Mmp17b Is Essential for Proper Neural Crest Cell Migration In Vivo

The extracellular matrix plays a critical role in neural crest (NC) cell migration. In this study, we characterize the contribution of the novel GPI-linked matrix metalloproteinase (MMP) zebrafish mmp17b. Mmp17b is expressed post-gastrulation in the developing NC. Morpholino inactivation of mmp17b function, or chemical inhibition of MMP activity results in aberrant NC cell migration with minimal change in NC proliferation or apoptosis. Intriguingly, a GPI anchored protein with metalloproteinase inhibitor properties, Reversion-inducing-Cysteine-rich protein with Kazal motifs (RECK), which has previously been implicated in NC development, is expressed in close apposition to NC cells expressing mmp17b, raising the possibility that these two gene products interact. Consistent with this possibility, embryos silenced for mmp17b show defective development of the dorsal root ganglia (DRG), a crest-derived structure affected in RECK mutant fish sensory deprived (sdp). Taken together, this study has identified the first pair of MMP, and their putative MMP inhibitor RECK that functions together in NC cell migration.


Introduction
The neural crest (NC) is a transient multipotential progenitor cell population arising from the dorsal folds of the neural tube [1,2]. NC cells undergo epithelial-to-mesenchymal (EMT) transition, and begin to migrate throughout the developing body, giving rise to the craniofacial skeleton, melanocytes and sympathetic and sensory ganglia [3,4]. Tissues surrounding the neural tube produce both positive and negative cues that guide NC cells along defined paths [5]. Trunk NC cell migration is guided by signals emerging from adjacent somites [6]. Once in the trunk, NC cell migrate either ventromedially or dorsolaterally, which eventually predicts their fate [5], however trunk NC cell are specified before reaching their final location [3]. NC cells migrating via the ventromedial route without invading the sclerotome (somite) become neurons and glia of the sympathetic ganglia, and adrenal chromaffin cells [1]. NC cells taking the ventromedial route invade and remain within the sclerotome to form Schwann cells, sensory neurons and glia of the dorsal root ganglia (DRG). NC cells that take a dorsolateral route in between the dorsal ectoderm and the dermamyotome differentiate into melanocytes [5]. In terms of molecular cues, trunk NC cells that migrate via the ventromedial route enter the somite via attraction cues from CXCR4/CXCL12 signaling molecules [7,8]. These NC cells are confined to the rostral sclerotome by Neuropilin2/Semphorin 3F repulsion molecules working in concert with Eph/ephrin signaling, F-spondin and proteoglycans that reinforce this migration route [8]. As with most cell migration processes, the extracellular matrix surrounding tissues play a critical role in directing NC cells to their final destination [9]. Matrix metalloprotease (MMPs) enzymes are responsible for degradation of extracellular matrix, and facilitate the migration and invasion of NC cells [10,11]. Recently, a group of molecules has been identified that control MMPs activity via direct inhibition. These molecules also contribute to the migratory routes taken by motile cells. For example, a zebrafish genetic mutant, sensory deprived (sdp) [12] carries a mutation in a glycophosphatidyl-inositol (GPI) anchored metalloprotease inhibitor protein Reversion-inducing-Cysteine-rich protein with Kazal motifs (RECK). Sdp mutant zebrafish show NC cell migration defects leading to defective DRG formation [12]. Despite these accepted concepts, the identification of specific MMPs involved in NC cell migration is still in its infancy. In this study, we extend the role of MMPs to NC cell development by identifying and characterizing the function of a zebrafish ortholog of MMP17 in embryonic NC migration.

Biochemical characterization of the zebrafish mmp17b gene
We performed a search for transcripts in the zebrafish information network (ZFIN) using expression pattern keywords such as vessels (Methods S1). These searches identified an expressed sequence tag (EST) sb:eu434 showing a segmental expression pattern along the embryonic trunk reminiscent of the intersomitic vasculature. A BLAST search revealed that the sb:eu434 EST sequence showed 56% sequence homology with matrix metalloprotease17 (MMP17; MT4-MMP) ( Figure  S1A). As part of a whole genome effort to characterize MMPs, a previous study has identified using bioinformatics a MMP17 ortholog (mmp17a) in zebrafish [13]. We therefore renamed the EST as mmp17b. The mmp17b gene consists of 11 exons, predicted to encode a 613 AA protein. Analysis of the predicted Mmp17b protein product using multiple web based programs [14,15] identified a zinc-dependent metalloprotease domain (AA141-309) ( Figure S1C), two hemopexin-like (HX) domains ( Figure S1C) (AA 356-399 and 487-541) and GPI anchor site (AA 592). To confirm the bioinformatic prediction, we cloned the full-length zebrafish mmp17b cDNA into pcDNA3.1 with a myc tag ( Figure 1C, Methods S1) to determine expression via western blot using myc ( Figure 1C) and MMP17 antibodies ( Figure 1D). We also included cDNAs for two of the identified human GPI-linked MMPs, MMP17 and MMP25 [16] as controls. We transfected 293T cells with myc expression constructs, and generated soluble (S) supernatant and insoluble pellet (P) fractions, that were probed with myc and MMP17 antibodies. Western blots with myc antibody (Methods S1) showed a ~70 kDa band in Mmp17b pellet lysate lanes ( Figure 1C) similar to human MMP17 (~73 kDa) and MMP25 (~68 kDa). The S fractions also showed bands at the respective sizes although the intensity was quite weak. Interestingly, the human MMP17 antibody cross-reacted with the zebrafish Mmp17b protein in 293T lysate ( Figure 1D), which was again observed predominantly in the P fraction. Since GPI-anchored proteins are often localized in caveolinrich membrane fractions [17], we performed immunofluorescence (IF) with human MMP17 and caveolin antibodies on 293T cells transfected with human MMP17 ( Figure 1E-G) and zebrafish Mmp17b ( Figure 1H-J) cDNAs (Methods S1). Both human MMP17 ( Figure 1G) and Mmp17b proteins ( Figure 1J) are co-localized with caveolin in 293T cells. Taken together, these results support the prediction based on sequence homology that the mmp17b gene is a GPIanchored MMP, and is a putative ortholog of the GPI-family member MMP17.

Expression analysis of mmp17b across embryonic development
To investigate the temporal expression pattern of mmp17b during embryonic development, we performed RT-PCR on total RNA isolated from stage-specific wild type strain TübingenAB (TuAB) zebrafish embryos using gene-specific primer sets for mmp17b, mmp17a and actin (Table 1 and Figure 2A). Mmp17a, but not mmp17b was observed between 1-10 hpf which means mmp17a is a maternally derived transcript. Expression of mmp17b is first detected at 18 hpf and continues through 72 hpf with most robust expression observed between 24-72 hpf. To determine the temporal and spatial expression pattern for mmp17b, whole mount in situ hybridization (WISH) was carried out on 17, 20 and 26 hpf embryos. At 17 hpf, discontinuous mmp17b ISH staining is visible in the neural tube (NT) region ( Figure 2B). At 20 hpf, mmp17b transcript is observed in the ventral mesoderm tissue -intermediate cell mass (ICM) region adjacent to the yolk extension ( Figure 2C). At 26 hpf ( Figure 2D, arrows), mmp17b + cells appear throughout the middle of the trunk region, which on high power are observed adjacent to the somite boundaries ( Figure 2D''). To elucidate the cell types that express mmp17b in the trunk, we performed double staining by combined fluorescent in situ hybridization (FISH) for mmp17b and immunostaining for Flk-GFP (vascular) ( Figure 2E) or znp-1 (cell body -motor neuron)/zn-1 (axon -motor neuron) ( Figure S2A) or 4D9 (muscle pioneer cell) ( Figure 2F) or immunostaining for RFP using a Tg(sox10(7.2):mrfp) line or ISH for crestin (NC) ( Figure  2G-J) marker. Mmp17b messenger RNA is not detected in flk positive ISVs ( Figure 2E) but is readily detected within the intervening somites. Mmp17b expression appears to colocalize with znp-1/zn-1 + motor neurons ( Figure S2A), and to some extent with 4D9 + muscle pioneer cells ( Figure 2F). Mmp17b clearly co-localizes with crestin positive (crestin + ) NC cells in the trunk at 26 hpf ( Figure 2G & 2J), and interestingly this co-localization is observed in the anterior somites ( Figure  2G & 2I) where mmp17b + cells have migrated more ventrally than crestin + cells. Our staining procedure includes a tyramide signal amplification process. Tyramide is a precipitate that prevents the combination of colors (e.g., red and green equals yellow) that is normally used to indicate co-localization. Since one precipitate covers the other, we looked for cells that appeared to have both colors represented on them to indicate co-localization. To confirm this observation, we also conducted FISH for mmp17b in Tg(sox10 (7.2):mrfp) [18] embryos, and we also observed co-localization in sox10 + NC cells at the same stage ( Figure 2N, arrowhead). In parallel experiments, mmp17a expression was detected across the dorsal side of the embryo, more dorsally than seen for mmp17b ( Figure 2K). Taken together, we observe mmp17b expression post gastrulation in tissues that undergo morphogenesis and patterning, and in particular in crestin + NC cells.

Mmp17b loss of function (LOF) embryos show defects in trunk NC cell patterning
To assess function of mmp17b, we designed two separate splice morpholinos (MOs) that target the exon-intron junction of exons 3 and 4 of mmp17b ( Figure S2B), and validated the MOs using RT-PCR with exon-specific primers spanning the targeted junctions ( Figure S2B). Injection of 7.5 ng of MO1 resulted in expression of an abnormal higher molecular weight gene product ( Figure S2C, lane MO1) because of intron retention, which resulted in an in-frame stop codon. Injection of MO2 at 12.5 ng dosage resulted in multiple aberrant RT-PCR products ( Figure S2C, white asterisk). We performed WISH at 14 hpf, a stage prior to initial detection of mmp17b transcript ( Figure 2A) for myoD (somites) [19], and for etv-2 (early endothelial hemato-vascular precursor cells) [20] in control and mmp17b injected morphants ( Figure S2D, compare CMO vs. mmp17b MO1). We observed no alteration in these structures implying that the loss of mmp17b does not affect the development of somites or endothelial precursors during embryonic development. Based on the co-expression pattern of mmp17b and crestin ( Figure 2G), we investigated crestin + NC cells in mmp17b morphants, and observed that both mmp17b MO1 ( Figure 3B Figure 3A). This abnormal crestin patterning was quantified (Methods S1), and greater than 80% of embryos displayed crestin patterning defects in mmp17b morphants ( Figure 3G). The crestin patterning defect is defined as the irregular spacing of crestin + cells in the trunk presumably due to altered migration. Complementary gain of function with mRNA injection and rescue experiments were attempted by coinjecting human MMP17 or zebrafish mmp17b mRNA along with mmp17b MO. Unfortunately both gain of function (GOF) and rescue experiments caused embryonic death, even at low doses implying stress-like responses in early embryos to presence of MMP RNAs at stages when they are not expressed (before 18 hpf). Our mmp17b LOF study indicates a critical role for mmp17b in crestin patterning in the trunk of developing zebrafish embryos.

Small molecule inhibition of MMPs results in a neural crest phenotype similar to mmp17b morphants
We utilized a complementary chemical biology approach to address the function of mmp17b in embryonic development, an approach extensively used in zebrafish to perturb developmental processes [21]. We investigated several broadspectrum MMP inhibitors [22,23] and selected Marimastat, a collagen peptidomimetic, and ONO-4817, a synthetic hydroxamic acid-based non-peptidomimetic ( Figure S3A-B), for further experiments. We incubated dechorionated or chorion punctured Tg(kdrl:EGFP) embryos in 0.5 mM of each compound from 10 to 26 hpf followed by fixation in 4% PFA. Crestin WISH was performed on the MMP inhibitor-treated as well as DMSO-treated control embryos. The embryos treated with both MMP inhibitors ( Figure  There are fewer crestin + cells in MMP-inhibitor-treated embryos compared to the DMSO-treated embryos, and the crestin + cells that are present in MMP-inhibitor-treated embryos ( Figure 3E & 3F, compare red bracket) are mislocalized when compared to controls. Quantification of the defects shows greater than 80% of embryos have defective crestin patterning in both MMP inhibitor-treated embryos ( Figure 3H).
To determine whether crestin + cells in mmp17b morphants were undergoing apoptosis, we performed a terminal deoxynucleotidyl transferase dUTP end labeling (TUNEL) assay as described in Methods S1. As a positive control for our assay, we treated embryos with 6 mM hydrogen peroxide, which results in generalized apoptosis, and was detected in our staining method ( Figure S3D). However, in uninjected ( Figure  S3C, UI) or mmp17b MO1-injected ( Figure S3E) embryos, we noticed no difference in TUNEL staining. We next investigated the possibility of altered cell proliferation in mmp17b morphants by performing phospho-histone3 (pH3) staining ( Figure S3F-H, Methods S1). Although we noticed a slight increase in the overall pH3 + cells in the trunk of the mmp17b KD embryos ( Figure S3H), these results were not statistically significant. Taken together, the LOF-MO and the MMP inhibitor results support the hypothesis that mmp17b affects trunk NC cell development in embryonic zebrafish.

Mmp17b is involved in trunk NC cell migration
In both mmp17b morphants ( Figure 3B & 3C, compare white bracket area) and MMP-inhibitor-treated ( Figure 3E & 3F, compare red bracket area) embryos, the uniform distribution of crestin + cells in the dorsal region of the embryo observed in control-MO ( Figure 3A) or DMSO-treated ( Figure 3D) fish was disturbed. In morphant or inhibitor treated fish, crestin + cells are detected more posteriorly along the dorsal axis compared to controls, suggesting a change in migration pattern. Trunk NC cells that migrate and differentiate along dorsolateral routes become melanocytes [5]. Therefore, we investigated melanocyte spatial distribution in embryos injected with cMO ( Figure 3I) or mmp17b MO1 ( Figure 3J) at 48 hpf. Embryos injected with mmp17b MO1 showed an increase in melanocyte cells that were located on either side of the midline (lateral melanocytes) ( Figure 3K, lateral cells, p<0.05) compared to   control-MO while no difference ( Figure 3K, medial cells, p>0.05) was noted in cells located at the midline (medial melanocytes). However, the total number of melanocytes between the control-MO and the mmp17b-MO1-injected fish were roughly equal. These observations suggest that melanocyte differentiation from crestin + cells is unaffected but that their final destination in the trunk has changed. Together, these results suggest that loss of mmp17b results in trunk NC cell migration defects in the developing zebrafish embryo.

Mmp17b and its putative inhibitor (Reck) are proximally expressed in the trunk, interact biochemically, and are involved in DRG formation
A recent paper identified a GPI-anchored inhibitor of MMP's called Reck, which was involved in DRG formation [12]. The authors demonstrated that reck and crestin co-localize between 22 hpf and 30 hpf, and we show co-localization here at 26 hpf ( Figure 4F). We next investigated whether reck and mmp17b are co-expressed spatially and temporally during embryonic development. We performed FISH for reck and mmp17b on 24 hpf fixed embryos, and imaged double-labelled embryos using confocal microscopy. Reck and mmp17b transcripts are not detected on the same cells ( Figure 4E). Triple staining for vessels (Flk-GFP), reck and mmp17b ( Figure 4G & 4H), showed that mmp17b expressing migrating NC cells in close apposition to reck expressing cells in the mid-trunk region ( Figure 4I). Reck + cells appear to delimit the position of the crestin + /mmp17b + double positive cells in the trunk. We also performed immunostaining for endogenous MMP17 ( Figure 4J, white asterisk) and RECK ( Figure 4K) proteins in HeLa cells, and observed that MMP17 and RECK co-localize in the same cellular compartment ( Figure 4L, merge). These RNA and protein data suggest that RECK and MMP17 can interact by expression within the same cell or adjacent cells.
Zebrafish mutant for Reck, sensory deprived (sdp), have shown failure of sensory neuron precursors to migrate to the position of the DRG [12]. To determine whether mmp17b and reck functionally interact we investigated DRG formation in mmp17b morphant embryos using the pan-neuronal marker elavl1. Unfortunately, knockdown of mmp17b in sdp mutant fish did not rescue the DRG phenotype, indicating that the loss of control of mmp17b is not the mechanism for DRG loss in this mutant. However, the mmp17b morphant fish embryos showed elavl1 + neurons located more ventrally and dorsally ( Figure 5B) compared to controls ( Figure 5A) suggesting defects in migration and arrangement of these neurons. Quantitation (Methods S1) in Figure 5C shows that the number of DRG's has not altered ( Figure 5C, p=0.43) but the position of the DRG's (migrated DRG) along the axis of the embryos is clearly altered (Figure 5C, p<0.01). Next, we investigated at the biochemical level whether Mmp17b and RECK immunoprecipitated (IP) in a heterologous overexpressing Cos 7 cell system. We overexpressed tagged versions of both proteins (myc-Mmp17b, RFP-Reck) in Cos7 cells, and performed IP with RFP antibody and western blotted for myc epitope as described in Methods S1. The myc-Mmp17b protein immunoprecipitates in the RFP-Reck lane ( Figure 5D, arrow), which was also observed with human MMP17 ( Figure 5E).
Next, we investigated whether Reck blocks MMP17 activity. We incubated pure MMP17 protein with 1 mM 4aminophenylmercuric acetate (APMA), and performed western blot with MMP17 antibody. APMA is used to activate MMP's in the pro(inactive)-form. The MMP17 activation event occurs in vivo via furin cleavage, and APMA is a chemical mimic that is used for this purpose. The activation reaction is autocatalytic and is triggered in the presence of APMA (or furin). APMA activates MMP17 as indicated by numerous smaller molecular weight bands that are noticed in APMA-treated MMP17 ( Figure  5F, lane 2) compared to the untreated MMP17 ( Figure 5F, lane 1). When RECK pure protein was co-incubated with APMAactivated MMP17 protein, the activation was blocked (Figure 5F, lane 3) since no lower molecular weight bands were observed. We confirmed this result by incubating the same samples in a FRET-peptide based MMP assay ( Figure 5G). In this assay, the cleavage of the peptide substrate results in fluorescence. Incubation of the activated MMP17 with MMP17specific FRET-peptide substrate results in high fluorescence values compared to unactivated MMP17 or activated MMP17 co-incubated with RECK protein. Taking the western and FRET assay data together indicates that RECK blocks the activation step of MMP17. These data collectively suggest that Reck-Mmp17b show functional interaction during DRG formation during trunk NC cell development.

Discussion
In this study, we have characterized the function of a novel matrix metalloprotease gene, mmp17b, in zebrafish NC cell development. We demonstrate mmp17b expression in migrating NC cells, present the first demonstration of an in vivo function in vertebrates for a GPI-anchored MMP, and implicate functional interaction of Mmp17b and the MMP inhibitor Reck during NC cell development.
NC cell migration falls under three distinct phases (1), Directed migration resulting from contact with the ectoderm and cues from the microenvironment (2), Contact-mediated guidance facilitating homing to the target site, and (3) Contactinhibition of movement upon entry and colonization of the target site (i.e. the ventral trunk for sympathoadrenal neurons) [4]. This behavior occurs as streams of cells that migrate directionally [24]. Although it has long been speculated that the extracellular matrix plays an active role in this process, to date the role(s) of MMPs in NC migration have remained poorly understood. Here, we implicate Mmp17b in the directed migration of NC cells in the trunk from the overlying ectoderm.
Of the 28 known human MMP's, only two are predicted to be attached to the cell surface via a GPI anchor. Both of those, MMP17 and MMP25, have been implicated in cancer progression [25,26] but a role for these MMP's during development has not been previously described. Here, we identify zebrafish Mmp17b whose close sequence similarity and predicted catalytic domains indicate that it is an ortholog of human MMP17. In terms of expression patterns, the only other known mmp to express in NC cells during development is MMP8 [27], and MMP2 [28]. At present, the GPI-anchor, the domain similarities to MMP17, and co-localization with caveolin protein argue that Mmp17b is most similar to human MMP17.
Mmp17b expressing NC cells at 26 hpf in the developing trunk are found in a sub-set of crestin + cells that have migrated ventrally in the anterior somites ( Figure 2G), and are also found in sox10 + cells ( Figure 2N). This feature is more prominent in the posterior somites where the majority of crestin + cells are located in the dorsal portion of the neural tube region. Most of the mmp17b expressing cells are positioned further ventrally ( Figure 2H) near the dorsal aorta. The mmp17b expression in migrating NC cells implies a functional contribution to this process, and indeed mmp17b LOF embryos show defects specifically in NC cell migration with little change in NC cell proliferation or apoptosis. We also examined mmp17a, a close ortholog of mmp17b, and determined that mmp17a expression occurred much earlier, and was not expressed in the trunk region as mmp17b ( Figure 2K). Since 20% of the zebrafish genome is duplicated [29], it is feasible that the mmp17a and mmp17b have disparate function. To assess this we will need to perform combination knockdown analysis, which is likely to result in pre-gastrulation defect owing to the early expression of mmp17a. Interestingly, broad-spectrum inhibition of MMP function also leads to similar deficiencies in NC cell migration, implying that Mmp17b activity is critical for NC cell migration. Further, we did not assess for additional defects in NC migration such as precocious migration or dorsal position location of these cells. Taken together, the MMP inhibitor and mmp17b knockdown embryo phenotypes are similar in that the crestin + cells are misplaced, but there is no crestin + cell death ( Figure S3E).
Tissue-specific metalloprotease inhibitors (TIMPs) regulate MMP activity. For MMP17, TIMP2 [30,31], TIMP3 [30], and to a lesser extent TIMP1 [30] have all been shown to inhibit MMP17 activity, and these were considered first as probable inhibitors of Mmp17b. In zebrafish TIMP1 has not been identified, TIMP3 is predicted, and two transcripts of TIMP2, timp2a and timp2b have been identified. We conducted WISH for timp2a and timp2b in 26 hpf zebrafish embryos ( Figure S4D & E), and observed that timp2a is only expressed in the tail tip and hindbrain regions, while timp2b is expressed in plexus and regions adjoining or in the vasculature. We investigated the expression pattern of a GPI-anchored inhibitor of MMPs called reck that was recently shown to be critical for DRG formation [12]. Reck is co-expressed in crestin + expressing cells. At 26 hpf, reck (presumably crestin + ) expressing cells migrate ventrally from the trunk and envelops mmp17b expressing NC cells ( Figure 4G, asterisk) in the middle of the trunk. Loss of mmp17b in embryos show crestin + cells near the dorsal region of the neural tube at the posterior end of the embryo at 26 hpf. However, loss of reck causes sox10 + cells (presumably crestin + ) NC cells to increase their velocity of migration ventrally [12] implying that presence of Mmp17b is responsible for directional migration of NC cells from dorsal ectoderm to ventral regions of the embryo.
The juxtaposition of mmp17b expressing cells with reck expressing cells implies that the putative interaction of the two cells will most likely trigger signals that dictate the final location of the differentiated DRGs. At a molecular level, RECK protein is capable of blocking the activation of MMP17 in vitro, which in itself suggests that MMP17 is a new, previously unreported target of RECK. Further support for this hypothesis is from mmp17b LOF embryos data where the number of DRGs has not changed presumably because of the presence of Reck but their absolute position on the dorsal-ventral axis is affected. This suggests that Reck is necessary for specification of DRGs from NC precursors, and Mmp17b is necessary for the DRGs final position. Taking our data and those of Prendergast et al [12] together, we hypothesize that Reck is responsible for the NC cell differentiation into DRGs post migration dorsally after 30 hpf. Concomitantly or slightly earlier, mmp17b expressed in a dorsal NC cell population migrate ventrally until they are properly oriented with reck expressing cells from the ventral region. In this scenario, loss of reck would result in NC cells not receiving the proper cues to differentiate into DRGs resulting in loss of DRGs as observed in sdp mutants. Loss of mmp17b however, results in NC cells differentiating into DRGs because of Reck presence but they do not migrate to the proper position in the zebrafish trunk. Future studies will need to elaborate the nature of this interaction and signalling mechanisms in this process.
In terms of melanocytes that differentiate from trunk NC cells, disruption of NC cell migration in mmp17b LOF embryos results in no change in number of melanocytes but alters their migration to the final position, which implies that the matrix and surrounding cells does influence the final positioning of the differentiated NC cells. In addition to melanocytes, mmp17b morphants also showed defects in tail fin skeletal structure ( Figures S4B & S4C) that is likely to be derived from NC cells [32]. This defect was also observed in MMP-inhibitor-treated embryos, and is likely to be secondary to alteration in trunk NC cell migration events. However, recent studies show that NC does not play a role in larval fin development [33]. We speculate that the tail and fin abnormalities could be due to disruption in cartilage development in this region via collagen processing, because human MMP17 is known to play a role in the activation of a disintegrin and metalloproteinase with thrombospondin-like motif-4 (ADAMTS-4) [34,35] during cartilage breakdown. Other possibilities include regulating the levels of growth factor sequestered in the matrix. For example, MMP17 could behave like MMP9 in that it facilitates the availability of sequestered VEGF in the matrix [36].
In summary, this study is the first to unravel the molecular role for Mmp17b in vertebrate development, and suggests a functional role for this molecule and its cognate TIMP Reck during NC cell development.

Ethics Statement
All zebrafish studies were performed according to Medical College of Wisconsin animal protocol guidelines under protocol AUA320. The MCW's IACUC committee approved studies conducted here.

Morpholino Injections
Gene Tools, Inc. designed splice MOs targeting the exonintron junction of exons 3 (mmp17b MO1) and 4 (mmp17b MO2) of the mmp17b gene ( Figure S2B). All primers and morpholino sequences used in this study are provided in Table  1. Both gene-specific and control MOs were reconstituted to a final concentration of 2 mM using water. Microinjection was performed with two nanoliters (nL) injected into each embryo at 1-2 cell stage. Doses of MO were empirically determined and were 15 ng -Control MO (cMO), 12.5 ng -mmp17b MO2, 7.5 ng -mmp17b MO1. Concentration of MO was chosen that showed the most minimal gastrulation delay (i.e., most normal looking), and yet showed defects in marker staining. Each injection experiment was repeated three independent times with approximately 40 embryos/condition each time. For MO targeting, RT-PCR was performed with primers located in the adjacent exons as shown in Figure S2. To determine the consequence of MO targeting, the alternate bands identified in the MO-targeted lanes were cloned into the CloneSmart ® HCKan vector using the CloneSmart ® Blunt Cloning Kit (Lucigen ® ) and sequenced.

Whole Mount In Situ Hybridization (WISH)
Zebrafish mmp17b, timp2a, and timp2b cDNA was generated from total zebrafish 24 hpf RNA using RT-PCR. The RT reaction conditions included: 1.5 µg of RNA, 1 µL of 50 µM oligo dT 20 (Invitrogen ™ ) 10 mM dNTP, in a 15 µL reaction. Standard RT-PCR protocols were carried out with a slight modification in that the Phusion ® DNA Polymerase (Finnzymes) was used in the PCR reaction to amplify the full-length mmp17b, timp2a or timp2b fragment. The amplified PCR fragment was used as a template in future PCR amplifications for probe generations. Initially, T3 and T7 promoter sequences were designed into the forward and reverse primers for generating template that were used in subsequent in vitro transcription to produce digoxigenin-labeled RNA-probes. Subsequently, we cloned the full-length mmp17b fragment into pcr4-TOPO vector (Life Technologies), linearized the plasmid with NotI enzyme, and the T3 sequences in the plasmid were used for probe generation as described previously [38]. The WISH procedure was performed as described before [38]. Twocolor fluorescent in-situ-hybridization (FISH) was performed with the following modifications: Instead of single riboprobe incubation, embryos were incubated with pairs of digoxygenin (DIG)-and dinitrophenol (DNP)-labeled riboprobes. Labeling of riboprobes was done by in vitro transcription from PCRgenerated probe templates. We used a DIG RNA labeling mix (Roche) to produce DIG-labeled riboprobes, and DNP-labeled riboprobes were produced using a mix of ribonucleotides and DNP-11-UTP (PerkinElmer). After hybridization, stringent washes in PBST (PBS, 0.1% Tween-20) and blocking of embryos in block solution (1% BSA and 1% DMSO in PBST, 2 h at room temperature [RT]) were done. RNA-DIG probe hybrids of the first transcript were detected by overnight incubation at 4°C with anti-DIG Fab fragments conjugated to horseradish peroxidase (Roche, 1:1000 dilution) in block solution. Fluorescent signal was produced using Cy3, fluorescein or Cy5 tyramide signal amplification (TSA) systems (PerkinElmer, 1:100 dilution, 30-40 min incubation at room temperature) as the first color. Following washes in PBST, quenching of peroxidase activity (at least 1 hour incubation in 1.5% H 2 O 2 in PBST), and further extensive PBST washes and another blocking step, embryos were incubated with anti-DNP antibody (PerkinElmer, 1:500 dilution in block solution) overnight at 4°C. Then the second fluorescent signal was developed using a different tyramide substrate than in the first round. Embryos were washed thoroughly in PBST and ready for imaging. Details on the double staining procedure are available upon request. The embryos were mounted using PolyMount solution (Polysciences) and allowed to dry overnight in darkness. The embryos were imaged the next day using a Zeiss LSM 510 confocal microscope. The images were analyzed using Volocity software (PerkinElmer).

Chemical Inhibitor Analysis
Two broad-spectrum inhibitors Marimastat and ONO-4817 (Tocris Bioscience) were reconstituted to a concentration of 100 mM in 100% DMSO. The drug was diluted in E3 embryo water to produce working stocks. Ranges of concentration (0.5 mM, 50 µM, 5 µM, and 0.5 µM) were titrated for both drugs to determine a dose that yielded a phenotype without grossly disrupting the development of the embryo similar to the MO protocol. Total DMSO concentration was kept at 0.5%. Tg(kdr:EGFP) zebrafish embryos were punctured at 10 hpf with forceps to facilitate drug diffusion into the embryo. Ten embryos (10 hpf stage) were placed per well in a 24-well plate, and treated with 0.5 mL of either a drug or a DMSO control until the embryos reached 24 hpf stage at 28.5°C. The embryos were then dechorionated and analyzed for abnormalities similar to the MO injection described earlier.
Photography and imaging was also performed as described above. The concentration of drug that was optimal for these studies was 0.5 mM. The experiment was repeated twice, and all abnormalities were quantitated, documented, and graphed. WISH for crestin and mmp17b was conducted on the drug treated embryos as described previously.

MMP17 Activation and FRET-peptide MMP Assay
Human recombinant MMP17 (R&D Systems) was first diluted to 100 µg/mL in Activation Buffer (1M Tris, 1M CaCl 2 , 5M NaCl, 30% Brij-35, 0.1M ZnCl 2 , pH 7.5) and was then activated by incubating in 1 mM APMA (4-Aminophenylmercuric acetate, Sigma-Aldrich) for 2 h at room temperature. After activation, the MMP17 protein was diluted 1:50 in Assay Buffer (1M Tris and 1M CaCl 2 , pH 7.5) and 50 µl was plated in a black 96-well plate. MMP17 fluorogenic substrate (Mca-PLAQAV-Dpa-RSSSR-NH 2 -Fluorogenic Peptide Substrate III, R&D Systems) was diluted 1:200 in Assay Buffer and 50 µl was added to either experimental sample or 50 µl Assay Buffer for substrate control. After mixing the substrate with the samples, the plate was covered and protected from light and allowed to incubate at room temperature for 1 h. The plate was then read in a Molecular Devices Gemini EM fluorescent plate reader on the EndPoint setting. Each sample was read for 10 seconds at Excitation 320 nm and Emission 405 nm. This assay was performed twice, and gave reproducible results. were used to amplify mmp17b fragments. In both MO1 and MO2, the mmp17b amplicon was larger than the control MO injected sample. In MO1, there appeared to be a major higher band along with a minor band that was consistent with the normal mmp17b band. In MO2, there was a major band much higher than control and a minor band consistent with the higher band observed in the MO1 sample. Asterisks indicate the aberrant transcripts. β-actin was used as a loading control. D shows knockdown of mmp17b does not affect early development. When etv-2 (upper and lower left panels) and myoD (upper and lower right panels) are probed for in 14 hpf mmp17b MO1 injected embryos (lower left and right panels), there is no difference observed compared to control MO injected embryos (upper left and right panels). (TIF) Panel B illustrates the cupped-fin phenotype seen in mmp17b MO1-injected embryos compared to controls. This defect was quantitated in panel C that shows mmp17b MO1-injected embryos have a statistically significant increase in the cuppedfin defect compared to controls. Panels D and E show timp2a and timp2b 26 hpf ISH embryos (head is to the left). Yellow arrow in D shows plexus region and the black arrow indicates regions adjoining the vasculature. (TIF) Methods S1. Supporting Methods. (DOCX)