Introduction

Adenoviral vectors mediate gene transfer at a high efficacy compared with other vector systems, and they are currently the most frequently used vectors for cancer gene therapy.¹ Both a non-replicating p53-expressing adenoviral vector² and a replication-selective adenovirus (H101) have received regulatory approval in China.³ The success of recombinant adenoviruses as a cancer therapy is, however, limited by inefficient delivery. The suboptimal transduction of cancer cells is compounded by poor distribution of the virus within the tumor mass.⁴

The use of replicating adenoviruses is a logical development, in that by repeated rounds of infection, release, and re-infection of adjacent cells, the virus may spread from cell to cell through the tumor mass, despite any initial problems with uniformity of delivery.^5,6 This is certainly the case in cell culture monolayers, where a replicating virus can rapidly spread throughout a monolayer cell culture despite a low proportion of cells being initially infected. In contrast, xenograft tumors in immune-incompetent mice are rarely eradicated despite the persistence of high levels of infectious virus within the tumors.^4,7,8 Also, in clinical studies, despite some evidence of replication and efficacy, overall results have been disappointing.^9,10,11

Our previous study showed that after local injection of replication-competent adenovirus into xenograft tumors, high levels of titratable virus could be recovered from a tumor as late as 100 days after initial viral injection. Tumors even persisted at a time when infectious virus could be detected in the circulation, several weeks after initial viral injection.^4,7 The viral persistence without tumor eradication suggested that viral spread is limited through tumor tissue, and this concept is supported by the patchy and uneven intra-tumoral distribution of virus. Virus-infected cells can often be seen flanked by tumor necrosis and murine connective tissue. These data suggest that human adenoviral spread within tumor xenografts may be impaired by connective tissue.⁴

The hypothesis that is addressed in this study is that matrix within the tumor hinders the free cell-to-cell spread of replicating adenoviral vectors. Collagen I, a major component of tumor stroma,¹² has been shown to be of particular importance in creating a diffusive hindrance for therapeutic agents within tumors.¹³ Interstitial collagen fibrils are resistant to degradation by most proteases. However, members of the fibrillar collagenase matrix metalloproteinase (MMP) family, in particular MMP-1, MMP-8, and MMP-13, can break down intact triple-helical collagen.^14,15 MMP-8 is a Zn²⁺ metalloendopeptidase, predominantly expressed by neutrophils, but also by a few melanoma cell lines, chondrocytes, rheumatoid synovial fibroblasts, activated macrophages, smooth muscle cells, and endothelial cells.¹⁶ MMP-1 degrades type III collagen more efficiently than it does type I or type II collagen. MMP-8 cleaves types I, II, and III collagen with 20-fold selectivity for type I over type III, but MMP-8 does not degrade types IV or V collagen.¹⁴ MMP-13, in turn, degrades type II collagen sixfold more efficiently than it does type I or type III collagen.

In the present study we found that collagen I can block adenoviral diffusion in in vitro experiments. On the basis of this finding, AdMMP8, a non-replicating adenoviral vector that expresses MMP-8, was constructed and shown to break down collagen in vitro. To evaluate the effects of MMP-8 expression from a non-replicating virus on the oncolytic activity of a replicating virus, AdMMP8 was administered in combination with wild-type virus in murine xenograft tumor models. Intra-tumoral injection of non-replicating AdMMP8 in combination with wild-type virus led to reduced tumor cell growth and reduced expression of collagen within areas of virus-induced necrosis compared with wild-type virus given together with a non-replicating control virus.

Results

Effect of collagen I on adenoviral diffusion

Membrane inserts pre-coated with collagen I (3-m pore) were placed into the chambers of 24-well plates that had been pre-seeded with 293 cells growing as a monolayer on the base. Ad-gal, an E1a-deleted -galactosidase (-gal)-expressing virus, was added to the upper chamber. To infect the 293 cells plated on the base of the 24-well plates, the virus had to spread through the matrix-coated insert. Only Ad-gal virus–infected 293 cells would be expected to express -gal. The collagen I insert blocked the spread of Ad-gal virus so that only a few 293 cells stained blue after the addition of X-gal. In contrast, control non-coated inserts allowed the passage of virus, with subsequent intense blue staining of the 293 indicator cells (Figure 1a and b). Collagenase treatment of the collagen I insert restored the ability of Ad-gal to diffuse and infect the 293 indicator cells. The 293 cells were intensely stained blue compared with the untreated collagen I insert, in which only a few cells stained blue (Figure 1b and c).

Figure 1.

Virus diffusion across a collagen matrix. The bases of 24-well plates were pre-seeded with 293 cells. After 24 hours' incubation, inserts containing Ad-gal virus (multiplicity of infection of 5) diluted in 0.3 ml of minimum essential medium were placed onto the wells. The inserts had a 3-m pore size and were pre-coated with collagen I. The control insert had no coating. After 24 hours' infection, the ability of the virus to spread through the insert was analyzed. In situ staining of cells for -galactosidase (-gal) is shown. (a) Control wells allowed free diffusion of the virus. (b) Collagen I markedly reduced the proportion of cells expressing -gal compared with (a) control, and this can be restored to control levels by (c) pre-treatment with collagenase.

Full figure and legend (134K)

To determine whether the virus bound to the collagen I, the Ad-gal virus was placed onto the collagen I insert as previously, but then after 24 hours was aspirated and applied directly onto the 293 indicator cells. Similar proportions of positive blue cells were seen when taking the medium containing Ad-gal virus from the collagen I insert upper compartment and infecting 293 cells as when the same amount of Ad-gal virus was applied to and aspirated from the control insert (data not shown). This suggests that adenoviruses do not have a strong binding affinity to collagen and that the collagen matrix may serve as a physical barrier.

Tumor stroma is largely composed of collagen I, but the physical form within the tumor may be different from that on the inserts. To evaluate the potential importance of collagen I in reducing viral spread in tumors in vivo, we constructed an adenoviral vector that expresses a metalloproteinase with the potential to break down collagen I.

Expression of functional MMP-8 by a non-replicating adenovirus

On the basis of reports of the effective collagen I–degrading activity of MMP-8,¹⁴ we chose to construct an adenovirus to express the MMP-8 transgene as described in Materials and Methods. In the recombinant AdMMP8 adenovirus, a cytomegalovirus promoter drives the MMP-8 complementary DNA (cDNA) (Figure 2a). A control vector without a transgene, Adcon, was constructed in a similar manner.

Figure 2.

Construction and activity of AdMMP8. (a) Schematic of AdMMP8, an E1-deleted non-replicating adenovirus expressing the human matrix metalloproteinase-8 (MMP-8) full-length complementary DNA under the control of a cytomegalovirus (CMV) promoter. (b) A549 cells infected with AdMMP8 express MMP-8 messenger RNA as detected by real-time polymerase chain reaction (lane 3). The expression of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is shown as a control. Controls include non-infected cells (lane 1) and control vector–infected cells (lane 2). (c) Medium from cultures of infected cells was separated on a 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) glycine gel for immunoblotting with a goat polyclonal anti-MMP-8 antibody. Evidence of MMP-8 protein expression is seen in the AdMMP8-infected group in lane 3. (d) Supernatants from AdMMP8-infected cells show collagen-degrading activity (lane 3), as shown on 10% SDS-PAGE glycine with 0.1% gelatin gel (zymogen gel). Adcon, AdMMP8-infected control virus; LITR, left inverted terminal repeat; RITR, right inverted terminal repeat.

Full figure and legend (16K)

To investigate whether the AdMMP8 adenovirus directed the expression of functional MMP-8 protein, lung A549 cancer cells were infected with recombinant AdMMP8. Total RNA was extracted from AdMMP8-infected, control virus (Adcon)–infected, and uninfected A549 cells and used as template for real-time polymerase chain reaction (PCR). Control A549 cells and Adcon-infected A549 cells did not express MMP-8 messenger RNA (mRNA), whereas AdMMP8-infected A549 cells expressed MMP-8 mRNA, as demonstrated by a PCR product of the appropriate size (Figure 2b). Immunoblotting was also performed on the serum-free conditioned supernatants from the A549 cells under similar experimental conditions. The blots show a single band recognized by the polyclonal anti-MMP-8 antibody at 65 kd in size (Figure 2c). The collagenase activity of the supernatant was also demonstrated by zymography (Figure 2d). The gelatin hydrolysis visible in the zymogram gel indicates that the recombinant MMP-8 generated by infection of A549 cells with AdMMP8 was not only secreted into the culture medium but also functional on the gelatin substrate.

A549 cells infected with AdMMP8 break down a fibrillar collagen matrix

The tumor content of fibrillar collagen has been shown to influence diffusive transport within tumors.¹³ In this experiment we used a fibrillar collagen matrix insert that resembles the tumor architecture more closely than amorphous collagen.¹⁷ A fibrillar collagen membrane was conditioned by a monolayer of A549 cells infected with AdMMP8. The collagen membrane conditioned by AdMMP8-infected cells but not cells infected with a control virus became permeable to the adenovirus (Figure 3). The in vitro forms of collagen appear to form a physical barrier to viral diffusion. To evaluate this physical property we studied albumin diffusion. An albumin gradient of 400 was established (insert to well), and this was reduced to a gradient of less than 4 by 24 hours for the control membranes, whereas at 72 hours a gradient of 6 for amorphous collagen I and 16 for fibrillar collagen persisted. These experiments suggest that impaired viral diffusion results from the physical properties of collagen I.

Figure 3.

Effects of matrix metalloproteinase-8 (MMP8)–expressing cells on fibrillar collagen. Fibrillar collagen inserts conditioned by A549 cells infected with AdMMP8, but not Adcon-infected cells, facilitate diffusion of an adenoviral reporter construct. Adcon, control virus; AdMMP8, non-replicating adenoviral vector that expresses MMP-8; -gal, -galactosidase.

Full figure and legend (3K)

AdMMP8 infection of A549 lung cancer cells does not alter cell viability

A549 lung cancer cells were infected in a dose-response experiment (0.1–200 multiplicity of infection (MOI)) with AdMMP8 or control virus Adcon. At 6 days cell viability was determined using a WST-1 assay, and the data normalized to uninfected control cells. No differences were seen in the viability of cells infected with Adcon compared with cells infected with AdMMP8 (Figure 4a). Expression of MMP-8 from AdMMP8 did not, therefore, modulate cell survival compared with the control virus.

Figure 4.

Effects of AdMMP8 on cell survival and viral replication. (a) Viability of A549 cells infected with AdMMP8. A549 cells were infected with AdMMP8 or Adcon at the multiplicities of infection (MOIs) shown. Control cells remained uninfected. Cell viability was measured using a WST-1 assay and is presented as a percentage of uninfected cell control values. AdMMP8 did not modulate cell viability compared with control virus. The experiment was performed on four occasions and the titers measured in duplicate for each experiment. (b) Replication of wild-type adenovirus in the presence of AdMMP8. A549 cells were infected with Adwt300 at an MOI of 1 alone (gray) or in combination with AdMMP8 (solid) or control virus Adcon (white) in a dose-response as shown. Total viral yield was no different in the AdMMP8 dose-response compared with the Adcon dose-response. The experiment was performed twice and the titer measured in duplicate. Mean values are shown. AdMMP8, non-replicating adenoviral vector that expresses matrix metalloproteinase-8.

Full figure and legend (11K)

AdMMP8 does not modify the production of wild-type virus in A549 lung cancer cells

A549 lung cancer cells were infected with wild-type adenovirus (fixed MOI of 1) together with AdMMP8 or Adcon in a dose-response experiment (MOI 0.1–200). There were no differences in viral yield from Adcon co-infected samples compared with the AdMMP8 co-infected samples as determined by titration on 293 cells (Figure 4b). Of interest, on the basis of repeat titration of the lysates on A549 cells to detect only replicating virus, very limited replication of the non-replicating virus had occurred (Adcon MOI 200, 100% of Adwt300 control; AdMMP8 MOI 200, 97% of Adwt300 control: mean of two experiments titered in duplicate). That is, adding Adcon or AdMMP8 at an MOI of 200 to Adwt300 at an MOI of 1 did not influence the total viral yield, which was essentially entirely comprised of wild-type virus (which can replicate on both A549 and 293 cells).

MMP-8 expression from AdMMP8 improves the oncolytic efficacy of a replicating adenovirus

On the basis of the in vitro studies just described, matrix components, and in particular collagen I, may present a hurdle for effective adenoviral spread. To determine whether MMP-8 expression within the tumor would improve the efficacy of the replicating adenovirus, we studied two murine models: an A549 lung cancer xenograft model and a BxPC-3 pancreatic xenograft model.

Human lung A549 xenografts were injected with a wild-type replicating adenovirus Adwt300 (5 10⁸ plaque forming unit (PFU)) together with the non-replicating AdMMP8 virus (5 10⁸ PFU). The addition of AdMMP8 to the wild-type virus significantly increased the survival (survival endpoint defined as a threefold increase in tumor size) of animals compared with injecting control tumors with the wild-type virus in combination with Adcon (P = 0.008), as shown in Figure 5a. At day 19, Adwt300/AdMMP8 group tumors were one-third the size of vehicle-treated control group tumors (455 113 versus 1,354 358) and approximately half the size of Adwt300/Adcon group tumors (455 113 versus 788 228), as shown in Figure 5c.

Figure 5.

AdMMP8 in combination with Adwt300 reduces tumor growth. (a) A Kaplan–Meir analysis and log-rank test of animals bearing A549 lung cancer xenografts. Significant differences were seen between AdMMP8/Adwt300-injected and Adwt300/Adcon-injected animals (P = 0.008) on the basis of the time to a threefold increase in tumor volume (n= 6 each group). (b) A Kaplan–Meir analysis and log-rank test of animals bearing BxPC-3 pancreatic cancer xenografts. Significant differences were seen between AdMMP8/Adwt300-injected and Adwt300/Adcon-injected animals (P = 0.002) on the basis of the time to a threefold increase in tumor volume (n= 7 each group). (c) Tumor growth curve for A549 lung cancer. At day 26, tumors injected with Adwt300/AdMMP8 were approximately one-fifth the size of vehicle-injected control group tumors (508 158 versus 2,530 770) and one-third the size of Adwt300/AdCon-injected group tumors (508 158 versus 1,436 627). (d) Tumor growth curve for BxPC-3 pancreatic cancer. At day 44, Adwt300/AdMMP8 group tumors were one-seventh of the size of vehicle-treated control group tumors (200 42 versus 1,488 363) and approximately one-quarter the size of Adwt300/Adcon-injected group tumors (200 42 versus 772 196). Adcon, control virus; AdMMP8, non-replicating adenoviral vector that expresses matrix metalloproteinase-8.

Full figure and legend (19K)

Human pancreatic BxPC-3 xenografts were similarly injected with Adwt300 (5 10⁸ PFU) together with AdMMP8 virus (5 10⁸ PFU). The addition of AdMMP8 to the wild-type virus significantly increased the survival of animals compared with injecting control tumors with the wild-type virus in combination with Adcon (P = 0.002), as shown in Figure 5b. At day 44, Adwt300/AdMMP8 group tumors were one-seventh of the size of vehicle-treated control group tumors (200 42 versus 1,488 363) and approximately one-third the size of Adwt300/Adcon group tumors (200 42 versus 595 158), as shown in Figure 5d.

AdMMP8 injection alone did not affect the growth of the lung or pancreatic tumor xenografts. Tumor xenografts treated with AdMMP8 (plus control Adcon virus) showed similar growth kinetics to vehicle-treated controls. The lungs and livers of all animals in both xenograft models were evaluated for macroscopic evidence of metastases, but none were found. Lung sections were evaluated by microscopy in the A549 model Adwt300/AdMMP8 group for the presence of metastases, and none were detected.

MMP-8 mRNA expression is persistent

To confirm MMP-8 expression after viral administration, total RNA was extracted from the fresh A549 tumor tissues when mice were killed and was used as a template for reverse transcriptase PCR for MMP-8 mRNA. Four of six tumors in the Adwt300/AdMMP8 virus group were positive for MMP-8 mRNA at 42 days, and three of six tumors in the AdMMP8/Adcon group at day 26 (Figure 6a).

Figure 6.

Matrix metalloproteinase-8 (MMP8) messenger RNA (mRNA) and protein expression and collagenase activity in tumor lysates. (a) In vivo mRNA expression. Total RNA extracted from the fresh tumor tissues was used as template for real-time polymerase chain reaction of MMP-8. Four of six mice tumors in the Adwt300/AdMMP8 virus group were positive at 42 days from the time of injection; three of six mice tumors in the Adwt300/Adcon group were positive at day 26, when these animals with rapidly growing tumors were killed. The positive control is A549 cells infected with AdMMP8 in vitro (lane 1); the negative control is from a vehicle-injected tumor (lane 2). The expression of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is shown as a control. (b) MMP-8 protein expression is persistent. An MMP-8 enzyme-linked immunosorbent assay was performed on tumor lysates obtained from BxPC-3 pancreatic xenografts harvested between days 49 and 52. Compared with the control group, significantly higher levels of MMP-8 protein were detectable in the tumors that received AdMMP8/Adcon (P = 0.0001) or AdMMP8/Adwt300 (P = 0.0001). There was no difference between the AdMMP8/Adcon and AdMMP8/Adwt300 groups. (c) Collagenase activity can be detected in tumors infected with both AdMMP8 and Adwt300. All of the BxPC-3 tumor lysates were subject to zymography, apart from one tumor in the Adwt300/AdMMP8 group that was too small for further analysis. A clear band of collagenase activity is seen in the Adwt300/AdMMP8 group (lanes 5 and 6), which is at the same size as the positive control band (lane 9). Low background levels of activity are seen in the other groups. One representative blot of three is shown, but all of the AdMMP8/Adwt300 tumors (n = 6) showed activity on zymography, whereas none of the AdMMP8/Adcon group (n = 7) showed activity. Adcon, control virus; AdMMP8, non-replicating adenoviral vector that expresses MMP-8.

Full figure and legend (45K)

MMP-8 protein expression is persistent

An MMP-8 enzyme-linked immunosorbent assay (ELISA) was performed on tumor lysates obtained from BxPC-3 pancreatic xenografts harvested between days 49 and 52. Compared with the control group, higher levels of MMP-8 protein were detectable in all of the tumors that received AdMMP8, irrespective of the presence or absence of the wild-type virus (Figure 6b). Furthermore, there was no difference between the Adwt300/AdMMP8 (0.146 0.014 ng MMP-8/mg total protein) and AdMMP8/Adcon (0.142 0.024 ng MMP-8/mg total protein) groups. The MMP-8 antibody that was used is specific for human MMP-8 and detects both active and inactive forms of MMP-8.

Collagenase activity can be detected in tumors infected with AdMMP8 and Adwt300

All of the BxPC-3 tumor lysates were subject to zymography, apart from one tumor in the Adwt300/AdMMP8 group that was too small for further analysis. A representative gel (one of three) is shown in Figure 6c. A clear band of collagenase activity is seen in the Adwt300/AdMMP8 group, which is at the same size as the positive control band. Low background levels of activity were seen in the other groups. All of the AdMMP8/Adwt300 tumors (n = 6) showed activity on zymography, whereas none of the AdMMP8/Adcon group (n = 7) showed activity. At this late time point it would appear that the MMP-8 protein that is detected by ELISA in the AdMMP8/Adcon tumor lysates was not active or was inactivated by the large quantities of tumor tissue during sample preparation.

AdMMP8 combined with wild-type virus reduced tumor collagen expression

Virus distribution within the tumor and collagen expression were evaluated with Masson's trichrome staining for collagen and immunohistochemistry for adenoviral capsid proteins (Figure 7). A549 xenograft tumors contained abundant and dense collagen bands in the vehicle-injected control group tumors. In Adwt300 virus–injected tumors adenoviral spread was inefficient and patchy, and abundant collagen bands persisted in necrotic virus-infected and surrounding areas. In contrast, in Adwt300 plus AdMMP8–injected tumors collagen degradation was extensive within the virus-induced necrotic areas. The intensity of collagen staining in and around necrotic areas of all of the tumors was evaluated by two blinded observers (Figure 8). On the basis of the extent of collagen staining within the section, the observers created a rank order of all the samples from the least to most extensive amount of stained collagen visible. A significant reduction in collagen within the Adwt300/AdMMP8 tumors compared with Adwt300/Adcon was seen (P < 0.0001). The AdMMP8/Adcon group also scored with less collagen than the vehicle control (P = 0.0005).

Figure 7.

Collagen expression and viral protein expression within xenograft tumors. Immunohistochemistry (IHC) for adenoviral capsid proteins (a, c, e, g) and Masson's trichrome staining for collagen (b, d, f, g) in aligned serial sections. (a) As anticipated, no evidence of viral protein expression was seen in the control uninfected A549 xenograft tumors. (b) Abundant and dense collagen bands staining blue were seen in the trichrome stained control uninfected tumors. (c) The presence of adenovirus as detected by IHC was seen in the Adwt300/AdMMP8 injected tumors (many brown staining cells apparent in upper right quadrant of image, insert shows enlarged image). (d) This was associated with collagen degradation and extensive necrosis. (e) In the Adwt300/AdCon virus-injected tumors replicating adenoviral spread was inefficient, only a few cell in these tumors were infected on the basis of IHC for viral proteins (insert shows enlarged image) and (f) abundant collagen bands were present in necrotic and surrounding areas. However, AdMMP8/AdCon injected tumors showed (g) no adenoviral staining and (h) little evidence of collagen degradation. Adcon, control virus; AdMMP8, non-replicating adenoviral vector that expresses MMP-8.

Full figure and legend (211K)

Figure 8.

Intensity of collagen expression within xenograft tumors. The amounts of collagen within necrotic areas were scored by two blinded observers by ranking the intensity of collagen staining for all the tumors. A significant reduction in collagen within the Adwt300/AdMMP8 tumors compared with Adwt300/Adcon was seen (P < 0.0001). The AdMMP8/Adcon group also scored with less collagen than the vehicle control group (P = 0.0005) and Adwt300/Adcon group (P < 0.0001). Adcon, AdMMP8-infected control virus; AdMMP8, non-replicating adenoviral vector that expresses MMP-8.

Full figure and legend (6K)

Discussion

Clinical experience and animal models using replicating adenoviral vectors indicate surprisingly poor intra-tumoral spread of the replicating virus.^4,5,7,11,18 Effective oncolytic virotherapy is likely to be dependent on more efficient and dispersed delivery of the replicating adenovirus to the tumor mass than is currently achieved. Ineffective spread may render the virus unable to overwhelm the actively growing tumor.¹⁹ The causes of poor viral spread within tumors, even in immune-compromised hosts, are not clear, but there are several possibilities. Previous data from our laboratory and from another group have shown that hypoxia, a condition prevalent within large tumors, reduces viral replication.^20,21 Virus spread may also be impaired by the presence of stromal cells and stromal matrix within tumors. Human adenovirus replicates very poorly in murine cells, and murine stromal cells may limit viral spread in xenografts. The aim of this work was to determine whether matrix components within tumors might impede the spread of replicating adenovirus.

Previous studies that support this supposition are the findings by Brown et al. that collagen can provide a diffusive hindrance to the penetration of therapeutic molecules within tumors, and that this can be improved by matrix modification.¹³ Also, direct administration of collagenase/dispase or trypsin into glioma xenografts has been shown to enhance the extent of infection of a non-replicating adenoviral vector expressing a reporter gene.²²

Using cell culture inserts coated with collagen I, we have demonstrated that collagen I significantly blocked adenoviral diffusion. When the collagen I on the insert was treated with collagenase, the ability of the adenoviruses to diffuse through the insert was restored. We also confirmed that this blocking effect was associated not with the adenovirus binding to collagen I but with limiting of the passage of virus particles through the membrane by collagen I.

In vitro studies of this kind clearly have limitations, so we constructed AdMMP8, a non-replicating virus expressing human MMP-8, for further studies. MMP-8 was chosen to break down collagen because MMP-8 has high activity and some selectivity towards collagen I and has previously been shown to reduce the amount of collagen in several conditions. The administration of MMP-8 directed by a non-replicating adenovirus has been shown to reduce fibrosis in cirrhotic mouse livers.^23,24,25 MMP-8 activity has also been associated with collagen degradation in humans. In particular, secretion of MMP-8 by neutrophils may play a role in resolving fibrotic scar formation during cholestasis,²⁶ and MMP-8 may play a role in involution of the postpartum uterus.²⁷

AdMMP8 was evaluated in an in vitro fibrillar collagen model and in two murine models: human A549 lung cancer and BxPC-3 pancreatic cancer xenografts. Fibrillar collagen, like amorphous collagen, blocked viral diffusion. However, when the fibrillar collagen was conditioned with A549 cells infected by AdMMP8, diffusion was restored. In both the lung and pancreatic xenograft models, addition of the non-replicating MMP-8-expressing virus to the wild-type virus reduced tumor growth compared with the addition of a non-replicating control virus. Reduced amounts of collagen were visible within necrotic areas of the tumors that received the non-replicating AdMMP8 virus together with the wild-type replicating adenovirus.

These findings support the hypothesis that matrix components, in particular collagen I, limit viral diffusion and spread within tumors. In addition, others have shown that collagen I fibers within a tumor have an important effect on diffusive transport and can be modulated by collagen dissolution using the hormone relaxin, which up-regulates several MMPs.¹³ Relaxin expression from a replicating adenovirus has also recently been shown to improve tumor responses and viral spread in murine models.²⁸

Our data show that MMP-8 can improve tumor responses in combination with a single low dose of a replicating virus and that this is associated with reduced amounts of collagen within the tumor. Importantly, we show that infection with AdMMP8 does not modify tumor cell survival or production of wild-type virus in vitro experiments. In addition to collagen degradation, other mechanisms may also play a role in the positive effect of MMP-8. For instance, cytokines can be bound to matrix components and released by the activities of MMPs. In addition, MMP-8 has been shown to have anti-inflammatory effects during allergen-induced lung inflammation, partly owing to a regulation of inflammatory cell apoptosis,²⁹ and may also protect against lethal endotoxin-induced hepatitis.³⁰ Although clear changes in collagen composition within the tumors were noticed, no obvious changes in inflammatory cell infiltrates within these tumors in nude mice were evident.

Some of the matrix metalloproteases have been found to have an important role in tumor invasion and metastasis. This is a potential deleterious effect of the administration of MMP-8. However, so far MMP-8 has not been associated in a positive way with tumorigenesis or metastasis. In fact, in isogenic breast cancer cell lines, expression of MMP-8 has been associated with inhibition of metastases and invasion.³¹ Furthermore, genetic manipulation of a metastatic cell line to up-regulate the activity of the MMP-8 gene has been shown to decrease metastatic spread; conversely, its down-regulation by incorporation of a targeted ribozyme into a non-metastatic line resulted in the cells becoming metastatic.³² Mice deficient in MMP-8 (MMP-8–^/– mice) have been reported to have an increased incidence of skin tumors,³³ and female MMP-8–^/– mice submitted to oopherectomy or treated with tamoxifen were more susceptible to tumors than wild-type mice. Furthermore, in this study, bone marrow transplantation experiments revealed that MMP-8 supplied by neutrophils was sufficient to restore the natural protection against tumor development mediated by MMP-8 in this model.³³

Our data also suggest that MMP-8 expression has no deleterious effect on cancer progression. AdMMP8-treated tumors in the absence of wild-type virus grew with similar growth kinetics to the control group in two animal models. There was also no evidence of metastatic spread to the lung after treatment of mice with Adwt300/AdMMP8. All these findings suggest that there was no increase in tumor invasiveness or enhanced tumor cell spread after AdMMP8 administration.

It might be anticipated that the non-replicating MMP-8-expressing virus would replicate in the presence of the wild-type replicating virus on the basis of complementation of the E1a gene defect with the E1a expressed from the replicating virus.³⁴ However, MMP-8 protein levels within the pancreatic tumor at the end of the experiment showed similar levels of MMP-8 in the tumor receiving AdMMP8 in the presence or absence of the replicating virus. Of interest, the in vitroviral production experiments also showed that the overwhelming viral product of the co-transfection experiments was wild-type virus. It is possible that the AdMMP8 genome replicated in the presence of Adwt300, but the wild-type genome was far more efficiently packaged into replication-competent virus.

Although at the time of termination of the experiment, MMP-8 protein was detected in all the tumors that had received AdMMP8 in the presence or absence of Adwt300 virus, clear evidence of collagenase activity was shown only in the tumors that received both AdMMP8 and Adwt300 virus. The explanation for this result remains unclear. The MMP-8 ELISA detects active and inactive MMP-8. Active MMP-8 is found in the supernatants of cells infected with AdMMP8, suggesting the ease with which MMP-8 can be activated in a tissue culture milieu. Although less noticeable than in the Adwt300/AdMMP8 group, decreased collagen staining within tumors was noted in the AdMMP8/Adcon group compared with controls, suggesting that MMP-8 had been active during the experiment. The zymography results were also obtained at one time point at the end of the experiment, when the Adwt300/AdMMP8 tumors were much smaller than the AdMMP8/Adcon tumors. It is possible that there is a shift in the protease/antiprotease balance during the course of tumor growth.

In conclusion, incorporation of the MMP-8 transgene within a replicating adenovirus might be a beneficial strategy for improving viral spread and improving oncolytic activity of a replicating adenovirus.

Materials and Methods

Cell Lines. A549 lung adenocarcinoma cells and BxPC-3 pancreatic cancer cells were obtained from American Type Culture Collection (Rockville, MD) and cultured in recommended media. Human 293 cells, which are embryonic kidney cells transformed by human adenovirus type 5, were obtained from Microbix (Toronto, Canada).

Total RNA isolation and cDNA synthesis. Total RNA was extracted from cells or tissues using the Micro-to-Midi Total RNA Purification System (Invitrogen, Carlsbad, CA). mRNA was reverse transcribed using avian myeloblastosis virus reverse transcriptase and oligo(dT)¹⁵ primers (Promega, Madison, WI). PCR primers were based on the human MMP-8 cDNA sequence (GenBank accession no. NM_002424) specifically to amplify full-length mRNA. Forward primer: 5'-AAAGAAAGCCAGGAGGGGTA-3'; reverse primer: 5'-CGGAGGACAGGTAGAATGGA-3'; predicted size of cDNA amplification product is 1,585 base pairs; coding sequence is 1,401 base pairs. A second pair of forward and reverse primers that produced a 639-base-pair fragment were used for semi-quantitative reverse transcription PCR. Forward primer: 5'-ATCTCACAGGGAGAGGCAGA-3'; reverse primer: 3'-CCTTGGGATAACCTTGCAGA-3'.

Adenoviruses and recombinant adenoviral construction. The wild-type virus used in these experiments was Adwt300. Ad-gal is a non-replicating adenovirus expressing the -gal transgene. AdMMP8, an E1- and E3-gene-deleted adenoviral DNA vector expressing the human MMP-8 under the control of the cytomegalovirus promoter with a simian virus 40 polyadenylation signal, was constructed using the AdEasy XL Adenoviral Vector System (Stratagene, La Jolla, CA). The human MMP-8 full-length cDNA product was ligated into pcDNA3.1/V5-His-TOPO-TA Expression Vector (Invitrogen), and the sequence was confirmed. In the AdEasy XL system, the human MMP-8 cDNA was cloned into pShuttle- cytomegalovirus, and once constructed the shuttle vector was linearized with PmeI and co-transformed into Escherichia coli BJ5183 electroporation-competent cells along with the adenoviral backbone vector. Electroporation was performed in 2.0-mm cuvettes at 2,500 V, 200 , and 25 microfarads in a Bio-Rad Gene Pulser electroporator. Transformants were selected for kanamycin resistance, and recombinants were subsequently identified by restriction digestion. Purified recombinant Ad plasmid DNA was digested with PacI to expose its inverted terminal repeats and was then used to transfect 293 cells for large-scale preparation. Adcon, a control non-replicating virus without a transgene, was constructed in the same way. Titration was performed by serial dilution on 293 cells as previously described.²⁰

Cell survival analysis. Cell survival was analyzed using a WST-1 assay (Roche, Indianapolis, IN) as described previously.³⁵ In brief, cells were seeded in 96-well plates and infected with dilutions of virus in the range 0.1–200 PFU/cell. Cells were analyzed 6 days after infection. WST-1 reagent (10 l) was added and the absorbance of the dye solution was measured in a multi-well spectrophotometer (FL600; Bio-Tek Instruments, Winooski, VT) at a wavelength of 450 nm.

Immunoblotting. A549 cells were infected with AdMMP8 at an MOI 200 for 24 hours. To prepare serum-free conditioned medium, cells were washed six times with serum-free medium and then re-suspended in fresh serum-free medium. After 24 hours, the culture supernatant was aspirated, spun at 1,500g to remove cellular debris, and concentrated using Biomax Ultrafree centrifugal filters (Millipore, Bedford, MA). Total protein concentration was determined by bicinchoninic acid protein assay (Pierce, Rockford, IL), equal amounts of denatured and reduced protein were separated on a 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis glycine gel, and proteins were transferred onto a polyvinylidene difluoride membrane (Immobilon-P; Millipore, Bedford, MA). The membranes were incubated with goat polyclonal anti-MMP-8 primary antibody (Santa Cruz Biotechnology, Santa Cruz, CA) at a 1:500 dilution and donkey anti-goat secondary antibody (Santa Cruz Biotechnology, Santa Cruz, CA) for 1 hour at room temperature. Antibody binding was visualized by enhanced chemiluminescence (Supersignal West Pico Chemiluminescent Substrate; Pierce, Rockford, IL).

Zymography. The analyses were performed as has been previously described.³² In brief, 25 g of unheated protein samples from concentrated serum-free conditioned medium or 100 g from tumor lysates was loaded on 10% Tris-glycine polyacrylamide gel with 0.1% gelatin incorporated as a substrate (Bio-Rad Life Science, Hercules, CA). After a room-temperature migration under non-reducing conditions, the gel was incubated twice for 15 minutes at room temperature in re-naturing buffer (Bio-Rad Life Science, Hercules, CA), equilibrating for 30 minutes at room temperature in the developing buffer, and incubated overnight at 37 °C in fresh developing buffer. The gel was stained with 0.5% Coomassie Brilliant Blue R in 50% methanol/10% acetic acid for 30 minutes and destained in 7.5% acetic acid/5% methanol. The clear bands represent gelatinase activity.

Diffusion assays. Virus diffusion was evaluated using BD Biocoat inserts pre-coated with collagen I on 3-m and 0.45-m membranes (BD Bioscience, San Diego, CA). In brief, 293 cells were plated on the base of 24-well plates (1.5 10⁵ cells per well). After 24 hours' incubation, an insert containing Ad-gal virus (MOI = 5) was placed in the well. After 24 hours' infection, the ability of the virus to spread through the insert was analyzed by in situ staining of cells for -gal and by -gal assay. In an additional experiment, the collagen I insert was treated with collagenase from clostridium histolyticum (100 g/ml) in minimum essential medium for 2 hours at 37 °C and then washed three times with phosphate-buffered saline (PBS) before application of the virus.

Albumin diffusion through collagen and fibrillar collagen matrix was evaluated. In brief, an insert containing bovine serum albumin (2 mg/ml) was placed in the 24-well plate, and PBS was plated on the base of the well. At the 24-hour, 48-hour, and 72-hour time points, the albumin concentration of both the inserts and the wells were measured by bicinchoninic acid protein assay. To determine the effects of MMP-8 on fibrillar collagen inserts, A549 cells were seeded on the fibrillar collagen membrane and infected with AdMMP8 or Adcon. After 7 days, wells were rinsed with PBS, and viral diffusion was evaluated.

In situ staining of cells for -galactosidase activity. The cell monolayers were rinsed twice with PBS and fixed with 0.5% glutaraldehyde solution for 15 minutes. The glutaraldehyde was removed and the staining solution (0.5 ml of 100 mm potassium ferricyanide, 5 mm potassium ferriocynide, 2 mm MgCl₂, X-Gal 1 mg/ml in PBS) was added and the cells were incubated at 37 °C for 3 hours until the control cells were visibly stained.

-Galactosidase assay. The growth medium was removed, and reporter lysis buffer was then added to the cells. The cell lysates were centrifuged at 14,000 rpm, and the supernatants were used for the -gal assay using kit reagents (-Gal Enzyme Assay System; Promega Madison, WI). The absorbance of the samples was read at 420 nm in a microplate reader.

Established tumor xenograft model. A549 or BxPC-3 cells (5 10⁶) were injected into the right flank of NCrNU-M nude mice (Taconic, Germantown, NY) that were 6–7 weeks of age. Tumor size (length and breadth) was measured twice weekly, and tumor volume was calculated on the basis of the following formula: length (breadth/2)² . Virus was injected when the tumors reached approximately 250 mm³ in size for A549 cells and 150 mm³ in size for the more slowly growing BxPC-3 cells. A total of 1 10⁹ PFU of equal amounts of two viruses were injected in a total volume of 50 l split into the four quadrants of the tumors. The viral combinations were as follows: Adwt300 plus non-replicating AdMMP8; Adwt300 plus non-replicating Adcon vector; AdMMP8 plus Adcon vector; vehicle control. Animals bearing A549 tumors were killed at day 42 and animals bearing BxPC-3 tumors at day 52, or before if tumor volumes were estimated to exceed 10% of an animal's body weight. All animal experiments were performed following the policies and procedures of the New York University Institutional Animal Care and Use Committee.

MMP-8 ELISA. Tumor lysates were evaluated for MMP-8 protein expression using a Biotrak MMP-8 ELISA system following the manufacturer's instructions (Amersham Biosciences, Piscataway, NJ).

Histology and immunohistochemistry. Tumor and lung tissue was fixed in formalin (10%), embedded in paraffin, and cut into 5-m sections. Representative sections were stained with hematoxylin and eosin and examined by light microscopy. Masson's trichrome staining for collagen was performed using aniline blue (Wax-it Histology Services, Vancouver, Canada). For immunohistochemistry, adenovirus antigen detection was performed using methods and reagents supplied by Dako ("DAKO ARK"; Dako, Carpentaria, CA). The primary mouse anti-adenovirus antibody (MAB8052; Chemicon, Temecula, CA) was first complexed with a biotinylated anti-mouse secondary antibody (Dako, Carpentaria, CA). Mouse serum was added, and the antibody complex was then applied to the sections and incubated for 15 minutes, rinsed in buffer, and incubated in streptavidin peroxidase. Diaminobenzidine/hydrogen peroxidase was used as the chromogen substance.

Statistical analyses. Data are reported as mean SEM. Analysis of variance with post hoc testing was used to compare multiple groups. For the animal experiments, Kaplan–Meier survival curves were generated on the basis of the pre-determined endpoint of time to tumor trebling after viral administration. Survival curves were compared using a log-rank test.

References

REFERENCES

1. Gene Therapy Clinical Trials Worldwide. (2007). Journal of Gene Medicine. Wiley, http://www.wiley.co.uk/genetherapy/clinical/.
2. Peng, Z (2005). Current status of gendicine in China: recombinant human Ad-p53 agent for treatment of cancers. Hum Gene Ther 16: 1016–1027. | Article | PubMed | ISI | ChemPort |
3. Kirn, DH (2006). The end of the beginning: Oncolytic Virotherapy achieves clinical proof-of-concept. Mol Ther 13: 237–238. | Article | PubMed | ISI | ChemPort |
4. Sauthoff, H, Hu, J, Maca, C, Goldman, M, Heitner, S, Yee, H et al. (2003). Intratumoral spread of wild-type adenovirus is limited after local injection of human xenograft tumors: virus persists and spreads systemically at late time points. Hum Gene Ther 14: 425–433. | Article | PubMed | ISI | ChemPort |
5. Vile, R, Ando, D and Kirn, D (2002). The oncolytic virotherapy treatment platform for cancer: unique biological and biosafety points to consider. Cancer Gene Ther 9: 1062–1067. | Article | PubMed | ISI | ChemPort |
6. Parato, KA, Senger, D, Forsyth, PA and Bell, JC (2005). Recent progress in the battle between oncolytic viruses and tumours. Nat Rev Cancer 5: 965–976. | Article | PubMed | ISI | ChemPort |
7. Harrison, D, Sauthoff, H, Heitner, S, Jagirdar, J, Rom, WN and Hay, JG (2001). Wild-type adenovirus decreases tumor xenograft growth, but despite viral persistence complete tumor responses are rarely achieved--deletion of the viral E1b-19-kD gene increases the viral oncolytic effect. Hum Gene Ther 12: 1323–1332. | Article | PubMed | ISI | ChemPort |
8. Doronin, K, Toth, K, Kuppuswamy, M, Ward, P, Tollefson, AE and Wold, WS (2000). Tumor-specific, replication-competent adenovirus vectors overexpressing the adenovirus death protein. J Virol 74: 6147–6155. | Article | PubMed | ISI | ChemPort |
9. Chiocca, EA, Abbed, KM, Tatter, S, Louis, DN, Hochberg, FH, Barker, F et al. (2004). A phase I open-label, dose-escalation, multi-institutional trial of injection with an E1B-Attenuated adenovirus, ONYX-015, into the peritumoral region of recurrent malignant gliomas, in the adjuvant setting. Mol Ther 10: 958–966. | Article | PubMed | ISI | ChemPort |
10. Galanis, E, Okuno, SH, Nascimento, AG, Lewis, BD, Lee, RA, Oliveira, AM et al. (2005). Phase I-II trial of ONYX-015 in combination with MAP chemotherapy in patients with advanced sarcomas. Gene Ther 12: 437–445. | Article | PubMed | ISI | ChemPort |
11. Kirn, D (2001). Clinical research results with dl1520 (Onyx-015), a replication-selective adenovirus for the treatment of cancer: what have we learned? Gene Ther 8: 89–98. | Article | PubMed | ISI | ChemPort |
12. Pupa, SM, Menard, S, Forti, S and Tagliabue, E (2002). New insights into the role of extracellular matrix during tumor onset and progression. J Cell Physiol 192: 259–267. | Article | PubMed | ISI | ChemPort |
13. Brown, E, McKee, T, diTomaso, E, Pluen, A, Seed, B, Boucher, Y et al. (2003). Dynamic imaging of collagen and its modulation in tumors in vivo using second-harmonic generation. Nat Med 9: 796–800. | Article | PubMed | ISI | ChemPort |
14. Hasty, KA, Jeffrey, JJ, Hibbs, MS and Welgus, HG (1987). The collagen substrate specificity of human neutrophil collagenase. J Biol Chem 262: 10048–10052. | PubMed | ISI | ChemPort |
15. Owen, CA, Hu, Z, Lopez-Otin, C and Shapiro, SD (2004). Membrane-bound matrix metalloproteinase-8 on activated polymorphonuclear cells is a potent, tissue inhibitor of metalloproteinase-resistant collagenase and serpinase. J Immunol 172: 7791–7803. | PubMed | ISI | ChemPort |
16. Herman, MP, Sukhova, GK, Libby, P, Gerdes, N, Tang, N, Horton, DB et al. (2001). Expression of neutrophil collagenase (matrix metalloproteinase-8) in human atheroma: a novel collagenolytic pathway suggested by transcriptional profiling. Circulation 104: 1899–1904. | Article | PubMed | ISI | ChemPort |
17. Swiderek, MS and Mannuzza, FJ (1997). Effets of ECM Proteins on Barrier Formation In Caco-2 Cells. Becton Dickinson Technical Bulletin #421.
18. Thorne, SH, Hermiston, T and Kirn, D (2005). Oncolytic virotherapy: approaches to tumor targeting and enhancing antitumor effects. Semin Oncol 32: 537–548. | Article | PubMed | ISI | ChemPort |
19. Peng, KW, Hadac, EM, Anderson, BD, Myers, R, Harvey, M, Greiner, SM et al. (2006). Pharmacokinetics of oncolytic measles virotherapy: eventual equilibrium between virus and tumor in an ovarian cancer xenograft model. Cancer Gene Ther 13: 732–738. | Article | PubMed | ISI | ChemPort |
20. Pipiya, T, Sauthoff, H, Huang, YQ, Chang, B, Cheng, J, Heitner, S et al. (2005). Hypoxia reduces adenoviral replication in cancer cells by downregulation of viral protein expression. Gene Ther 12: 911–917. | Article | PubMed | ISI | ChemPort |
21. Shen, BH and Hermiston, TW (2005). Effect of hypoxia on Ad5 infection, transgene expression and replication. Gene Ther 12: 902–910. | Article | PubMed | ISI | ChemPort |
22. Kuriyama, N, Kuriyama, H, Julin, CM, Lamborn, K and Israel, MA (2000). Pretreatment with protease is a useful experimental strategy for enhancing adenovirus-mediated cancer gene therapy. Hum Gene Ther 11: 2219–2230. | Article | PubMed | ISI | ChemPort |
23. Siller-Lopez, F, García-Bañuelos, J, Hasty, KA, Segura, J, Ramos-Márquez, M, Qoronfleh, MW et al. (2000). Truncated active matrix metalloproteinase-8 gene expression in HepG2 cells is active against native type I collagen. J Hepatol 33: 758–763. | Article | PubMed | ISI | ChemPort |
24. Siller-Lopez, F, Sandoval, A, Salgado, S, Salazar, A, Bueno, M, Garcia, J et al. (2004). Treatment with human metalloproteinase-8 gene delivery ameliorates experimental rat liver cirrhosis. Gastroenterology 126: 1122–1133; discussion 1949. | Article | PubMed | ISI | ChemPort |
25. Garcia-Banuelos, J, Siller-Lopez, F, Miranda, A, Aguilar, LK, Aguilar-Cordova, E and Armendariz-Borunda, J (2002). Cirrhotic rat livers with extensive fibrosis can be safely transduced with clinical-grade adenoviral vectors. Evidence of cirrhosis reversion. Gene Ther 9: 127–134. | Article | PubMed | ISI | ChemPort |
26. Harty, MW, Huddleston, HM, Papa, EF, Puthawala, T, Tracy, AP, Ramm, GA et al. (2005). Repair after cholestatic liver injury correlates with neutrophil infiltration and matrix metalloproteinase 8 activity. Surgery 138: 313–320. | Article | PubMed | ISI |
27. Balbin, M, Fueyo, A, Knäuper, V, Pendás, AM, López, JM, Jiménez, MG et al. (1998). Collagenase 2 (MMP-8) expression in murine tissue-remodeling processes. Analysis of its potential role in postpartum involution of the uterus. J Biol Chem 273: 23959–23968. | Article | PubMed | ISI | ChemPort |
28. Kim, JH, Lee, YS, Kim, H, Huang, JH, Yoon, AR and Yun, CO (2006). Relaxin expression from tumor-targeting adenoviruses and its intratumoral spread, apoptosis induction, and efficacy. J Natl Cancer Inst 98: 1482–1493. | PubMed | ChemPort |
29. Gueders, MM, Balbin, M, Rocks, N, Foidart, JM, Gosset, P, Louis, R et al. (2005). Matrix metalloproteinase-8 deficiency promotes granulocytic allergen-induced airway inflammation. J Immunol 175: 2589–2597. | PubMed | ISI | ChemPort |
30. Van Lint, P, Wielockx, B, Puimege, L, Noel, A, Lopez-Otin, C and Libert, C (2005). Resistance of collagenase-2 (matrix metalloproteinase-8)-deficient mice to TNF-induced lethal hepatitis. J Immunol 175: 7642–7649. | PubMed | ISI | ChemPort |
31. Agarwal, D, Goodison, S, Nicholson, B, Tarin, D and Urquidi, V (2003). Expression of matrix metalloproteinase 8 (MMP-8) and tyrosinase-related protein-1 (TYRP-1) correlates with the absence of metastasis in an isogenic human breast cancer model. Differentiation 71: 114–125. | Article | PubMed | ISI | ChemPort |
32. Montel, V, Kleeman, J, Agarwal, D, Spinella, D, Kawai, K and Tarin, D (2004). Altered metastatic behavior of human breast cancer cells after experimental manipulation of matrix metalloproteinase 8 gene expression. Cancer Res 64: 1687–1694. | Article | PubMed | ISI | ChemPort |
33. Balbin, M, Fueyo, A, Tester, AM, Pendás, AM, Pitiot, AS, Astudillo, A et al. (2003). Loss of collagenase-2 confers increased skin tumor susceptibility to male mice. Nat Genet 35: 252–257. | Article | PubMed | ISI | ChemPort |
34. Thorne, SH, Tam, BY, Kirn, DH, Contag, CH and Kuo, CJ (2006). Selective intratumoral amplification of an antiangiogenic vector by an oncolytic virus produces enhanced antivascular and anti-tumor efficacy. Mol Ther 13: 938–946. | Article | PubMed | ISI | ChemPort |
35. Sauthoff, H, Heitner, S, Rom, WN and Hay, JG (2000). Deletion of the adenoviral E1b-19kD gene enhances tumor cell killing of a replicating adenoviral vector. Hum Gene Ther 11: 379–388. | Article | PubMed | ISI | ChemPort |

Acknowledgments

This research was supported by National Institutes of Health (NIH) National Cancer Institute grant R01CA102053, NIH National Center for Research Resources grant MO1RR-00096, and a Veterans Administration Advanced Career Development Award (H.S.). The authors thank Robert J Schneider, New York University School of Medicine, for the gift of Adwt300 and Ad-gal. It is with great sadness that we note the untimely death of one of the authors, YaoQi Huang, who made important contributions to this work.