for cartilage tissue engineering, the orthogonally wovenPCL scaffold possesses controlled anisotropic and nonlinearmechanical properties before cell seeding (50, 51). Although theproperties of this 3D woven scaffold have been extensivelycharacterized previously, here we show that PLL treatment doesnot alter the porosity, surface topography, or hydrophobicity ofthe scaffold (SI Results). These properties closely mimic thebiomechanical features of articular cartilage and would allow forimmediate loading upon implantation. This aspect may be crucialto the goal of inducing tissue generation in situ upon implantationof a degradable, bioactive scaffold, because functional restorationof a joint is likely to be facilitated by a biomaterial possessingmechanical properties approximating those of articular cartilage,while also permitting the formation of a viable, cartilage neotissuethat could eventually serve as the load-bearing tissue after thecomplete degradation of the PCL. As demonstrated here, thedevelopment of cartilaginous ECM constituents served to en-hance (i.e., decrease) the hydraulic permeability of the constructby four orders of magnitude compared with D0 constructs, aspores of the woven scaffold were filled in with neotissue thatconsolidated the construct. Importantly, however, the aggregatemodulus of the tissue engineered constructs was the same orderof magnitude as that of native articular cartilage before depositionof ECM, and these properties were maintained throughout theculture period as the compressive properties are dominated bythose of the 3D woven scaffold (50). In contrast, scaffolds com-posed of collagen, alginate, or fibrin-based hydrogels capable ofencapsulating gene delivery vectors possess compressive proper-ties below those of native articular cartilage (52–54). In thisregard, the 3D woven structure provides several highly desirableproperties as a scaffold for biomaterial-mediated gene delivery forin situ tissue engineering.An important feature of the current study includes the use of

self-inactivating lentiviral vectors in place of γ-retrovirus (43),nonintegrating viral vectors (16, 40–42, 55), or naked plasmidDNA (15). Although considerable progress has been made withregard to spatial control of gene delivery to cells, transient geneexpression and poor transfection/transduction efficiencies maynot allow sufficient control of cell phenotype and tissue develop-ment. LV represents an attractive option for efficient and sus-tained gene delivery, in addition to other advantageous propertiesincluding high packaging capacity (>10 kb), low immunogenicity,low cytotoxicity, and broad tropism (56–59). The development ofself-inactivating vectors has addressed the serious safety concernsof earlier generations of viral systems (60) and made LV an at-tractive option for clinical use that overcomes the shortcomings ofmany alternative vector systems.

The use of LV likely explains the efficient transduction of theslowly dividing hMSCs compared with that in other studies in-volving γ-retroviral vectors. Using LV, an efficiency of scaffold-mediated transduction of roughly 82% was achieved, whereasa previous study of scaffold-mediated gene delivery to fibroblastswith retrovirus showed a transduction efficiency of ≥20% (43).Therefore, biomaterial-mediated delivery of LV may serve as apromising approach for applications requiring robust and uniformtransgene expression, such as those seeking to generate homo-geneous tissue formation or those using biomaterials as deliverydepots of anti-inflammatory gene products for protection againstfurther degeneration of musculoskeletal tissues.Although effective at inducing chondrogenesis, the safety of

constitutive overexpression of TGF-β3 serves as one of the draw-backs of this study. To circumvent this, future studies may be di-rected toward the replacement of the constitutive EF-1α promoterwith a drug-inducible promoter (61), which could confer engineeredcontrol of the duration and magnitude of transgene expression.Moreover, the technique of biomaterial-mediated gene deliverymay be applied toward engineering complex tissues or tissue in-terfaces for regenerative medicine (62–64).Toward the goal of in situ tissue engineering, it is important to

consider the effects of the synovial environment on viral trans-duction efficiency and the ability to transduce target cells in vivo.Although we did not directly test transduction efficiency in thecontext of synovial fluid, target cells were transduced in mediumcontaining 10% FBS. Synovial fluid is considered a plasma di-alysate, with the major difference between serum and synovialfluid being the abundance of hyaluronic acid in the latter (65).The presence of synovial fluid does not preclude lentiviral trans-duction, as previous studies have demonstrated transduction ofsynovial fibroblasts as well as cells in ligaments, tendons, and jointcapsule via direct intra-articular injection of virus to synovial joints(66–68). Importantly, we showed that material-mediated trans-duction of hMSCs remained >90% after rinsing the substratesupon which lentivirus was immobilized. Moreover, we have alsoshown that incubation of PLL-coated surfaces with retrovirus for16 h results in high levels of transduction of fibroblasts (∼70%)(20), suggesting that the half-life of the vector does not precludeefficient transduction. Collectively, these results support the po-tential of viral transduction in the joint environment.These studies demonstrate that scaffold-mediated gene de-

livery may serve as a promising strategy for in situ engineering ofmusculoskeletal tissues. By coupling the biomechanically rele-vant woven scaffold to an effective gene delivery vehicle carryingthe appropriate transgene, we have demonstrated both the proof-of-principle of scaffold-mediated delivery to hMSCs and the po-tential utility of this method for tissue engineering of articularcartilage. In this work, a relatively inert material such as PCL wasfunctionalized with the ability to specify cell differentiation andECM production. As such, this technique for LV immobilizationcan be applied in the context of other PCL-based scaffolds andpotentially using other materials than PCL. More importantly, itis possible that our approach to scaffold-mediated transductioncould be extended to a variety of applications using numerousother cell types, transgenes, and drug-inducible promoters (69).When combined with the appropriate scaffold structure andmaterial, we envision broad utility of this technique in the area ofrecapitulating the complex organization of heterogeneous tissues,such as osteochondral tissues, and in the delivery of therapeutictransgenes capable of mediating protection and repair of tissuesin the context of a diseased or damaged joint.

MethodsHuman Mesenchymal Stem Cell Isolation and Expansion. Bone marrow wasobtained in concordance with an approved Institutional Review Board ex-emption as discarded and de-identified waste tissue from adult bone mar-row transplant donors at Duke University Medical Center. Adherent cells

Brunger et al. PNAS | Published online February 18, 2014 | E803

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

4, 2

020

were expanded in DMEM-low glucose (Gibco) supplemented with L-gluta-mine, sodium pyruvate, 1 ng/mL basic fibroblast growth factor (Roche),penicillin/streptomycin (Gibco), and 10% lot-selected FBS (HyClone). Isolatedcells from three donors were combined in equal numbers in a superlot afterpassage two and used as described below.

Lentivirus Production.All plasmids used in the production of LVwere obtainedvia the Addgene Plasmid Repository. The cDNA for TGF-β3 was obtained fromHEK-293T cells via RT-PCR and cloned into a lentiviral transfer vector[Addgene Plasmid 12250 (70) modified with a custom multiple-cloning sitedownstream of the EF-1α promoter]. Virus driving overexpression of eGFPwas produced using either Addgene Plasmid 21320 (71) or 11645 (72)modified to contain an EF1-α-eGFP cassette. To produce VSV-G pseudotypedLV, 4.2e6 HEK-293T cells were plated onto each 10-cm dish in DMEM–highglucose supplemented with L-glutamine, sodium pyruvate, and 10% FBS(D-10 medium). The following day, cells on each dish were cotransfected withthe appropriate lentiviral transfer plasmid (20 μg), the second-generationpackaging plasmid, psPAX2 (Addgene 12260, 15 μg), and the pMD2.G(Addgene 12259, 6 μg) envelope plasmid via calcium phosphate precipita-tion (73). After an overnight incubation, transfection medium was removedand replaced with 12 mL of fresh 293T medium. Conditioned medium con-taining LV was collected at 36 h and 60 h posttransfection. The raw lentiviralsupernatant was cleared of producer cells via centrifugation and subsequentfiltration through 0.45-μm cellulose acetate filters, snap frozen, and storedat −80 °C until future use.

Polycaprolactone Film Production and LV Immobilization. PCL (molecularweight 70,000–90,000; Sigma-Aldrich) was dissolved in glacial acetic acid ata 10% wt/vol ratio as in ref. 40. PCL solution was pipetted into a well plate,and the acid was allowed to evaporate overnight. Films were quenched with1 N NaOH, washed in DPBS, and aseptically processed through an ethanolgradient in preparation for cell culture experiments. Films were then in-cubated in PLL solution (Sigma-Aldrich) diluted to 0.002% with DPBS for 1 hto allow for passive adsorption. PLL was aspirated, and LV-eGFP supernatantwas pipetted into the well plate and allowed to incubate for 4 h at 37 °C(20). DPBS was used in place of PLL in untreated control wells. After LVimmobilization, viral supernatant was aspirated, the wells were rinsed withDPBS, and then human MSCs at passages 4–6 were plated in wells ata density of 5,000/cm2 or 10,000/cm2 and cultured for an additional 3–5 days.Cells were trypsinized, washed with DPBS, and fixed with 1% paraformaldehyde(PFA) before flow cytometry. A total of two experiments were performedwith n = 2–3 wells per group per experiment. Data presented are from theaverage of all experiments.

Scaffold-Mediated Transduction and Chondrogenic Differentiation. Three-dimensional woven PCL scaffolds were produced as previously described (50,51). The orthogonally woven scaffold contained three warp and four weftlayers interlocked by a third set of fibers passing through the thickness ofthe structure. Scaffold pore sizes ranged from 100 μm to 300 μm, with a porefraction volume of ∼50%, which was not altered by PLL treatment (Fig. S3C).Cylindrical samples (6 mm ø) were punched from the scaffold material,treated with 1 N NaOH to increase wettability, rinsed in water, and asepti-cally processed through an ethanol gradient for tissue culture experiments.Sterilized scaffolds were soaked overnight in 0.002% PLL before viral im-mobilization and cell seeding. Atomic force microscopy (74), contact anglemeasurements (75), and microcomputed tomography (76) were performedas described in SI Methods to compare PLL-treated scaffolds to control PBS-soaked scaffolds to determine the effect of the PLL coating on the originalscaffold properties, which have been extensively characterized previously.

Before transduction, conditioned 293T medium containing LV was con-centrated roughly 50-fold in a 100-kDa molecular mass cutoff filter (Milli-pore). In the pretransduced LVT group, passage 4 hMSCs were culturedovernight in 12 mL hMSC expansion medium supplemented with 300 μL ofconcentrated LV and 4 μg/mL polybrene (Sigma-Aldrich). Transduction me-dium was then exchanged for fresh expansion medium until the end of thepassage and scaffold seeding as described below.

Scaffolds were transferred from PLL-containing wells to low-attachmentwell plates (Corning), and 50 μL of concentrated LV was pipetted onto eachscaffold in the iLVT group. Control scaffolds were treated with 50 μL of freshD-10 medium. Scaffolds were incubated at room temperature for 1 h. Pas-sage 4 hMSCs were trypsinized, counted, and resuspended at a density of25e6/mL. Each scaffold received 20 μL of the cell suspension, and the re-sultant constructs were incubated for 1 h in a humidified incubator at 37 °Cfor 1 h before the addition of 1 mL of hMSC expansion medium. Freshmedium was exchanged 3 days later. Six days after seeding, constructs were

induced to differentiate by changing the medium to serum-free, DMEM–

high glucose (Gibco) supplemented with ITS+ (BD), penicillin/streptomycin(Gibco), dexamethasone (100 nM), ascorbic acid (50 μg/mL), and L-proline (40μg/mL) and, in the rhTGF-β3 group only, 10 ng/mL rhTGF-β3 (R&D Systems). Ahalf medium change occurred every 3 days for all groups until samples wereharvested at 14 days or 28 days after inducing chondrogenic differentiation.An aliquot of medium was collected from four samples during culture andused for an ELISA (R&D Systems) to measure TGF-β3 concentrations per themanufacturer’s instructions.

Biochemical and Biomechanical Assays. Samples used for biochemical andbiomechanical analyses were harvested, rinsed with DPBS, and stored at −20 °Cuntil testing. Test specimens were obtained by taking a 3-mm ø biopsy punchfrom constructs. The biopsy was used for mechanical testing whereas the an-nulus was lyophilized and then digested in 125 μg/mL papain at 60 °C over-night for biochemical analysis, using the picogreen assay (Invitrogen) tomeasure double-stranded DNA, the ortho-hydroxyproline assay (77) for mea-suring total collagen content, and the dimethylmethylene blue assay (78) formeasuring the total sulfated glycosaminoglycan content of constructs (n = 6per group).

Aggregate modulus and hydraulic permeability were determined viaconfined compression testing, using a Bose ELF 3200 materials testingsystem. Specimens (n = 4 per group) were subjected to a 10-gF tare loadand a 30-gF step compressive load until equilibration. To calculate HA and k,creep data were fitted to a three-parameter, nonlinear least-squares re-gression fit based on the biphasic, viscoelastic model describing articularcartilage (23).

Gene Expression Analysis. Samples for gene expression analysis were har-vested, rinsed in DPBS, and snap frozen until further processing. Total RNAwas isolated per manufacturer’s recommendations, using TRIzol (Invitrogen),following tissue homogenization with a biopulverizer chilled in liquid ni-trogen. Reverse transcription was performed using the superscript VILOcDNA synthesis kit (Invitrogen) per manufacturer’s instructions. QuantitativeRT-PCR was performed with n = 4 samples per group on a StepOnePlus, usingPower Sybr (Applied Biosystems) per manufacturer’s instructions. Gene ex-pression was probed using the following primer pairs designed to containintervening introns to avoid aberrant genomic DNA amplification: ACAN[forward (F), 5′-CACTTCTGAGTTCGTGGAGG-3′; reverse (R), 5′-ACTGGACT-CAAAAAGCTGGG-3′]; COL1A1 (F, 5′-TGTTCAGCTTTGTGGACCTC-3′; R, 5′-TTCTGTACGCAGGTGATTGG-3′); COL2A1 (F, 5′-GGCAATAGCAGGTTCACGTA-3′; R, 5′-CTCGATAACAGTCTTGCCCC-3′); COL10A1 (F, 5′-CATAAAAGGCCC-ACTACCCAAC-3′; R, 5′-ACCTTGCTCTCCTCTTACTGC-3′); and TGFB3 (F, 5′-GGAAAACACCGAGTCGGAATAC-3′; R, 5′-GCGGAAAACCTTGGAGGTAAT-3′).Ribosomal 18s (F, 5′-CGGCTACCACATCCAAGGAA-3′; R, 5′-GGGCCTCGAAA-GAGTCCTGT-3′) was used as a reference gene for normalization. All oligo-nucleotides were synthesized by IDT. Efficiencies for each primer pair weredetermined and fold changes in expression levels were calculated using thePfaffl method (79).

Histological and Immunohistochemical Processing. Samples for histology andimmunohistochemistry were rinsed in DPBS upon harvest, fixed in 4%paraformaldehyde for 48 h, paraffin embedded, and sectioned at 10 μmthickness. Samples were stained with Safranin-O/fast green/hematoxylin,using standard protocols. For immunohistochemistry, sections were strippedof paraffin using xylenes and hydrated in an ethanol series, and epitopeswere retrieved with Digest-All pepsin solution (Invitrogen) before peroxi-dase quenching. Sections were blocked with goat serum, incubated witha primary antibody (Collagen Type I, mouse monoclonal, Abcam ab90395;Collagen Type II, mouse monoclonal, Developmental Studies HybridomaBank, University of Iowa, II-II6B3; Collagen Type X, mouse monoclonal,Sigma C7974) for 1 h at room temperature, incubated with a biotinylatedsecondary antibody (goat anti-mouse; Abcam ab97021) at room tempera-ture, treated with streptavidin HRP conjugate, developed using AEC, andmounted with VectaMount (Vector Labs). Human osteochondral tissue wasused as positive controls and samples incubated without primary antibodywere used to confirm staining specificity.

Fluorescence Microscopy. Fluorescence microscopy was performed on a ZeissS100 Axiovert inverted microscope, and confocal microscopy was performedon a Zeiss LSM 510 laser scanning microscope.

Statistical Analysis. Statistical analysis was performed in the Statistica 7software package. ELISA data comparing pLVT and iLVT treatments wereanalyzed using an independent paired t test. Other analyses were performed

E804 | www.pnas.org/cgi/doi/10.1073/pnas.1321744111 Brunger et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

4, 2

020

using ANOVA with Fisher’s protected least significance difference post hoctest with α = 0.05. For biomechanical analyses, all treatments were alsocompared with reference samples harvested 1 h after seeding, using a pairedt test with α = 0.05. For qRT-PCR comparisons, fold change values were log-transformed before statistical analysis. Average group values and SEMs werecalculated in the logarithmic space before transforming data to linear valuesfor reporting fold changes.

ACKNOWLEDGMENTS. The authors thank Arnold Caplan and Don Lennonfor providing hMSCs used in pilot experiments, and thank Dr. Nelson Chaofor providing bone marrow [National Institutes of Health (NIH) Grant P01CA47741]. This work was supported by NIH Grants AR061042, AR50245,AR48852, AG15768, AR48182, AG46927, and OD008586; National Science Foun-dation Faculty Early Career Development (CAREER) Award CBET-1151035; theCollaborative Research Center; AO Foundation, Davos, Switzerland; the NancyTaylor Foundation for Chronic Diseases; and the Arthritis Foundation.

1. Kock L, van Donkelaar CC, Ito K (2012) Tissue engineering of functional articularcartilage: The current status. Cell Tissue Res 347(3):613–627.

2. Butler DL, Goldstein SA, Guilak F (2000) Functional tissue engineering: The role ofbiomechanics. J Biomech Eng 122(6):570–575.

3. Gao J, Yao JQ, Caplan AI (2007) Stem cells for tissue engineering of articular cartilage.Proc Inst Mech Eng H 221(5):441–450.

4. Kretlow JD, Klouda L, Mikos AG (2007) Injectable matrices and scaffolds for drugdelivery in tissue engineering. Adv Drug Deliv Rev 59(4–5):263–273.

5. LeBaron RG, Athanasiou KA (2000) Ex vivo synthesis of articular cartilage. Bio-materials 21(24):2575–2587.

6. Bian L, Zhai DY, Zhang EC, Mauck RL, Burdick JA (2012) Dynamic compressive loadingenhances cartilage matrix synthesis and distribution and suppresses hypertrophy inhMSC-laden hyaluronic acid hydrogels. Tissue Eng Part A 18(7–8):715–724.

7. Moffat KL, et al. (2009) Novel nanofiber-based scaffold for rotator cuff repair andaugmentation. Tissue Eng Part A 15(1):115–126.

8. Corsi KA, Schwarz EM, Mooney DJ, Huard J (2007) Regenerative medicine in ortho-paedic surgery. J Orthop Res 25(10):1261–1268.

9. Stern-Straeter J, Riedel F, Bran G, Hörmann K, Goessler UR (2007) Advances in skeletalmuscle tissue engineering. In Vivo 21(3):435–444.

10. Wu Y, et al. (2010) Electrosprayed core-shell microspheres for protein delivery. ChemCommun 46(26):4743–4745.

11. Lee CH, et al. (2010) Regeneration of the articular surface of the rabbit synovial jointby cell homing: A proof of concept study. Lancet 376(9739):440–448.

12. Reyes CD, Petrie TA, Burns KL, Schwartz Z, García AJ (2007) Biomolecular surfacecoating to enhance orthopaedic tissue healing and integration. Biomaterials 28(21):3228–3235.

13. Stahl PJ, Yu SM (2012) Encoding cell-instructive cues to PEG-based hydrogels via triplehelical peptide assembly. Soft Matter 8:10409–10418.

14. Hwang NS, Varghese S, Li H, Elisseeff J (2011) Regulation of osteogenic and chon-drogenic differentiation of mesenchymal stem cells in PEG-ECM hydrogels. Cell TissueRes 344(3):499–509.

15. Shea LD, Smiley E, Bonadio J, Mooney DJ (1999) DNA delivery from polymer matricesfor tissue engineering. Nat Biotechnol 17(6):551–554.

16. Basile P, et al. (2008) Freeze-dried tendon allografts as tissue-engineering scaffoldsfor Gdf5 gene delivery. Mol Ther 16(3):466–473.

17. Dang JM, Leong KW (2006) Natural polymers for gene delivery and tissue engi-neering. Adv Drug Deliv Rev 58(4):487–499.

18. Liao IC, Chen S, Liu JB, Leong KW (2009) Sustained viral gene delivery through core-shell fibers. J Control Release 139(1):48–55.

19. Fishbein I, et al. (2013) Adenoviral vector tethering to metal surfaces via hydrolyzablecross-linkers for the modulation of vector release and transduction. Biomaterials 34(28):6938–6948.

20. Gersbach CA, Coyer SR, Le Doux JM, García AJ (2007) Biomaterial-mediated retroviralgene transfer using self-assembled monolayers. Biomaterials 28(34):5121–5127.

21. Cucchiarini M, Orth P, Madry H (2013) Direct rAAV SOX9 administration for durablearticular cartilage repair with delayed terminal differentiation and hypertrophy invivo. J Mol Med 91(5):625–636.

22. Mueller MB, Tuan RS (2008) Functional characterization of hypertrophy in chondro-genesis of human mesenchymal stem cells. Arthritis Rheum 58(5):1377–1388.

23. Mow VC, Kuei SC, Lai WM, Armstrong CG (1980) Biphasic creep and stress relaxationof articular cartilage in compression? Theory and experiments. J Biomech Eng 102(1):73–84.

24. Ousema PH, et al. (2012) The inhibition by interleukin 1 of MSC chondrogenesis andthe development of biomechanical properties in biomimetic 3D woven PCL scaffolds.Biomaterials 33(35):8967–8974.

25. Valonen PK, et al. (2010) In vitro generation of mechanically functional cartilagegrafts based on adult human stem cells and 3D-woven poly(epsilon-caprolactone)scaffolds. Biomaterials 31(8):2193–2200.

26. Kim M, Erickson IE, Choudhury M, Pleshko N, Mauck RL (2012) Transient exposure toTGF-β3 improves the functional chondrogenesis of MSC-laden hyaluronic acid hy-drogels. J Mech Behav Biomed Mater 11:92–101.

27. Ballock RT, Heydemann A, Izumi T, Reddi AH (1997) Regulation of the expression ofthe type-II collagen gene in periosteum-derived cells by three members of thetransforming growth factor-beta superfamily. J Orthop Res 15(3):463–467.

28. Mackay AM, et al. (1998) Chondrogenic differentiation of cultured human mesen-chymal stem cells from marrow. Tissue Eng 4(4):415–428.

29. Serra R, et al. (1997) Expression of a truncated, kinase-defective TGF-beta type II re-ceptor in mouse skeletal tissue promotes terminal chondrocyte differentiation andosteoarthritis. J Cell Biol 139(2):541–552.

30. Seyedin SM, et al. (1986) Cartilage-inducing factor-A. Apparent identity to trans-forming growth factor-beta. J Biol Chem 261(13):5693–5695.

31. Leonard CM, et al. (1991) Role of transforming growth factor-beta in chondrogenicpattern formation in the embryonic limb: Stimulation of mesenchymal condensation

and fibronectin gene expression by exogenenous TGF-beta and evidence for en-dogenous TGF-beta-like activity. Dev Biol 145(1):99–109.

32. Madry H, Rey Rico A, Venkatesan JK, Johnstone B, Cucchiarini M (2013) TGF-beta-releasing scaffolds for cartilage tissue engineering. Tissue Eng Part B Rev, 10.1089/ten.teb.2013.0271.

33. Neumann AJ, Alini M, Archer CW, Stoddart MJ (2013) Chondrogenesis of human bonemarrow-derived mesenchymal stem cells is modulated by complex mechanical stim-ulation and adenoviral-mediated overexpression of bone morphogenetic protein 2.Tissue Eng Part A 19(11–12):1285–1294.

34. Palmer GD, et al. (2005) Gene-induced chondrogenesis of primary mesenchymal stemcells in vitro. Mol Ther 12(2):219–228.

35. Steinert AF, et al. (2009) Enhanced in vitro chondrogenesis of primary mesenchymalstem cells by combined gene transfer. Tissue Eng Part A 15(5):1127–1139.

36. Kramer J, et al. (2000) Embryonic stem cell-derived chondrogenic differentiation invitro: Activation by BMP-2 and BMP-4. Mech Dev 92(2):193–205.

37. Madry H, Kaul G, Zurakowski D, Vunjak-Novakovic G, Cucchiarini M (2013) Cartilageconstructs engineered from chondrocytes overexpressing IGF-I improve the repair ofosteochondral defects in a rabbit model. Eur Cell Mater 25:229–247.

38. Weimer A, et al. (2012) Benefits of recombinant adeno-associated virus (rAAV)-me-diated insulinlike growth factor I (IGF-I) overexpression for the long-term re-construction of human osteoarthritic cartilage by modulation of the IGF-I axis. MolMed 18:346–358.

39. Evans CH, Ghivizzani SC, Robbins PD (2013) Arthritis gene therapy and its tortuouspath into the clinic. Transl Res 161(4):205–216.

40. Hu WW, et al. (2009) The use of reactive polymer coatings to facilitate gene deliveryfrom poly (epsilon-caprolactone) scaffolds. Biomaterials 30(29):5785–5792.

41. Hu WW, Lang MW, Krebsbach PH (2008) Development of adenovirus immobilizationstrategies for in situ gene therapy. J Gene Med 10(10):1102–1112.

42. Hu WW, Wang Z, Hollister SJ, Krebsbach PH (2007) Localized viral vector delivery toenhance in situ regenerative gene therapy. Gene Ther 14(11):891–901.

43. Phillips JE, Burns KL, Le Doux JM, Guldberg RE, García AJ (2008) Engineering gradedtissue interfaces. Proc Natl Acad Sci USA 105(34):12170–12175.

44. Shin S, Salvay DM, Shea LD (2010) Lentivirus delivery by adsorption to tissue engi-neering scaffolds. J Biomed Mater Res A 93(4):1252–1259.

45. Shin S, Shea LD (2010) Lentivirus immobilization to nanoparticles for enhanced andlocalized delivery from hydrogels. Mol Ther 18(4):700–706.

46. Shin S, Tuinstra HM, Salvay DM, Shea LD (2010) Phosphatidylserine immobilization oflentivirus for localized gene transfer. Biomaterials 31(15):4353–4359.

47. Padmashali RM, Andreadis ST (2011) Engineering fibrinogen-binding VSV-G envelopefor spatially- and cell-controlled lentivirus delivery through fibrin hydrogels. Bio-materials 32(12):3330–3339.

48. Lei P, Padmashali RM, Andreadis ST (2009) Cell-controlled and spatially arrayed genedelivery from fibrin hydrogels. Biomaterials 30(22):3790–3799.

49. Jang JH, Schaffer DV, Shea LD (2011) Engineering biomaterial systems to enhanceviral vector gene delivery. Mol Ther 19(8):1407–1415.

50. Moutos FT, Freed LE, Guilak F (2007) A biomimetic three-dimensional woven com-posite scaffold for functional tissue engineering of cartilage. Nat Mater 6(2):162–167.

51. Moutos FT, Guilak F (2010) Functional properties of cell-seeded three-dimensionallywoven poly(epsilon-caprolactone) scaffolds for cartilage tissue engineering. TissueEng Part A 16(4):1291–1301.

52. LeRoux MA, Guilak F, Setton LA (1999) Compressive and shear properties of alginategel: Effects of sodium ions and alginate concentration. J Biomed Mater Res 47(1):46–53.

53. Weisel JW (2004) The mechanical properties of fibrin for basic scientists and clinicians.Biophys Chem 112(2–3):267–276.

54. Spiller KL, Maher SA, Lowman AM (2011) Hydrogels for the repair of articular carti-lage defects. Tissue Eng Part B Rev 17(4):281–299.

55. Hu WW, Lang MW, Krebsbach PH (2009) Digoxigenin modification of adenovirus tospatially control gene delivery from chitosan surfaces. J Control Release 135(3):250–258.

56. Wiznerowicz M, Trono D (2005) Harnessing HIV for therapy, basic research and bio-technology. Trends Biotechnol 23(1):42–47.

57. Kumar M, Keller B, Makalou N, Sutton RE (2001) Systematic determination of thepackaging limit of lentiviral vectors. Hum Gene Ther 12(15):1893–1905.

58. Zufferey R, et al. (1998) Self-inactivating lentivirus vector for safe and efficient in vivogene delivery. J Virol 72(12):9873–9880.

59. Dull T, et al. (1998) A third-generation lentivirus vector with a conditional packagingsystem. J Virol 72(11):8463–8471.

60. Schambach A, Baum C (2008) Clinical application of lentiviral vectors - concepts andpractice. Curr Gene Ther 8(6):474–482.

61. Barde I, et al. (2006) Efficient control of gene expression in the hematopoietic systemusing a single Tet-on inducible lentiviral vector. Mol Ther 13(2):382–390.

62. Atala A, Kasper FK, Mikos AG (2012) Engineering complex tissues. Sci Transl Med 4(160):160rv112.

Brunger et al. PNAS | Published online February 18, 2014 | E805

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Mar

ch 2

4, 2

020

63. Reed CR, et al. (2009) Composite tissue engineering on polycaprolactone nanofiberscaffolds. Ann Plast Surg 62(5):505–512.

64. Smith L, Xia Y, Galatz LM, Genin GM, Thomopoulos S (2012) Tissue-engineeringstrategies for the tendon/ligament-to-bone insertion. Connect Tissue Res 53(2):95–105.

65. Newton CD, Nunamaker DM (1985) Textbook of Small Animal Orthopaedics (Lip-pincott, Philadelphia).

66. Gouze E, et al. (2007) Transgene persistence and cell turnover in the diarthrodial joint:Implications for gene therapy of chronic joint diseases. Mol Ther 15(6):1114–1120.

67. Gouze E, et al. (2003) Lentiviral-mediated gene delivery to synovium: Potent intra-articular expression with amplification by inflammation. Mol Ther 7(4):460–466.

68. Gouze E, et al. (2002) In vivo gene delivery to synovium by lentiviral vectors.Mol Ther5(4):397–404.

69. Gersbach CA, Phillips JE, García AJ (2007) Genetic engineering for skeletal re-generative medicine. Annu Rev Biomed Eng 9:87–119.

70. Wiznerowicz M, Trono D (2003) Conditional suppression of cellular genes: Lentivirusvector-mediated drug-inducible RNA interference. J Virol 77(16):8957–8961.

71. Hotta A, et al. (2009) Isolation of human iPS cells using EOS lentiviral vectors to selectfor pluripotency. Nat Methods 6(5):370–376.

72. Szulc J, Wiznerowicz M, Sauvain MO, Trono D, Aebischer P (2006) A versatile tool forconditional gene expression and knockdown. Nat Methods 3(2):109–116.

73. Salmon P, Trono D (2007) Production and titration of lentiviral vectors. Curr ProtocHum Genet Chap 12:Unit 12.10.

74. Zhong Q, Inniss D, Kjoller K, Elings VB (1993) Fractured polymer/silica fiber surfacestudied by tapping mode atomic force microscopy. Surf Sci Lett 290(1–2):L688–L692.

75. Yaszemski MJ (2004) Tissue Engineering and Novel Delivery Systems (Marcel Dekker,New York).

76. Yavari SA, et al. (2013) Fatigue behavior of porous biomaterials manufactured usingselective laser melting. Mater Sci Eng C Mater Biol Appl 33(8):4849–4858.

77. Woessner JF, Jr. (1961) The determination of hydroxyproline in tissue and proteinsamples containing small proportions of this imino acid. Arch Biochem Biophys 93:440–447.

78. Farndale RW, Buttle DJ, Barrett AJ (1986) Improved quantitation and discriminationof sulphated glycosaminoglycans by use of dimethylmethylene blue. Biochim BiophysActa 883(2):173–177.

79. Pfaffl MW (2001) A new mathematical model for relative quantification in real-timeRT-PCR. Nucleic Acids Res 29(9):e45.

E806 | www.pnas.org/cgi/doi/10.1073/pnas.1321744111 Brunger et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

4, 2

020