REVIEWS
`
`Valerie S. Salazar, Laura W. Gamer and Vicki Rosen
`
`BMP signalling in skeletal
`development, disease and repair
`
`Abstract | Since the identification in 1988 of bone morphogenetic protein 2 (BMP2) as a potent
`inducer of bone and cartilage formation, BMP superfamily signalling has become one of the most
`heavily investigated topics in vertebrate skeletal biology. Whereas a large part of this research has
`focused on the roles of BMP2, BMP4 and BMP7 in the formation and repair of endochondral bone,
`a large number of BMP superfamily molecules have now been implicated in almost all aspects of
`bone, cartilage and joint biology. As modulating BMP signalling is currently a major therapeutic
`target, our rapidly expanding knowledge of how BMP superfamily signalling affects most tissue
`types of the skeletal system creates enormous potential to translate basic research findings into
`successful clinical therapies that improve bone mass or quality, ameliorate diseases of skeletal
`overgrowth, and repair damage to bone and joints. This Review examines the genetic evidence
`implicating BMP superfamily signalling in vertebrate bone and joint development, discusses a
`selection of human skeletal disorders associated with altered BMP signalling and summarizes the
`status of modulating the BMP pathway as a therapeutic target for skeletal trauma and disease.
`
`The human skeleton includes over 200 bones and 340
`joints, as well as an intricate network of tendons, lig-
`aments and cartilage. During development and post-
`natal life, bone and joint health is profoundly affected
`by genetics and environmental factors such as nutrition
`and exercise. Unsurprisingly, the skeletal system is a
`major site of human disease. As the name implies, bone
`morphogenetic proteins (BMPs) were originally discov-
`ered by their ability to induce new bone formation1–4;
`accordingly, recombinant human BMPs have been
`exploited as osteoinductive agents to repair bone defects
`in clinical settings5. However, our current understanding
`of BMP superfamily molecules further establishes these
`signals as mediators of normal skeletogenesis as well as
`the underlying aetiology of several debilitating skeletal
`pathologies including fibrodysplasia ossificans progres-
`siva (FOP)6, Marfan syndrome7, Loeys–Dietz syndrome8
`and osteoarthritis9,10. In this Review, we describe BMP
`superfamily signalling in the context of skeletal devel-
`opment and joint morphogenesis, with the premise
`that the pathway is poised as a promising therapeutic
`target for treating skeletal trauma and diseases beyond
`bone repair. We open with a historical account of how
`BMPs were discovered, present a phylogenetic analysis
`of key molecules in the BMP signalling pathway and
`summarize fundamental BMP family signalling mech-
`anisms in vertebrates. We then discuss developmental
`skeletogenesis, focusing on the genetic evidence from
`
`mice and humans supporting a decisive role for the
`BMP pathway in skeletal development and disease and
`conclude by summarizing nodes of the pathway that are
`currently or potentially accessible as therapeutic targets
`for clinical medicine.
`
`Historical perspective
`
`Marshall Urist practiced orthopaedic surgery and con-
`ducted scientific research at the University of California,
`Los Angeles Medical School, USA, for nearly half of the
`twentieth century. At the time of his practice, the thera-
`peutic potential of applying shavings from healthy bone
`to heal major bone defects had long been recognized
`in orthopaedic settings11, although the mechanism for
`repair was unknown. In the 1960s, Urist identified an
`interfibrillar protein complex1 in demineralized rabbit
`bone able to induce calcified cartilage from minced mus-
`cles in vitro2 and bone formation at nonskeletal sites in
`rats3. Urist named this factor bone morphogenetic pro-
`tein. Although initially ignored, Urist’s work was eventu-
`ally reproduced and published by Nobel laureate Charles
`Huggins12, sparking intense efforts to identify and purify
`bone morphogenetic protein. The challenging purifica-
`tion of BMPs from bone matrix took many years and,
`in the end, researchers were unable to purify a homo-
`geneous BMP13,14. Human BMPs were finally cloned
`in 1988, and it was then realized that the BMP activity
`Urist first identified consisted of multiple individual
`
` VOLUME 12 | APRIL 2016 | 203
`
`(cid:49)(cid:35)(cid:50)(cid:35)(cid:49)(cid:53)(cid:35)(cid:34)(cid:421)
`
`(cid:17)(cid:52)(cid:32)(cid:43)(cid:40)(cid:50)(cid:39)(cid:35)(cid:49)(cid:50)
`(cid:613)
`
`(cid:12)(cid:40)(cid:44)(cid:40)(cid:51)(cid:35)(cid:34)(cid:421)
`(cid:613)
`
`(cid:1)(cid:43)(cid:43)
`(cid:613)
`
`(cid:49)(cid:40)(cid:37)(cid:39)(cid:51)(cid:50)
`(cid:613)
`
`(cid:400)(cid:398)(cid:399)(cid:406)
`(cid:613)
`
`(cid:13)(cid:31)(cid:33)(cid:44)(cid:40)(cid:43)(cid:43)(cid:31)(cid:45)
`(cid:613)
`
`(cid:479)(cid:613)
`
`Department of Developmental
`Biology, Harvard School
`of Dental Medicine,
`188 Longwood Avenue,
`Boston, Massachusetts
`02115, USA.
`
`Correspondence to V.R.
`vicki_rosen@
`hsdm.harvard.edu
`
`doi:10.1038/nrendo.2016.12
`
`Published online 19 Feb 2016
`
`NATURE REVIEWS | ENDOCRINOLOGY
`
`Lassen - Exhibit 1025, p. 1
`
`

`

`REVIEWS
`Key points
`• Phylogenetic analysis indicates that the bone morphogenetic protein (BMP) pathway
`is ancient and highly conserved across the animal kingdom
`• Gene duplication and divergence has created a diverse matrix of BMP ligand–
`receptor pairs that achieve sophisticated control of signalling through variable
`activity profiles and functional redundancy
`• Members of the BMP superfamily affect almost all aspects of bone, cartilage and
`• Altered BMP signalling is a major underlying cause of human skeletal disorders
`• Modulation of BMP signalling is emerging as a promising therapeutic strategy for
`improving bone mass and bone quality, ameliorating diseases of skeletal overgrowth
`and repairing damage to bones and joints
`
`(cid:76)(cid:81)(cid:75)(cid:80)(cid:86)(cid:124)(cid:68)(cid:75)(cid:81)(cid:78)(cid:81)(cid:73)(cid:91)
`
`related gene products4. Since that time, recombinant
`human BMP2 and BMP7 have been used in ortho paedic
`applications, where enhancing bone repair by activat-
`ing BMP signalling has become standard practice in
`treating non-union fractures, spinal surgeries and oral
`maxillofacial procedures5,15.
`
`Signalling mechanisms of the BMP pathway
`Essential components
`
`The BMP pathway is at least 1.2–1.4 billion years old,
`emerging in the evolutionary record with multi-cellu-
`lar animals16. Consistent with the role of transmitting
`information between cells, BMP signalling coordinates
`many developmental processes including body axis
`determination17, germ layer specification, tissue mor-
`phogenesis and cell-fate specification. Phylogenetic
`analysis reveals that protein sequences for ligands,
`receptors and SMADs of the BMP pathway are highly
`conserved across distant species in the animal kingdom
`such as mice, flies and worms18. Full-length protein
`sequences of human and fly orthologues also exhibit
`considerable similarity19,20 (FIG. 1), and this evolutionary
`conservation is particularly striking in the amino acid
`sequence of active mature signalling proteins produced
`after post-translational processing of prepeptide and
`propeptide domains21,22. In fact, striking examples of
`cross-species activity have been documented in which fly
`orthologues of BMP2 and/or BMP4 and BMP7 (Dpp and
`Gbb, respectively) can successfully induce endochondral
`bone formation when implanted in mammals23.
`At the most empirical level, BMP signalling relies
`on a source of secreted ligands and a target cell express-
`ing type I and type II BMP receptors. Ligand-binding
`events activate a complex array of downstream intra-
`cellular mediators including, most notably, the canon-
`ical SMAD pathway 21,24. Although weak transcription
`factors on their own, SMADs are potent regulators of
`gene expression via their ability to recruit chromatin-
`remodelling machinery and tissue-specific transcrip-
`tion factors to the genomic landscape25–28. Despite the
`seemingly simple nature of this signal transduction cas-
`cade, >30 secreted ligands, seven type I receptors, five
`type II receptors and eight SMADs have been identi-
`fied in humans. Gene expression programs initiated by
`BMP superfamily signals are therefore highly diverse
`and tailored by factors such as ligand identity and
`
`204 | APRIL 2016 | VOLUME 12
`
`concentration, the type I and type II receptor profile
`on the target cell, the repertoire of tissue-specific tran-
`scription factors that define which SMAD-dependent
`gene targets are regulated27 and the status of the epi-
`genetic landscape26. The number of genes regulated by
`any single BMP superfamily ligand can therefore be
`either very low or very high, permitting the system to
`accommodate distinct transcriptional requirements of
`both quiescent stem cells and differentiated cells with
`complex physiological activity.
`
`Ligands. This extensive ligand family includes BMPs,
`growth/differentiation factors (GDFs), transform-
`ing growth factors (TGFs), activins, Nodal, and anti-
`Müllerian hormone (AMH). Collectively, these mol-
`ecules are typically referred to as the TGF-β superfam-
`ily, although this terminology is based on the order of
`their discovery as opposed to phylogenetic analysis,
`which identifies BMP2 as the founding family mem-
`ber22. Whereas BMPs were discovered as a result of their
`osteo inductive qualities, activins and inhibins were orig-
`inally discovered by their opposing control of follicle-
`stimulating hormone production29, and TGF-βs were first
`reported as secreted factors that conferred malignancy
`on cells via autocrine induction30. Aside from sequence
`similarity, these ligands can be further organized into
`three groups on the basis of preferred receptor usage and
`SMAD1/5/8 versus SMAD2/3 signalling activity (FIG. 2).
`In general, ligands are initially translated as prepropro-
`teins, which facilitates targeting to the secretory pathway
`for proteo lytic cleavage and enables noncovalent assem-
`bly into fully active dimers upon secretion via conserved
`cystine knot motifs31,32. Except for Nodal, proteolytic acti-
`vation and dimerization is essential for signalling33. Both
`homo dimers and heterodimers exhibit biological activ-
`ity34 that is well typified by activins, which can form active
`homodimers or heterodimers of activin βA, activin βB,
`activin βC or activin βE subunits. Activins can alterna-
`tively dimerize with inhibin α, and although this dimer
`retains receptor-binding activity, it constitutes a non-
`signal-generating ligand. Most ligands exhibit local
`para crine activity, although some BMPs, activins,
`TGF-βs and GDFs are thought to circulate and exert
`systemic effects35–39.
`
`Receptors. Type I and type II BMP receptors are the only
`known class of transmembrane cell surface receptors in
`humans with serine/threonine kinase activity. A mature
`receptor signalling complex requires one ligand dimer,
`two type I receptors and two type II receptors (FIG. 2).
`Several mechanisms are utilized to form activated
`ligand:receptor complexes, which affect the specific-
`ity of ligand-receptor pairing 40,41 and competition by
`distinct ligands for shared receptors42. Whereas type II
`receptors are constitutively active, type I receptors
`encode a Gly/Ser-rich domain that must be phospho-
`rylated by a type II receptor to activate intrinsic kinase
`activity (FIG. 2) and subsequently stimulate the recruit-
`ment and phosphorylation of the essential downstream
`pathway mediators known as receptor-activated SMADs
`(R-SMADs)43 (FIG. 2).
`
`www.nature.com/nrendo
`
`(cid:49)(cid:35)(cid:50)(cid:35)(cid:49)(cid:53)(cid:35)(cid:34)(cid:421)
`
`(cid:49)(cid:40)(cid:37)(cid:39)(cid:51)(cid:50)
`(cid:613)
`
`(cid:17)(cid:52)(cid:32)(cid:43)(cid:40)(cid:50)(cid:39)(cid:35)(cid:49)(cid:50)
`(cid:613)
`
`(cid:12)(cid:40)(cid:44)(cid:40)(cid:51)(cid:35)(cid:34)(cid:421)
`(cid:613)
`
`(cid:1)(cid:43)(cid:43)
`(cid:613)
`
`(cid:400)(cid:398)(cid:399)(cid:406)
`(cid:613)
`
`(cid:13)(cid:31)(cid:33)(cid:44)(cid:40)(cid:43)(cid:43)(cid:31)(cid:45)
`(cid:613)
`
`(cid:479)(cid:613)
`
`Lassen - Exhibit 1025, p. 2
`
`

`

`a
`
`LEFTYA
`LEFTYB
`
`2.0
`
`Ligands
`
`b
`
`REVIEWS
`Type I receptors
`Decapentaplegic
`BMP2BMP4BMP9/GDF2
`Thickveins
`ALK3/BMPR1A
`ALK6/BMPR1BSaxophone
`BMP10GDF5/BMP14
`ALK1/ACVRL1
`GDF7/BMP12
`ALK2/ACVR1
`GDF6/BMP13
`GDF1
`Baboon
`GDF3 GDF15GDF9
`ALK4/ACVR1B
`ALK7/ACVR1C
`ALK5/TGFBR1
`BMP15
`PuntACVR2B
`NODALScrew
`Type II receptors
`Glass bottom boat
`ACVR2C
`BMP8A
`TGFBR2
`BMP8B/OP2
`Wishful thinking
`BMP6BMP5
`BMPR2
`AMHR2
`BMP7/OP1
`AMH/MIS
`TGF-β1
`TGF-β2
`TGF-β3Activin β
`Co-SMADS
`Medea
`INHβA/ACTA
`SMAD4
`INHβB/ACTB
`Mad
`BMP/GDF
`INHβC/ACTC
`SMAD1
`R-Smads
`INHβE/ACTE
`SMAD5
`Dawdle
`GDF11/BMP11
`SMAD8
`GDF8/MSTNMyoglianin
`SMAD2
`TGF-β/Activin
`SMAD3
`R-SMADS
`BMP3BMP3B/GDF10
`Smox
`Dad
`I-SMADS
`Maverick
`SMAD6
`INHα
`SMAD7
`0.5
`Figure 1 | Phylogenetic analysis of BMP superfamily molecules. Protein sequences from flies and humans were
`aligned to assess evolutionary relationships between bone morphogenetic protein (BMP) superfamily molecules.
`Human proteins are designated in all capital letters; only the first letter of fly proteins is capitalized. For ligands,
`preproprotein sequences were used for alignments. The longest known isoform of each molecule was used when
`applicable. Molecules are grouped into a | ligands, b |(cid:2)(cid:86)(cid:91)(cid:82)(cid:71)(cid:124)(cid:43)(cid:2)(cid:67)(cid:80)(cid:70)(cid:2)(cid:86)(cid:91)(cid:82)(cid:71)(cid:124)(cid:43)(cid:43)(cid:2)(cid:84)(cid:71)(cid:69)(cid:71)(cid:82)(cid:86)(cid:81)(cid:84)(cid:85)(cid:2)(cid:67)(cid:80)(cid:70)(cid:2)c | SMADs. Branch lengths are
`drawn to scale; the scale bar indicates to the number of amino acid substitutions per site between two compared
`sequences. ACV, activin; ACVR, activin receptor; ALK, activin receptor-like kinase; AMH, anti-Müllerian hormone;
`(cid:35)(cid:47)(cid:42)(cid:52)(cid:20)(cid:14)(cid:2)(cid:35)(cid:47)(cid:42)(cid:2)(cid:84)(cid:71)(cid:69)(cid:71)(cid:82)(cid:86)(cid:81)(cid:84)(cid:15)(cid:20)(cid:29)(cid:2)(cid:36)(cid:47)(cid:50)(cid:52)(cid:14)(cid:2)(cid:36)(cid:47)(cid:50)(cid:2)(cid:84)(cid:71)(cid:69)(cid:71)(cid:82)(cid:86)(cid:81)(cid:84)(cid:29)(cid:2)(cid:41)(cid:38)(cid:40)(cid:14)(cid:2)(cid:73)(cid:84)(cid:81)(cid:89)(cid:86)(cid:74)(cid:17)(cid:70)(cid:75)(cid:72)(cid:72)(cid:71)(cid:84)(cid:71)(cid:80)(cid:86)(cid:75)(cid:67)(cid:86)(cid:75)(cid:81)(cid:80)(cid:2)(cid:72)(cid:67)(cid:69)(cid:86)(cid:81)(cid:84)(cid:29)(cid:2)(cid:43)(cid:48)(cid:42)β, inhibin β; co-SMAD,
`(cid:69)(cid:81)(cid:79)(cid:79)(cid:81)(cid:80)(cid:2)(cid:53)(cid:47)(cid:35)(cid:38)(cid:29)(cid:2)(cid:43)(cid:15)(cid:53)(cid:47)(cid:35)(cid:38)(cid:14)(cid:2)(cid:75)(cid:80)(cid:74)(cid:75)(cid:68)(cid:75)(cid:86)(cid:81)(cid:84)(cid:91)(cid:2)(cid:53)(cid:47)(cid:35)(cid:38)(cid:29)(cid:2)(cid:52)(cid:15)(cid:53)(cid:47)(cid:35)(cid:38)(cid:14)(cid:2)(cid:84)(cid:71)(cid:69)(cid:71)(cid:82)(cid:86)(cid:81)(cid:84)(cid:15)(cid:67)(cid:69)(cid:86)(cid:75)(cid:88)(cid:67)(cid:86)(cid:71)(cid:70)(cid:2)(cid:53)(cid:47)(cid:35)(cid:38)(cid:29)(cid:2)(cid:54)(cid:41)(cid:40)(cid:15)β(cid:14)(cid:2)(cid:86)(cid:84)(cid:67)(cid:80)(cid:85)(cid:72)(cid:81)(cid:84)(cid:79)(cid:75)(cid:80)(cid:73)(cid:2)(cid:73)(cid:84)(cid:81)(cid:89)(cid:86)(cid:74)(cid:2)(cid:72)(cid:67)(cid:69)(cid:86)(cid:81)(cid:84)(cid:124)β;
`TGFBR, TGF-β receptor.
`
`0.5
`
`c
`
`SMADs. SMADs are homologues of Drosophila melano-
`gaster Mad proteins (mothers against decapentaplegic)
`and Caenorhabditis elegans SMA proteins (small body
`size), and encode cytoplasmic proteins required for
`responsiveness to BMP superfamily ligands44. SMADs are
`modular in structure, with many highly conserved motifs.
`Among these, the N-terminal MH1 domain contains a
`sequence-selective45 DNA-binding motif 46 and nuclear
`localization signal47 essential for SMAD-dependent
`effects on gene expression in response to ligand-binding
`events48. A conserved L3 loop motif mediates direct bind-
`ing between R-SMADs and activated receptors and deter-
`mines SMAD1/5 versus SMAD2/3 pairing specificity49.
`A series of serine/threonine residues in the linker domain
`
`NATURE REVIEWS | ENDOCRINOLOGY
`
`enables SMADs to receive regulatory inputs from a vari-
`ety of intracellular kinase cascades including inhibitory
`regulation by mitogen-activated protein kinase (MAPK)50
`and glycogen synthase kinase 3β (GSK3β)51,52, facili-
`tating integration of BMP signals with other pathways
`including fibroblast growth factor (FGF) and WNT. The
`C-terminus of SMADs contains serine/threonine (Ser/
`Thr) residues directly phosphorylated by type I receptors,
`as well as protein–protein interaction domains that medi-
`ate R-SMAD/SMAD4 trimerization53 (FIG. 2). Activated
`SMAD complexes translocate to the nucleus where they
`target the genome via consensus SMAD-binding motifs,
`integrate with tissue-specific transcription factors and
`recruit chromatin remodelling machinery25–28 (FIG. 2).
`
` VOLUME 12 | APRIL 2016 | 205
`
`(cid:49)(cid:35)(cid:50)(cid:35)(cid:49)(cid:53)(cid:35)(cid:34)(cid:421)
`
`(cid:17)(cid:52)(cid:32)(cid:43)(cid:40)(cid:50)(cid:39)(cid:35)(cid:49)(cid:50)
`(cid:613)
`
`(cid:12)(cid:40)(cid:44)(cid:40)(cid:51)(cid:35)(cid:34)(cid:421)
`(cid:613)
`
`(cid:1)(cid:43)(cid:43)
`(cid:613)
`
`(cid:49)(cid:40)(cid:37)(cid:39)(cid:51)(cid:50)
`(cid:613)
`
`(cid:400)(cid:398)(cid:399)(cid:406)
`(cid:613)
`
`(cid:13)(cid:31)(cid:33)(cid:44)(cid:40)(cid:43)(cid:43)(cid:31)(cid:45)
`(cid:613)
`
`(cid:479)(cid:613)
`
`Lassen - Exhibit 1025, p. 3
`
`

`

`Activin
`• Activin βA
`• Activin βB
`e.g. Noggin
`• Activin βC
`• Activin βE
`• Myostatin
`• GDF1/3/11/15
`• BMP3
`• Activin β/Inhibin α
`• LEFTYA/B
`• ACVR2A
`• ACVR2A
`• ACVR2B
`• ACVR2B
`• BMPR2
`• AMHR2
`
`TGF-β
`e.g. Follistatin,
`• TGF-β1
`proprotein
`• TGF-β2
`latency
`• TGF-β3
`• ALK4 (βA)
`• ALK7 (βA/B, βB/B, Nodal)
`• ALK5 (MSTN)
`
`ALK5
`
`Proprotein
`latency
`TGFBR2
`
`REVIEWS
`BMP/GDF
`• BMP2/4/7
`• BMP5/6/7/8A/8B
`• BMP9/10/15
`• GDF5/6/7
`• AMH
`• MIS
`• ALK1
`• ALK2
`• ALK3
`• ALK6
`
`Ligands
`
`Receptor
`
`Recruitment
`Phosphorylation
`
`Trimerization
`
`expression
`Gene
`
`P
`
`PP
`P
`
`P P
`SMAD4
`
`SMAD1/5/8
`P
`
`PP
`
`PP
`
`P
`P
`
`SMAD2/3
`P
`
`P P
`
`P
`
`PP
`SMAD2/3
`P P
`
`P
`P
`P
`P
`SMAD1/5/8 responsive genes
`SMAD2/3 responsive genes
`• SMAD-dependent recruitment of chromatin remodelling factors (e.g. SWI/SNF) or histone modifying enzymes (e.g. p300, CBP)
`• (cid:52)(cid:71)(cid:69)(cid:84)(cid:87)(cid:75)(cid:86)(cid:79)(cid:71)(cid:80)(cid:86)(cid:2)(cid:67)(cid:80)(cid:70)(cid:17)(cid:81)(cid:84)(cid:2)(cid:75)(cid:80)(cid:86)(cid:71)(cid:73)(cid:84)(cid:67)(cid:86)(cid:75)(cid:81)(cid:80)(cid:2)(cid:89)(cid:75)(cid:86)(cid:74)(cid:2)(cid:69)(cid:71)(cid:78)(cid:78)(cid:15)(cid:86)(cid:91)(cid:82)(cid:71)(cid:2)(cid:85)(cid:82)(cid:71)(cid:69)(cid:75)(cid:558)(cid:69)(cid:2)(cid:86)(cid:84)(cid:67)(cid:80)(cid:85)(cid:69)(cid:84)(cid:75)(cid:82)(cid:86)(cid:75)(cid:81)(cid:80)(cid:2)(cid:72)(cid:67)(cid:69)(cid:86)(cid:81)(cid:84)(cid:85)(cid:2)(cid:10)(cid:71)(cid:16)(cid:73)(cid:16)(cid:2)(cid:52)(cid:55)(cid:48)(cid:58)(cid:20)(cid:11)
`Figure 2 | Fundamental mechanisms of canonical BMP superfamily signalling. Over 30 bone morphogenetic
`protein (BMP) superfamily ligands have been discovered in humans. Most are secreted as mature disulfide-linked dimers,
`with the exception of TGF-β1, TGF-β2 and TGF-β3, which can be secreted in a latent form and require proteolytic
`receptors recruit and phosphorylate pathway-specific R-SMADs (SMAD1, SMAD5 and SMAD8 (blue pathway), and
`SMAD2 and SMAD3 (orange pathway)), which can form trimers with SMAD4 and translocate to the nucleus. SMADs
`machinery and integration with tissue-specific transcription factors. SMAD8 is also known as SMAD9. The pathway can
`be antagonized by many mechanisms including neutralization of ligands by secreted traps such as noggin or follistatin,
`secretion of latent ligands bound to their propeptides, or via titration of receptors by nonsignalling ligands such
`(cid:67)(cid:85)(cid:124)(cid:36)(cid:47)(cid:50)(cid:21)(cid:14)(cid:2)(cid:67)(cid:69)(cid:86)(cid:75)(cid:88)(cid:75)(cid:80)(cid:2)β/inhibin α dimers or LEFTY monomers. ACVR, activin receptor; ALK, activin receptor-like kinase;
`Receptor/SMAD usage profiles
`
`(cid:67)(cid:69)(cid:86)(cid:75)(cid:88)(cid:67)(cid:86)(cid:75)(cid:81)(cid:80)(cid:16)(cid:2)(cid:36)(cid:47)(cid:50)(cid:85)(cid:2)(cid:85)(cid:75)(cid:73)(cid:80)(cid:67)(cid:78)(cid:2)(cid:86)(cid:74)(cid:84)(cid:81)(cid:87)(cid:73)(cid:74)(cid:2)(cid:67)(cid:2)(cid:79)(cid:87)(cid:78)(cid:86)(cid:75)(cid:79)(cid:71)(cid:84)(cid:75)(cid:69)(cid:2)(cid:69)(cid:71)(cid:78)(cid:78)(cid:2)(cid:85)(cid:87)(cid:84)(cid:72)(cid:67)(cid:69)(cid:71)(cid:2)(cid:69)(cid:81)(cid:79)(cid:82)(cid:78)(cid:71)(cid:90)(cid:2)(cid:69)(cid:81)(cid:80)(cid:85)(cid:75)(cid:85)(cid:86)(cid:75)(cid:80)(cid:73)(cid:2)(cid:81)(cid:72)(cid:2)(cid:86)(cid:89)(cid:81)(cid:2)(cid:86)(cid:91)(cid:82)(cid:71)(cid:124)(cid:43)(cid:2)(cid:84)(cid:71)(cid:69)(cid:71)(cid:82)(cid:86)(cid:81)(cid:84)(cid:85)(cid:2)(cid:67)(cid:80)(cid:70)(cid:2)(cid:86)(cid:89)(cid:81)(cid:2)(cid:86)(cid:91)(cid:82)(cid:71)(cid:124)(cid:43)(cid:43)(cid:2)
`(cid:84)(cid:71)(cid:69)(cid:71)(cid:82)(cid:86)(cid:81)(cid:84)(cid:85)(cid:16)(cid:2)(cid:54)(cid:91)(cid:82)(cid:71)(cid:124)(cid:43)(cid:2)(cid:67)(cid:80)(cid:70)(cid:2)(cid:86)(cid:91)(cid:82)(cid:71)(cid:124)(cid:43)(cid:43)(cid:2)(cid:36)(cid:47)(cid:50)(cid:2)(cid:84)(cid:71)(cid:69)(cid:71)(cid:82)(cid:86)(cid:81)(cid:84)(cid:85)(cid:2)(cid:67)(cid:84)(cid:71)(cid:2)(cid:85)(cid:75)(cid:80)(cid:73)(cid:78)(cid:71)(cid:2)(cid:82)(cid:67)(cid:85)(cid:85)(cid:2)(cid:86)(cid:84)(cid:67)(cid:80)(cid:85)(cid:79)(cid:71)(cid:79)(cid:68)(cid:84)(cid:67)(cid:80)(cid:71)(cid:2)(cid:82)(cid:84)(cid:81)(cid:86)(cid:71)(cid:75)(cid:80)(cid:85)(cid:2)(cid:89)(cid:75)(cid:86)(cid:74)(cid:2)(cid:67)(cid:80)(cid:2)(cid:75)(cid:80)(cid:86)(cid:84)(cid:67)(cid:69)(cid:71)(cid:78)(cid:78)(cid:87)(cid:78)(cid:67)(cid:84)(cid:2)(cid:85)(cid:71)(cid:84)(cid:75)(cid:80)(cid:71)(cid:17)
`(cid:86)(cid:74)(cid:84)(cid:71)(cid:81)(cid:80)(cid:75)(cid:80)(cid:71)(cid:2)(cid:77)(cid:75)(cid:80)(cid:67)(cid:85)(cid:71)(cid:2)(cid:70)(cid:81)(cid:79)(cid:67)(cid:75)(cid:80)(cid:16)(cid:2)(cid:35)(cid:72)(cid:86)(cid:71)(cid:84)(cid:2)(cid:78)(cid:75)(cid:73)(cid:67)(cid:80)(cid:70)(cid:2)(cid:68)(cid:75)(cid:80)(cid:70)(cid:75)(cid:80)(cid:73)(cid:14)(cid:2)(cid:86)(cid:91)(cid:82)(cid:71)(cid:124)(cid:43)(cid:43)(cid:2)(cid:84)(cid:71)(cid:69)(cid:71)(cid:82)(cid:86)(cid:81)(cid:84)(cid:85)(cid:2)(cid:82)(cid:74)(cid:81)(cid:85)(cid:82)(cid:74)(cid:81)(cid:84)(cid:91)(cid:78)(cid:67)(cid:86)(cid:71)(cid:2)(cid:10)(cid:50)(cid:11)(cid:2)(cid:86)(cid:74)(cid:71)(cid:2)(cid:86)(cid:91)(cid:82)(cid:71)(cid:124)(cid:43)(cid:2)(cid:84)(cid:71)(cid:69)(cid:71)(cid:82)(cid:86)(cid:81)(cid:84)(cid:85)(cid:16)(cid:2)(cid:35)(cid:69)(cid:86)(cid:75)(cid:88)(cid:67)(cid:86)(cid:71)(cid:70)(cid:2)(cid:86)(cid:91)(cid:82)(cid:71)(cid:124)(cid:43)(cid:2)
`
`(cid:74)(cid:67)(cid:88)(cid:71)(cid:2)(cid:75)(cid:80)(cid:86)(cid:84)(cid:75)(cid:80)(cid:85)(cid:75)(cid:69)(cid:2)(cid:38)(cid:48)(cid:35)(cid:15)(cid:68)(cid:75)(cid:80)(cid:70)(cid:75)(cid:80)(cid:73)(cid:2)(cid:67)(cid:69)(cid:86)(cid:75)(cid:88)(cid:75)(cid:86)(cid:91)(cid:2)(cid:67)(cid:80)(cid:70)(cid:2)(cid:67)(cid:84)(cid:71)(cid:2)(cid:67)(cid:68)(cid:78)(cid:71)(cid:2)(cid:86)(cid:81)(cid:2)(cid:84)(cid:71)(cid:73)(cid:87)(cid:78)(cid:67)(cid:86)(cid:71)(cid:2)(cid:73)(cid:71)(cid:80)(cid:71)(cid:2)(cid:71)(cid:90)(cid:82)(cid:84)(cid:71)(cid:85)(cid:85)(cid:75)(cid:81)(cid:80)(cid:2)(cid:68)(cid:91)(cid:2)(cid:84)(cid:71)(cid:69)(cid:84)(cid:87)(cid:75)(cid:86)(cid:79)(cid:71)(cid:80)(cid:86)(cid:2)(cid:81)(cid:72)(cid:2)(cid:69)(cid:74)(cid:84)(cid:81)(cid:79)(cid:67)(cid:86)(cid:75)(cid:80)(cid:15)(cid:84)(cid:71)(cid:79)(cid:81)(cid:70)(cid:71)(cid:78)(cid:78)(cid:75)(cid:80)(cid:73)(cid:2)
`
`(cid:35)(cid:47)(cid:42)(cid:14)(cid:124)(cid:67)(cid:80)(cid:86)(cid:75)(cid:15)(cid:47)(cid:216)(cid:78)(cid:78)(cid:71)(cid:84)(cid:75)(cid:67)(cid:80)(cid:2)(cid:74)(cid:81)(cid:84)(cid:79)(cid:81)(cid:80)(cid:71)(cid:29)(cid:2)(cid:35)(cid:47)(cid:42)(cid:52)(cid:20)(cid:14)(cid:2)(cid:35)(cid:47)(cid:42)(cid:2)(cid:84)(cid:71)(cid:69)(cid:71)(cid:82)(cid:86)(cid:81)(cid:84)(cid:2)(cid:20)(cid:29)(cid:2)(cid:36)(cid:47)(cid:50)(cid:52)(cid:14)(cid:2)(cid:36)(cid:47)(cid:50)(cid:2)(cid:84)(cid:71)(cid:69)(cid:71)(cid:82)(cid:86)(cid:81)(cid:84)(cid:29)(cid:2)(cid:41)(cid:38)(cid:40)(cid:14)(cid:2)(cid:73)(cid:84)(cid:81)(cid:89)(cid:86)(cid:74)(cid:17)(cid:70)(cid:75)(cid:72)(cid:72)(cid:71)(cid:84)(cid:71)(cid:80)(cid:86)(cid:75)(cid:67)(cid:86)(cid:75)(cid:81)(cid:80)(cid:2)(cid:72)(cid:67)(cid:69)(cid:86)(cid:81)(cid:84)(cid:29)(cid:2)
`(cid:54)(cid:41)(cid:40)(cid:14)(cid:124)(cid:86)(cid:84)(cid:67)(cid:80)(cid:85)(cid:72)(cid:81)(cid:84)(cid:79)(cid:75)(cid:80)(cid:73)(cid:2)(cid:73)(cid:84)(cid:81)(cid:89)(cid:86)(cid:74)(cid:2)(cid:72)(cid:67)(cid:69)(cid:86)(cid:81)(cid:84)(cid:29)(cid:2)(cid:54)(cid:41)(cid:40)(cid:36)(cid:52)(cid:14)(cid:2)(cid:54)(cid:41)(cid:40)(cid:15)β(cid:124)(cid:84)(cid:71)(cid:69)(cid:71)(cid:82)(cid:86)(cid:81)(cid:84)(cid:16)
`
`Ligand-receptor pairing specificity (reviewed elsewhere54)
`is summarized in FIG. 2. TGF-βs use the type I (ALK5)
`and type II (TGFBR2) TGF-β receptors to activate the
`SMAD2/3 pathway (FIG. 2, orange pathway). By contrast,
`BMPs and GDFs exhibit broad receptor usage patterns to
`activate the SMAD1/5/8 pathway (FIG. 2, blue pathway).
`Ser/Thr-protein kinase receptor R3 (ALK1), activin recep-
`tor type-1 (ALK2), BMP receptor type-1A (ALK3) and
`BMP receptor type-1B (ALK6) can all function as type I
`
`206 | APRIL 2016 | VOLUME 12
`
`BMP and GDF receptors; BMP receptor type-2 (BMPR2),
`activin receptor type-2A (ACVR2A) and ACVR2B serve
`as type II receptors. Nodal, GDF8 and GDF11 activate
`SMAD2/3 via ALK4, ALK5, or ALK7 type  I recep-
`tors and the ACVR2A and ACVR2B type II receptors.
`Activins utilize ALK4 (βA/βA) and ALK7 (βA/βB and
`βB/βB) for type I receptors, and ACVR2A and ACVR2B
`for type  II receptors (FIG.  2). Importantly, activins
`can also bind to ALK2, but these complexes do not
`normally signal55.
`
`www.nature.com/nrendo
`
`(cid:49)(cid:35)(cid:50)(cid:35)(cid:49)(cid:53)(cid:35)(cid:34)(cid:421)
`
`(cid:49)(cid:40)(cid:37)(cid:39)(cid:51)(cid:50)
`(cid:613)
`
`(cid:400)(cid:398)(cid:399)(cid:406)
`(cid:613)
`
`(cid:13)(cid:31)(cid:33)(cid:44)(cid:40)(cid:43)(cid:43)(cid:31)(cid:45)
`(cid:613)
`
`(cid:17)(cid:52)(cid:32)(cid:43)(cid:40)(cid:50)(cid:39)(cid:35)(cid:49)(cid:50)
`(cid:613)
`
`(cid:12)(cid:40)(cid:44)(cid:40)(cid:51)(cid:35)(cid:34)(cid:421)
`(cid:613)
`
`(cid:1)(cid:43)(cid:43)
`(cid:613)
`
`(cid:479)(cid:613)
`
`Lassen - Exhibit 1025, p. 4
`
`

`

`Pathway antagonism
`
`The BMP pathway is subject to many levels of regula-
`tory activity, including propeptide latency, antagonism
`by secreted receptors and ligands, receptor traffick-
`ing and negative intracellular feedback by SMAD6/7
`(REFS 54,56,57). As examples, noggin58, gremlin59 and
`follistatin60 are secreted antagonists that are expressed
`in skeletal tissues and bind to distinct subsets of BMPs,
`GDFs and/or activins to titrate active ligands out of the
`extracellular environment 61,62 (FIG. 2). GDF8, GDF11
`and TGF-βs can be secreted noncovalently attached to
`their prodomain, requiring additional processing to be
`activated from latency63 (FIG. 2). Receptor availability can
`be regulated by BMP3 (REF. 64), LEFTYA/B mono mers65
`and activin β/inhibin α heterodimers, which occupy but
`do not activate ACVR2A and/or ACVR2B (FIG. 2). This
`regulation dampens activin as well as BMP signalling, as
`ACVR2A and ACVR2B are shared receptors for these
`two ligand subtypes. Inside the cell, BMP and TGF-β
`signalling initiate negative feedback by transcriptional
`upregulation of SMAD6 and SMAD7, which are also
`known as the inhibitory SMADs (I-SMADs). By inter-
`acting with cytoplasmic domains of cell surface recep-
`tors, SMAD6 can sterically interfere with R-SMAD
`phosphorylation and recruit E3 ubiquitin ligases to mark
`signalling machinery for degradation66–70. Although long
`considered an intracellular signalling mediator of the
`canonical BMP pathway, new evidence suggests that
`SMAD8 (also known as SMAD9) is hypermorphic rela-
`tive to SMAD1 and SMAD5, and so attenuates canonical
`BMP signalling71. Additional details on signalling and
`regulatory mechanisms can be found elsewhere21,24,56,72.
`
`Genetics of the BMP pathway
`Developmental skeletogenesis
`
`A skeleton with articulated joints appeared >400 mil-
`lion years ago in Cambrian bony fishes. In modern day
`mammals, the axial skeleton includes the skull, ossicles
`of the middle ear, hyoid bone, ribs, sternum and ver-
`tebrae. The appendicular skeleton comprises the pelvic
`and pectoral girdles and bones in the limbs. All bones
`are formed during development from three embryonic
`lineages: neural crest, paraxial mesoderm and lateral
`plate mesoderm. Some bones, such as those found in the
`skull, form by intramembranous ossification, in which
`migratory cells from the neural crest and paraxial meso-
`derm condense into sheet-like structures, differentiate
`into bone-forming cells called osteoblasts and produce
`mineralized tissue. Most bones, however, form by endo-
`chondral ossification, where a cartilage template pro-
`duced by chondrocytes is segmented by joints, populated
`by haemato poietic progenitors during a primary wave
`of vascularization, remodelled by monocyte-derived
`resorbing cells called osteoclasts, and finally converted
`into bone by osteoblasts. The development of endochon-
`dral bones, therefore, requires the coordination of sig-
`nals from several distinct cell types developing within
`the cartilage rudiment73 (FIG. 3).
`Before bone and joint formation, the mesenchy-
`mal progenitor pool in the emerging limb bud must
`first undergo considerable expansion and patterning74.
`
`NATURE REVIEWS | ENDOCRINOLOGY
`
`REVIEWS
`
`Lineage tracing analysis reveals that most, if not all,
`connective tissue cell types in the limb skeleton and
`some structures in the cranial vault arise from Prx1+
`progenitors75 (Prx1 is also known as Prrx1; FIG. 3a).
`Accordingly, Prx1–Cre75 has become a useful tool for
`conditionally ablating genes selectively in the limb bud
`mesenchyme (FIG. 4a), without the embryonic lethality
`resulting from global-deficiency, such as is the case with
`Bmp2 (REF. 76). Prx1+ progenitors are highly responsive
`to BMP signalling as limb bud outgrowth and patterning
`are disrupted in mice lacking Alk3 (REF. 77), and severely
`impaired in mice with Prx1–Cre-mediated single dele-
`tion of Smad4 or compound deletion of Alk2, Alk3 and
`Alk6 (REFS 78,79). However, limb bud outgrowth ensues
`normally in mice with single or compound deletions
`of Bmp2, Bmp4 and Bmp7 (REFS  80–83), and is only
`modestly impaired by global compound deletions of
`Gdf5 and Gdf6 or Gdf5 and Bmp5 (REFS 84,85), which
`suggests that BMP signals essential for limb bud out-
`growth are normally provided by multiple BMP-like
`ligands. Both genetic methods as well as classic ‘cut and
`paste’ experiments have further demonstrated that tis-
`sue nonautonomous BMP signals essential for limb bud
`patterning and digit specification emerge from ecto-
`dermal cells in the limb bud organizing centre known
`as the apical ectodermal ridge (AER)86. Expression of
`Msx2 is highly enriched in the AER87 and Msx2–Cre
`has been used to make selective compound deletion of
`Bmp2, Bmp4 and Bmp7 (REF. 88). Consistent with a cell
`autonomous role for BMP signalling in the mesenchyme,
`loss of Bmp2, Bmp4 and Bmp7 in the AER (Bmp2; Bmp4;
`Bmp7; Msx1–Cre) has no effect on limb bud outgrowth,
`but instead leads to loss of the AER and striking defects
`in digit patterning88. Digit patterning is also affected by
`mesoderm-derived BMP signalling as overexpression of
`gremlin in the limb bud mesenchyme mediates specifi-
`cation of too few versus too many digits, depending on
`the timing of induction89.
`Although the confluence-sensing mechanism
`remains unclear, the expanding progenitor pool even-
`tually reaches a critical mass and triggers condensation,
`which is required for entry of progenitors into endo-
`chondral differentiation programs and imparts shape
`on presumptive skeletal elements. As these cells become
`specified to the chondrogenic lineage, they upregulate
`Col2a1 and Agc1 (FIG. 3b), and begin depositing a cartil-
`age matrix. Cells at the innermost regions of the con-
`densation upregulate Col10a1 as they differentiate into
`hypertrophic chondrocytes (FIG. 3c).
`Chondrocyte hypertrophy at the centre of the mes-
`enc