Factorial Design of Experiments to Optimize Multiple Protein Delivery

Apr 18, 2016 - *Phone: +1 412 624 7196. ... Protein therapies can potentially repair and regenerate the infarcted myocardium. ... This efficacy can be...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/journal/abseba

Factorial Design of Experiments to Optimize Multiple Protein Delivery for Cardiac Repair Hassan K. Awada,†,‡ Louis A. Johnson,¶ T. Kevin Hitchens,#,◆ Lesley M. Foley,◆ and Yadong Wang*,†,‡,§,∥,⊥,∇,○

Downloaded via NAGOYA UNIV on June 19, 2018 at 15:03:53 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.



Department of Bioengineering, ‡McGowan Institute for Regenerative Medicine, §Department of Chemical and Petroleum Engineering, ∥Department of Surgery, ⊥Department of Mechanical Engineering and Materials Science, #Department of Neurobiology, ∇Vascular Medicine Institute, ○Clinical and Translational Science Institute, and ◆Animal Imaging Center, University of Pittsburgh, 4200 Fifth Avenue, Pittsburgh, Pennsylvania 15260, United States ¶ SnapDat Inc., 733 West Foster Avenue, State College, Pennsylvania 16801, United States S Supporting Information *

ABSTRACT: Myocardial infarction (MI) is a major cardiovascular disease responsible for millions of deaths annually. Protein therapies can potentially repair and regenerate the infarcted myocardium. However, because of the short half-lives of proteins in vivo, their low retention at the target tissue, and the lack of spatiotemporal cues upon injection, the efficacy of protein therapy can be limited. This efficacy can be improved by utilizing controlled release systems to overcome shortcomings associated with a direct bolus injection. Equally important is the determination of an optimal combination of different proteins having distinct roles in cardiac function and repairs to prevent or reverse the multiple pathologies that develop after infarction. In this work, we used a rat MI model to test a combination of potentially complementary proteins: tissue inhibitor of metalloproteinases 3 (TIMP-3), interleukin-10 (IL-10), basic fibroblast growth factor (FGF-2), and stromal cell-derived factor 1 alpha (SDF-1α). To achieve controlled and timed release of the proteins per their physiologic cues during proper tissue repair, we used a fibrin gel-coacervate composite. TIMP-3 and IL-10 were encapsulated in fibrin gel to offer early release, while FGF-2 and SDF-1α were encapsulated in heparinbased coacervates and distributed in the same fibrin gel to offer sustained release. We utilized a powerful statistical tool, factorial design of experiments (DOE), to refine this protein combination based on its improvement of ejection fraction 4 weeks after MI. We found that TIMP-3, FGF-2, and SDF-1α demonstrated significant contributions toward improving the ejection fraction, while the IL-10’s effect was insignificant. The results also suggested that the higher doses tested for TIMP-3, FGF-2, and SDF-1α had greater benefit on function than lower doses and that there existed slight antagonism between TIMP-3 and FGF-2. Taken together, we conclude that factorial DOE can guide the evolution of multiple protein therapies in a small number of runs, saving time, money, and resources for finding the optimal dose and composition. KEYWORDS: factorial design, myocardial infarction, controlled release, coacervate, fibrin gel, proteins



INTRODUCTION

injection, the efficacy of protein therapy can be limited. To overcome these challenges, we designed a fibrin gel-coacervate composite to load, protect, and spatiotemporally release proteins. We have previously demonstrated the effectiveness of the fibrin gel-coacervate composite at sequential release of proteins for therapeutic angiogenesis post-MI.3 Here, the complex coacervate self-assembles by ionic interaction between a synthetic polycation, poly(ethylene argininylaspartate diglyceride) (PEAD), a natural polyanion heparin, and heparin-

Myocardial infarction (MI) affects approximately 8 million Americans and costs the United States billions of dollars annually in medical costs and lost productivity.1 MI can cause massive death of cardiomyocytes, excessive inflammation, ventricular dilation, interstitial fibrosis, and adverse cardiac remodeling, culminating in the development of congestive heart failure. Tissue repair and regeneration of the heart involve the signaling of many proteins to induce changes in cellular function. Protein-based therapies can potentially halt the progression of MI pathologies and set the infarcted myocardium on a path to recovery.2 However, because of the short half-lives of many proteins in vivo, their low retention at the target tissue, and the lack of spatiotemporal cues upon © 2016 American Chemical Society

Received: March 12, 2016 Accepted: April 18, 2016 Published: April 18, 2016 879

DOI: 10.1021/acsbiomaterials.6b00146 ACS Biomater. Sci. Eng. 2016, 2, 879−886

Article

ACS Biomaterials Science & Engineering

(HSV-1) and as a method to investigate the effects of different processing parameters for a tissue-engineering scaffold.15,16 Fractional factorial designs allow us to build statistical models using a small number of subjects. Such models can identify proteins important for therapeutic outcome, potential protein interactions, optimal protein doses, and optimal protein combinations. Thus, we set out to perform the initial test of these proteins in a rat MI model using cardiac function 4 weeks postop as a screening parameter.

binding proteins. The coacervate can release the loaded proteins over several weeks in vitro and in vivo.3−6 In this study, we explored the use of design of experiments (DOE) to streamline the research on controlled and timed release of four complementary proteins that are relatively distinct in the roles they play for cardiac function. Tissue inhibitor of metalloproteinases 3 (TIMP-3), interleukin-10 (IL10), fibroblast growth factor-2 (FGF-2), and stromal cellderived factor 1 alpha (SDF-1α) are proteins with therapeutic potential in cardiac repair and regeneration (Figure 1). TIMP-3



MATERIALS AND METHODS

Two-Level Half Fractional Factorial Design. We formulated a two-level half fractional factorial design to select the combination of proteins and their doses that are the most effective at recovering cardiac function post-MI. The two levels refer to upper and lower doses for each protein and were chosen based on previous experiments and literature review. The half fractional factorial design means half of the total runs were performed. For this design, we can estimate all main effects and some 2-factor interactions, which is reasonable in practice to evaluate the significance of each protein in the combination and find the corresponding optimal dose.17 Using the design formula 2(k‑p) with k = 4 factors, and p = 1 (for half fractional factorial), we studied 23 = 8 dose groups. From our previous studies, preliminary results, and the literature, we selected the upper and lower doses (one-fifth of upper dose) for each protein to create a range that encompasses commonly used doses in controlled delivery approaches when scaled to a rat model.9,12,14,18−25 The high doses for FGF-2, SDF-1α, IL-10 (PeproTech, Rocky Hill, NJ), and TIMP-3 (R&D Systems, Minneapolis, MN) were 3, 3, 2, and 4 μg, respectively; while the lower doses are 0.6, 0.6, 0.4, and 0.8 μg, respectively (Table 1). Here, we refer to the protein dose as the amount of each protein

Figure 1. FGF-2, SDF-1α, IL-10, and TIMP-3 have relatively distinct but complementary cardiac functions. FGF-2 promotes angiogenesis by endothelial sprouting and pericyte recruitment and protects cardiomyocytes after MI. SDF1-α has a critical role of recruiting cardiac, endothelial, hematopoietic, and mesenchymal stem and progenitor cells to the infarcted area, while also promoting angiogenesis and cardiomyocyte survival. IL-10 reduces inflammation by inhibiting the infiltration of immune cells into the myocardium and also reduces cardiomyocyte death. TIMP-3 helps preserve the cardiac ECM structure by inhibiting the activity and reducing the expression levels of MMPs and also promotes anti-inflammatory activities and cardiomyocyte survival.

Table 1. Treatment Groups According to Two-Level Half Fractional Factorial Design and the Corresponding Ejection Fraction (EF%) Obtained by MRI protein dose group dose group dose group dose group dose group dose group dose group dose group sham

inhibits the activity of matrix metalloproteinases (MMPs) that degrade the extracellular matrix (ECM).7 Therefore, TIMP-3 presence might be beneficial early after infarction when excessive ECM degradation occurs. IL-10 is an antiinflammatory cytokine known to reduce the infiltration of inflammatory cells into the infarcted myocardium, which is usually overly active after MI.8,9 FGF-2 is a potent stimulator of angiogenesis.10,11 SDF-1α can trigger strong chemotaxis of stem cells toward the infarct region.12−14 Therefore, we encapsulated TIMP-3 and IL-10 in fibrin gel for early release, while embedding FGF-2 and SDF-1α in heparin-based coacervates and distributing them in the same fibrin gel for sustained release (Figure 2). The endogenous biological system and tissue repair process are intrinsically complex with many proteins involved, some of which interact with each other. Considering the four proteins of interest in our study and the combinations of controls that can result from them, it is cost-prohibitive to test all possible combinations and doses. To address this challenge, we used factorial DOE, a powerful statistical method, to reduce study groups. Fractional factorial designs are commonly utilized in scientific studies and industrial applications. However, they have not been taken advantage of as commonly in biomedical research. These designs have been utilized previously to study drug combinations for treating Herpes simplex virus type 1

1 2 3 4 5 6 7 8

FGF-2

SDF-1α

IL-10

TIMP-3

EF %

SD

3 μg 3 μg 3 μg 3 μg 0.6 μg 0.6 μg 0.6 μg 0.6 μg

3 μg 3 μg 0.6 μg 0.6 μg 3 μg 3 μg 0.6 μg 0.6 μg

2 μg 0.4 μg 2 μg 0.4 μg 2 μg 0.4 μg 2 μg 0.4 μg

4 μg 0.8 μg 0.8 μg 4 μg 0.8 μg 4 μg 4 μg 0.8 μg

62.3 56.8 50.7 59.4 44.9 58.9 51.6 41.1 70.2

1 1 4.3 3.6 3.6 3.4 1.9 2.9 2.2

embedded in the delivery vehicle. For example, for the average rat used in this study, a high dose of 3 μg would correlate to approximately 15 μg/kg body weight (i.e., 3 μg/0.2 kg rat). TIMP-3 and IL-10 were encapsulated in fibrin gel, while FGF-2 and SDF-1α were encapsulated in heparin-based coacervates and distributed in the same fibrin gel (Figure 2). With n = 3 per dose group and a sham group, we used 27 rats for this initial-stage study. Using Minitab statistical software (State College, PA), this design provided a data collection plan with 8 groups of varying protein doses to be tested (Table 1). Each group with varying protein doses was tested in a rat model of acute MI. The key outcome measurement of cardiac function was the left ventricle (LV) ejection fraction (EF%) computed using cardiac magnetic resonance imaging (MRI) at 4 weeks post-MI. Ejection fractions were analyzed in Minitab software to provide estimates of the relative importance of each protein and its optimal dose in the context of improving cardiac function after MI. Rat Acute Myocardial Infarction Model. The University of Pittsburgh Institutional Animal Care and Use Committee (IACUC) approved all animal studies. MI and injections were performed as previously described.26 Briefly, 6−7 week old (175−225g) male 880

DOI: 10.1021/acsbiomaterials.6b00146 ACS Biomater. Sci. Eng. 2016, 2, 879−886

Article

ACS Biomaterials Science & Engineering

the following parameters: TR/TE = 9.0/3.0 ms, 40 × 40 mm FOV, 256 × 256 matrix, FA = 10°, and 200 repetitions. Then, 10−12 slices were collected to cover the area between the heart apex to the mitral valves with 1.5 mm slice thickness with common navigator slice. Endsystolic and end-diastolic phases were identified for each subject and the LV cavity manually traced using NIH ImageJ to determine LV endsystolic (ESV) and end-diastolic (EDV) volumes (Figure 3) (Table S1). These volumes were used to compute the ejection fraction as EF % = [(EDV − ESV)/EDV] × 100%.

Figure 2. Controlled release system comprised a fibrin gel embedding TIMP-3 and IL-10 to obtain early release and FGF-2-loaded and SDF1α-loaded coacervates distributed within the same gel to achieve sustained late release. The coacervate was formed through electrostatic interactions by combining FGF-2 or SDF-1α with heparin and then with the polycation, PEAD.

Figure 3. MRI was performed with an average of 10 contiguous slices, at a thickness of 1.5 mm covering the distance from the heart apex at level 1 to the mitral valve at level 10 in the representative images. Endsystolic and end-diastolic phases were identified for each subject and the LV cavity manually traced to determine LV end-systolic (ESV) and end-diastolic (EDV) volumes. The ejection fraction (EF %) was calculated from ESV and EDV values. The infarct reduces the contractility of the heart as seen at levels 1 and 5.

Sprague−Dawley rats (Charles River Laboratories, Wilmington, MA) were anesthetized first, then maintained with 2% isoflurane at 0.3 L/ min (Butler Schein, Dublin, OH), intubated, and connected to a mechanical ventilator to support breathing during surgery. The body temperature was maintained at 37 °C by a warm pad. The ventral side was shaved, and a small incision was made through the skin. Forceps, scissors, and Q-tips were used to dissect through the skin, muscles, and ribs. Once the heart was visible, the pericardium was torn. MI was induced by permanent ligation of the left anterior descending (LAD) coronary artery using a 6−0 polypropylene suture (Ethicon, Bridgewater, NJ). Infarct was confirmed by macroscopic observation of a change in color from bright red to light pink in the area below the ligation suture. Five minutes after the induction of MI, composite gel solutions with different protein doses (Table 1) were injected intramyocardially at 3 equidistant points around the infarct border zone using a 31-gauge needle (BD, Franklin Lakes, NJ). One hundred microliters of fibrin gel-coacervate vehicle was prepared as follows: 18 μL of coacervate solution (PEAD/heparin/ protein mass ratio at 50:10:1) containing an appropriate amount of FGF-2 and SDF-1α as outlined in Table 1, 65 μL of 20 mg/mL fibrinogen, 10 μL of solution containing heparin and an appropriate amount of TIMP-3 and IL-10 as outlined in Table 1, and 5 μL of 1 mg/mL aprotonin (Sigma-Aldrich, St. Louis, MO), and last, 2 μL of 1.5 mg/mL thrombin (Sigma-Aldrich, St. Louis, MO) was added, and the total solution was injected shortly before gelation occurred, approximately 50 s after mixing. The chest was closed, and the rat was allowed to recover. In the case of sham controls, the rats underwent surgery in which the chest was open, heart exposed, and pericardium torn, but no MI was induced. After 4 weeks, all animals (n = 27) were imaged using MRI and sacrificed. Cardiac MRI. MRI was preformed using a Bruker Biospec 4.7-T 40 cm scanner equipped with a 12 cm shielded gradient set, a 72 mm transmit RF coil (Bruker Biospin, Billerica MA), and a 4-channel rat cardiac receive array (Rapid MR International, Columbus, OH). Rats were induced with isoflurane, intubated, and ventilated at 1 mL/100g of body weight and maintained at 2% isoflurane in 2:1 O2/N2O gas mixture at 60 BPM. During the MRI procedure, rats were continually monitored, and rectal temperature was maintained at 37 °C with warm air (SA Instruments, Stony Brook, NY). Following pilot scans, rats were imaged using a self-gated cine FLASH sequence (IntraGate) with

Statistical Analysis. Results are presented as the means ± standard deviations (SD). Minitab software (State College, PA) was used for statistical analysis. Statistical significance was set at p < 0.05.



RESULTS AND DISCUSSION The use of TIMP-3, IL-10, FGF-2, or SDF-1α to promote cardiac repair post-MI is well supported by the literature.8,12,13,22,23,25,27−30 However, to our knowledge, they have never been tested in combination before. Therefore, the significance of each protein in the combination within the context of improving cardiac function and repair after MI is unknown. In addition, such a combination has not been evaluated using controlled delivery. Finally, there is no existing literature on optimal dose for each protein in free or controlled release form. Therefore, we utilized a two-level half fractional factorial design to build a statistical model with a small number of runs, saving time, resources, and money. The design of the controlled release for this study is based on our previous studies using the fibrin gel and coacervate, and also on literature that supports our claim that TIMP-3 and IL10, if used in therapy, might be beneficial early after MI to counter early ECM degradation and inflammation, while FGF-2 and SDF-1α should be provided in a more sustained manner because angiogenesis and stem cell homing require prolonged signaling.2,31−36 To this end, we used our heparin-coacervate as the prolonged release system as it is capable of releasing heparin-binding proteins over weeks of time for FGF-2 and SDF-1α.6,37 We have demonstrated the ability of the fibrin-coacervate composite gel to provide early vs delayed release in a previous study, where vascular endothelial growth factor (VEGF) was embedded directly in fibrin gel, while the platelet-derived 881

DOI: 10.1021/acsbiomaterials.6b00146 ACS Biomater. Sci. Eng. 2016, 2, 879−886

Article

ACS Biomaterials Science & Engineering growth factor (PDGF) was embedded within coacervates and distributed in the same gel.3 In the mentioned study, we demonstrated using an in vitro assay that this composite gel releases VEGF early and almost completely within 1 week, while PDGF release was delayed and sustained for at least 3 weeks. In this study, we speculate that the release of TIMP-3 and IL-10 from the fibrin gel directly would have happened in a similar fashion to that of VEGF, while FGF-2 and SDF-1α would release from the coacervate in a more sustained fashion similar to that of PDGF. The focus of this study was to use factorial DOE as a proof of concept to show its potential in identifying significant therapeutic proteins in multiple protein therapy and as a method that can be used to optimize protein doses. Controlled release was utilized in this study and justified as a method proven to be advantageous over bolus injections in order to achieve maximum benefit for MI treatment. The release kinetics of the four proteins were not the main focus of this study and will be published in a separate manuscript. In this optimization study, we generated, using statistical software, a table organizing 8 groups with different upper and lower doses for each protein, to be tested for efficacy on ejection fraction (EF %) in a rat MI model (Table 1). MRI images were used to compute ESV, EDV, and EF values for all 27 rats in the study (Table S1). The resulting mathematical model of the system provided valuable insight on the relative significance of each protein in the combination on improving cardiac function and the optimal dose needed to benefit for cardiac repair and regeneration. This factorial design has a resolution IV, meaning that no main effects are aliased with any other main effect or 2-factor interactions but that some 2-factor interactions are aliased with other 2-factor interactions and that main effects are aliased with 3-factor interactions. Aliasing means the loss of the ability to estimate some effects and/or interactions, which is the price paid for not running a full factorial design. However, a reasonable and common assumption in such designs is that higher-order interactions are assumed to be negligible because practically they are less likely to be important than lower-order interactions. The regression analysis assumes that the measurement and experimental errors in the data are normally distributed, random over time, and of equal variance at all levels of the response. These three assumptions were confirmed using analysis of residuals (Figure S1). Initial power calculations using a sample size of n = 3 per dose group and estimated standard error of 3 showed an 80% power of detecting an effect as small as 3.7% in EF and 90% power of detecting an effect as small as 4.2% in EF. The Rsquared of our regression model was 90%, indicating a good fit of the resultant model to the data. In ANOVA, the sums of squares help identify the level of contribution of various factors to changes in the outcome measurement, EF % in our study. The means of squares estimate population variance and can be calculated by dividing the respective sum of squares by the degrees of freedom. These means of squares are used to determine which factors are significant. The F-statistic is calculated by dividing the mean of squares of each factor by the mean of squares of error. The p values can be calculated with the aid of statistical software from the F-statistic, degrees of freedom of numerators (the treatment factors), and degrees of freedom of denominators (error). Results demonstrated significant main effects of TIMP-3, FGF-2, and SDF-1α (p < 0.001), while suggesting little effect of

IL-10 (p = 0.273) on improvement of cardiac function (Figure 4). This means that, within the context of improving EF %

Figure 4. Analysis of variance results show the relative significance of each of the 4 proteins: TIMP-3, IL-10, FGF-2, SDF-1α, and some of the 2-way protein interactions on improvement of ejection fraction (EF %).

using the controlled delivery of this combination, TIMP-3, FGF-2, and SDF-1α were beneficial for improving cardiac function, while IL-10’s effect was insignificant, suggesting that it can be removed from the combination. IL-10 is a potent anti-inflammation cytokine that is reported to improve cardiac function and angiogenesis, and reduce scar size and fibrosis after MI.8,9,23,38 However, a study using IL-10 knockout models deduced that it does not affect ventricular remodeling critically.39 The insignificant effect of IL-10 reported in our study, within the context of the protein combination, might have been influenced by two factors. The first factor is that IL-10’s role in improving cardiac function might have been diminished by the presence of TIMP-3, which itself has been reported to have anti-inflammatory effects by inhibiting the TNF-α-converting enzyme (TACE), the enzyme activator of TNF-α.40,41 TNF-α is a pro-inflammatory factor, involved in inducing inflammatory cell invasion of infarcted myocardium, MMP production, and cell apoptosis.40−42 The second factor is that an appropriate level of inflammatory response is necessary after MI to phagocytose dead cells and their debris, and that the presence of IL-10 alongside TIMP-3, within the protein combination, might have rendered IL-10 redundant and did not add additional benefit toward improving cardiac function. TIMP-3 had the greatest main effect on improvement of EF % accounting for 43% of the total sum of squares (598/1395), followed by FGF-2 accounting for 32% (440/1395), then SDF1α accounting for 12.5% (174/1395) (Figure 4). Together, the main effects of these proteins dominate the system and account for 87.5% of the total sum of squares, suggesting that the combined individual effects of these 3 proteins are responsible for 87.5% of the change in EF %, while higher-order protein interactions and error account for 12.5% of that change (Figure 4). Although the half fractional factorial design utilized in this study combined each of the 2-way protein interactions with one other 2-way interaction, this limitation does not affect the validity of conclusions and the significance of our findings as 882

DOI: 10.1021/acsbiomaterials.6b00146 ACS Biomater. Sci. Eng. 2016, 2, 879−886

Article

ACS Biomaterials Science & Engineering

Figure 5. (A) Main effects plot shows the individual effect of each protein on EF % from respective lower to upper doses. (B) The interaction between TIMP-3 and FGF-2, nearly significant at p = 0.076, suggests slight antagonism between the 2 proteins.

models.22,25 FGF-2’s potency at triggering angiogenesis through endothelial cell proliferation and sprouting is well documented.10,11 Our previous studies demonstrate that the FGF2 coacervate can induce persistent neovasculature formation and cardiac repair after MI.19,45 Therefore, FGF-2 presence within our combination might have helped ameliorate the hypoxic MI environment by revascularizing the infarct region and in turn improved cardiac function significantly. As for SDF1α, it has been shown to recruit different types of stem cells to the infarcted myocardium, which might support the cardiac repair through paracrine signaling or potential differentiation into vascular cells or cardiomyocytes.12,13,29,36 We have shown previously that scaffolds coated with SDF-1α coacervates can induce migration and infiltration of endothelial and mesenchymal progenitors.37 In addition, all three proteins, TIMP-3, FGF-2, and SDF-1α, have been suggested to support cardiomyocyte survival.25,31,46−48 These features of the complementary proteins employed in our approach might have contributed to the observed improvement in cardiac function. Taking the statistical results of our study and the cost of each protein into consideration, investigators can, for example, move forward with FGF-2, SDF-1α, and TIMP-3 in the protein combination. They can use a dose of 3 μg for each, thereby keeping the upper doses of FGF-2 and SDF-1α, while reducing TIMP-3’s upper dose slightly from 4 μg to 3 μg, and then employ more resources to understand the improvements at the cell and tissue levels to elucidate repair mechanisms. Reducing the regression model to include only the 3 significant main effects and the interaction term (of TIMP-3 and FGF-2) leads to a new model to predict EF % (Figure 6A). The contour plot constructed from this model predicts an EF % of approximately 62% if we utilize FGF-2, SDF-1α, and TIMP-3 at 3 μg each in the designed scheme within the fibrin gel-coacervate composite (Figure 6B). Because a two-level test will not reveal a plateau or oversaturation due to excessive dose, it is possible that higher doses of FGF-2, SDF-1α, and TIMP-3 than what was tested in

the results confirm the negligibility of higher-order protein interactions on the improvement of EF % in our combination where they only account for less than 13% of changes in EF %. This also supports our notion that the therapeutic proteins in this study are complementary with relatively distinct roles in cardiac function, which explains the limited value of the interaction terms toward improving EF %. As for the optimal protein doses, the most effective group of the 8 groups was group 1 restoring EF to 62%, which is closest to the average of 3 sham controls at 70% (Table 1). We observe, from the main effects plot, that all of the estimates for FGF-2, SDF-1α, and TIMP-3 have significant positive slopes, indicating that improvement of cardiac function, in this study, can be best achieved by using the higher doses of FGF-2, SDF1α, and TIMP-3 (Figure 5A). In addition to minimizing the required number of runs, another benefit of factorial experimentation is the ability to estimate 2-way interactions. Although the effects of the 2-way protein interactions that this fractional factorial design can estimate were all insignificant, the interaction between FGF-2 and TIMP-3 was worthy of consideration as it was nearly significant at p = 0.076 (Figure 4). The interaction plot suggests slight antagonism between the 2 proteins (Figure 5B). The improvement in EF % seen when FGF-2 is increased in dose is smaller when TIMP-3 is at the high dose than it is at the low dose (the slopes of the 2 lines are not equal) (Figure 5B). Physiologically, the 2 proteins might be fulfilling some similar functions, and therefore, their effects are overlapping when both are at the high dose. TIMP-3’s ability to block MMP’s catalytic domain and prevent its activation allows it to contribute to reducing the degradation of the cardiac ECM.7,43,44 Making TIMP-3 bioavailable early after MI as provided by the fibrin gelcoacervate composite may have reduced LV adverse remodeling and dilation, leading to an improvement in the cardiac function. Studies that delivered TIMP-3 by collagen or hyaluronan gels showed an increase of ejection fractions and a decrease of LV dilation and infarct size in rat and pig 883

DOI: 10.1021/acsbiomaterials.6b00146 ACS Biomater. Sci. Eng. 2016, 2, 879−886

Article

ACS Biomaterials Science & Engineering

effect. Upper doses of the 3 significant proteins caused a larger improvement in ejection fraction after MI. Therefore, we demonstrate that factorial design of experiments is a powerful tool in designing a treatment strategy for complex diseases such as MI. The factorial designs allow us to understand the relative significance of each drug, the drug−drug interactions, and guide the next step in dose optimization.



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsbiomaterials.6b00146. MRI data showing the ESV, EDV, and EF values for the 27 rats in the dose groups used in the study; and model regression fit analysis (PDF)



AUTHOR INFORMATION

Corresponding Author

*Phone: +1 412 624 7196. E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors thank Dr. Sanjeev Shroff for encouraging the publication of this study focusing on the application of DOE in biomedical engineering. This work was supported by the biomechanics in regenerative medicine (BiRM) T32 training program (grant # 5T32EB003392-09) of the National Institutes of Health, the American Heart Association (grant # 12EIA9020016), and the National Science Foundation (grant # DMR-1005766).



Figure 6. (A) Modified regression model, after removing IL-10. (B) A contour plot predicts the value of EF % upon choosing doses for TIMP-3 and FGF-2 among a range of values, while fixing the SDF-1α dose at 3 μg.

REFERENCES

(1) Go, A. S.; Mozaffarian, D.; Roger, V. L.; Benjamin, E. J.; Berry, J. D.; Blaha, M. J.; Dai, S.; Ford, E. S.; Fox, C. S.; Franco, S.; Fullerton, H. J.; Gillespie, C.; Hailpern, S. M.; Heit, J. A.; Howard, V. J.; Huffman, M. D.; Judd, S. E.; Kissela, B. M.; Kittner, S. J.; Lackland, D. T.; Lichtman, J. H.; Lisabeth, L. D.; Mackey, R. H.; Magid, D. J.; Marcus, G. M.; Marelli, A.; Matchar, D. B.; McGuire, D. K.; Mohler, E. R., 3rd; Moy, C. S.; Mussolino, M. E.; Neumar, R. W.; Nichol, G.; Pandey, D. K.; Paynter, N. P.; Reeves, M. J.; Sorlie, P. D.; Stein, J.; Towfighi, A.; Turan, T. N.; Virani, S. S.; Wong, N. D.; Woo, D.; Turner, M. B. American Heart Association Statistics, C.; Stroke Statistics, S., Heart disease and stroke statistics–2014 update: a report from the American Heart Association. Circulation 2014, 129 (3), e28− e292. (2) Awada, H. K.; Hwang, M. P.; Wang, Y. Towards comprehensive cardiac repair and regeneration after myocardial infarction: Aspects to consider and proteins to deliver. Biomaterials 2016, 82, 94−112. (3) Awada, H. K.; Johnson, N. R.; Wang, Y. Sequential delivery of angiogenic growth factors improves revascularization and heart function after myocardial infarction. J. Controlled Release 2015, 207, 7−17. (4) Awada, H. K.; Johnson, N. R.; Wang, Y. Dual delivery of vascular endothelial growth factor and hepatocyte growth factor coacervate displays strong angiogenic effects. Macromol. Biosci. 2014, 14 (5), 679−86. (5) Johnson, N. R.; Wang, Y. Coacervate delivery systems for proteins and small molecule drugs. Expert Opin. Drug Delivery 2014, 11 (12), 1829−32. (6) Chen, W. C.; Lee, B. G.; Park, D. W.; Kim, K.; Chu, H.; Kim, K.; Huard, J.; Wang, Y. Controlled dual delivery of fibroblast growth

this study will be more beneficial. Thus, a follow-up study can implement a three-level fractional factorial design to test three dose levels for each of the proteins, categorized as upper, medium, and lower doses, with the medium doses at the 3 μg level which gave the highest EF % recovery in this study. This would allow us to model the curved response surface caused by oversaturation.



CONCLUSIONS We developed a protein combination strategy for the repair of infarcted myocardium using TIMP-3, FGF-2, SDF-1α, and IL10. The delivery vehicle is a composite of fibrin gel and heparin-based coacervates, where complementary proteins are embedded differently to achieve spatiotemporal release. TIMP3 and IL-10 were embedded in the fibrin gel for early release, while FGF-2 and SDF-1α were encapsulated within heparinbased coacervates and distributed in the same gel for delayed releases. We found, by the utilization of a two-level half fractional factorial design, that TIMP-3, FGF-2, and SDF-1α significantly contributed to the changes in ejection fraction at 4 weeks after MI in a rat model, while IL-10 had an insignificant 884

DOI: 10.1021/acsbiomaterials.6b00146 ACS Biomater. Sci. Eng. 2016, 2, 879−886

Article

ACS Biomaterials Science & Engineering factor-2 and Interleukin-10 by heparin-based coacervate synergistically enhances ischemic heart repair. Biomaterials 2015, 72, 138−51. (7) Visse, R.; Nagase, H. Matrix metalloproteinases and tissue inhibitors of metalloproteinases: structure, function, and biochemistry. Circ. Res. 2003, 92 (8), 827−39. (8) Dhingra, S.; Sharma, A. K.; Arora, R. C.; Slezak, J.; Singal, P. K. IL-10 attenuates TNF-alpha-induced NF kappaB pathway activation and cardiomyocyte apoptosis. Cardiovasc. Res. 2009, 82 (1), 59−66. (9) Manukyan, M. C.; Alvernaz, C. H.; Poynter, J. A.; Wang, Y.; Brewster, B. D.; Weil, B. R.; Abarbanell, A. M.; Herrmann, J. L.; Crowe, B. J.; Keck, A. C.; Meldrum, D. R. Interleukin-10 protects the ischemic heart from reperfusion injury via the STAT3 pathway. Surgery 2011, 150 (2), 231−9. (10) Cochain, C.; Channon, K. M.; Silvestre, J. S. Angiogenesis in the infarcted myocardium. Antioxid. Redox Signaling 2013, 18 (9), 1100− 13. (11) Formiga, F. R.; Tamayo, E.; Simon-Yarza, T.; Pelacho, B.; Prosper, F.; Blanco-Prieto, M. J. Angiogenic therapy for cardiac repair based on protein delivery systems. Heart Failure Rev. 2012, 17 (3), 449−73. (12) Saxena, A.; Fish, J. E.; White, M. D.; Yu, S.; Smyth, J. W.; Shaw, R. M.; DiMaio, J. M.; Srivastava, D. Stromal cell-derived factor-1alpha is cardioprotective after myocardial infarction. Circulation 2008, 117 (17), 2224−31. (13) Segers, V. F.; Tokunou, T.; Higgins, L. J.; MacGillivray, C.; Gannon, J.; Lee, R. T. Local delivery of protease-resistant stromal cell derived factor-1 for stem cell recruitment after myocardial infarction. Circulation 2007, 116 (15), 1683−92. (14) Zhang, G.; Nakamura, Y.; Wang, X.; Hu, Q.; Suggs, L. J.; Zhang, J. Controlled release of stromal cell-derived factor-1 alpha in situ increases c-kit+ cell homing to the infarcted heart. Tissue Eng. 2007, 13 (8), 2063−71. (15) Jaynes, J.; Ding, X.; Xu, H.; Wong, W. K.; Ho, C. M. Application of fractional factorial designs to study drug combinations. Stat Med. 2013, 32 (2), 307−18. (16) Saraf, A.; Baggett, L. S.; Raphael, R. M.; Kasper, F. K.; Mikos, A. G. Regulated non-viral gene delivery from coaxial electrospun fiber mesh scaffolds. J. Controlled Release 2010, 143 (1), 95−103. (17) Wu, C.-F.; Hamada, M. Experiments: Planning, Analysis, and Parameter Design Optimization; Wiley: New York, 2009, 2nd edition. (18) Brudno, Y.; Ennett-Shepard, A. B.; Chen, R. R.; Aizenberg, M.; Mooney, D. J. Enhancing microvascular formation and vessel maturation through temporal control over multiple pro-angiogenic and pro-maturation factors. Biomaterials 2013, 34 (36), 9201−9. (19) Chu, H.; Chen, C. W.; Huard, J.; Wang, Y. The effect of a heparin-based coacervate of fibroblast growth factor-2 on scarring in the infarcted myocardium. Biomaterials 2013, 34 (6), 1747−56. (20) Hwang, H.; Kloner, R. A. Improving regenerating potential of the heart after myocardial infarction: factor-based approach. Life Sci. 2010, 86 (13−14), 461−72. (21) Hwang, H.; Kloner, R. A. The combined administration of multiple soluble factors in the repair of chronically infarcted rat myocardium. J. Cardiovasc. Pharmacol. 2011, 57 (3), 282−6. (22) Eckhouse, S. R.; Purcell, B. P.; McGarvey, J. R.; Lobb, D.; Logdon, C. B.; Doviak, H.; O’Neill, J. W.; Shuman, J. A.; Novack, C. P.; Zellars, K. N.; Pettaway, S.; Black, R. A.; Khakoo, A.; Lee, T.; Mukherjee, R.; Gorman, J. H.; Gorman, R. C.; Burdick, J. A.; Spinale, F. G. Local hydrogel release of recombinant TIMP-3 attenuates adverse left ventricular remodeling after experimental myocardial infarction. Sci. Transl. Med. 2014, 6 (223), 223ra21. (23) Krishnamurthy, P.; Rajasingh, J.; Lambers, E.; Qin, G.; Losordo, D. W.; Kishore, R. IL-10 inhibits inflammation and attenuates left ventricular remodeling after myocardial infarction via activation of STAT3 and suppression of HuR. Circ. Res. 2009, 104 (2), e9−18. (24) Song, M.; Jang, H.; Lee, J.; Kim, J. H.; Kim, S. H.; Sun, K.; Park, Y. Regeneration of chronic myocardial infarction by injectable hydrogels containing stem cell homing factor SDF-1 and angiogenic peptide Ac-SDKP. Biomaterials 2014, 35 (8), 2436−45.

(25) Uchinaka, A.; Kawaguchi, N.; Mori, S.; Hamada, Y.; Miyagawa, S.; Saito, A.; Sawa, Y.; Matsuura, N. Tissue inhibitor of metalloproteinase-1 and −3 improves cardiac function in an ischemic cardiomyopathy model rat. Tissue Eng., Part A 2014, 20 (21−22), 3073−84. (26) Dobner, S.; Bezuidenhout, D.; Govender, P.; Zilla, P.; Davies, N. A synthetic non-degradable polyethylene glycol hydrogel retards adverse post-infarct left ventricular remodeling. J. Card. Failure 2009, 15 (7), 629−36. (27) Garbern, J. C.; Minami, E.; Stayton, P. S.; Murry, C. E. Delivery of basic fibroblast growth factor with a pH-responsive, injectable hydrogel to improve angiogenesis in infarcted myocardium. Biomaterials 2011, 32 (9), 2407−16. (28) Laham, R. J.; Sellke, F. W.; Edelman, E. R.; Pearlman, J. D.; Ware, J. A.; Brown, D. L.; Gold, J. P.; Simons, M. Local perivascular delivery of basic fibroblast growth factor in patients undergoing coronary bypass surgery: results of a phase I randomized, double-blind, placebo-controlled trial. Circulation 1999, 100 (18), 1865−71. (29) Kanki, S.; Segers, V. F.; Wu, W.; Kakkar, R.; Gannon, J.; Sys, S. U.; Sandrasagra, A.; Lee, R. T. Stromal cell-derived factor-1 retention and cardioprotection for ischemic myocardium. Circ.: Heart Failure 2011, 4 (4), 509−18. (30) Fedak, P. W.; Smookler, D. S.; Kassiri, Z.; Ohno, N.; Leco, K. J.; Verma, S.; Mickle, D. A.; Watson, K. L.; Hojilla, C. V.; Cruz, W.; Weisel, R. D.; Li, R. K.; Khokha, R. TIMP-3 deficiency leads to dilated cardiomyopathy. Circulation 2004, 110 (16), 2401−9. (31) Bromage, D. I.; Davidson, S. M.; Yellon, D. M. Stromal derived factor 1alpha: a chemokine that delivers a two-pronged defence of the myocardium. Pharmacol. Ther. 2014, 143 (3), 305−15. (32) Christia, P.; Frangogiannis, N. G. Targeting inflammatory pathways in myocardial infarction. Eur. J. Clin. Invest. 2013, 43 (9), 986−95. (33) Daskalopoulos, E. P.; Janssen, B. J.; Blankesteijn, W. M. Myofibroblasts in the infarct area: concepts and challenges. Microsc. Microanal. 2012, 18 (1), 35−49. (34) Graham, H. K.; Horn, M.; Trafford, A. W. Extracellular matrix profiles in the progression to heart failure. European Young Physiologists Symposium Keynote Lecture-Bratislava 2007. Acta Physiol. 2008, 194 (1), 3−21. (35) Jugdutt, B. I. Ventricular remodeling after infarction and the extracellular collagen matrix: when is enough enough? Circulation 2003, 108 (11), 1395−403. (36) Penn, M. S.; Pastore, J.; Miller, T.; Aras, R. SDF-1 in myocardial repair. Gene Ther. 2012, 19 (6), 583−7. (37) Lee, K. W.; Johnson, N. R.; Gao, J.; Wang, Y. Human progenitor cell recruitment via SDF-1alpha coacervate-laden PGS vascular grafts. Biomaterials 2013, 34 (38), 9877−85. (38) Yu, Y.; Zhang, Z. H.; Wei, S. G.; Chu, Y.; Weiss, R. M.; Heistad, D. D.; Felder, R. B. Central gene transfer of interleukin-10 reduces hypothalamic inflammation and evidence of heart failure in rats after myocardial infarction. Circ. Res. 2007, 101 (3), 304−12. (39) Zymek, P.; Nah, D. Y.; Bujak, M.; Ren, G.; Koerting, A.; Leucker, T.; Huebener, P.; Taffet, G.; Entman, M.; Frangogiannis, N. G. Interleukin-10 is not a critical regulator of infarct healing and left ventricular remodeling. Cardiovasc. Res. 2007, 74 (2), 313−22. (40) Bradham, W. S.; Moe, G.; Wendt, K. A.; Scott, A. A.; Konig, A.; Romanova, M.; Naik, G.; Spinale, F. G. TNF-alpha and myocardial matrix metalloproteinases in heart failure: relationship to LV remodeling. Am. J. Physiol Heart Circ Physiol 2002, 282 (4), H1288−95. (41) Wisniewska, M.; Goettig, P.; Maskos, K.; Belouski, E.; Winters, D.; Hecht, R.; Black, R.; Bode, W. Structural determinants of the ADAM inhibition by TIMP-3: crystal structure of the TACE-N-TIMP3 complex. J. Mol. Biol. 2008, 381 (5), 1307−19. (42) Bradham, W. S.; Bozkurt, B.; Gunasinghe, H.; Mann, D.; Spinale, F. G. Tumor necrosis factor-alpha and myocardial remodeling in progression of heart failure: a current perspective. Cardiovasc. Res. 2002, 53 (4), 822−30. 885

DOI: 10.1021/acsbiomaterials.6b00146 ACS Biomater. Sci. Eng. 2016, 2, 879−886

Article

ACS Biomaterials Science & Engineering (43) Sternlicht, M. D.; Werb, Z. How matrix metalloproteinases regulate cell behavior. Annu. Rev. Cell Dev. Biol. 2001, 17, 463−516. (44) Yu, W. H.; Yu, S.; Meng, Q.; Brew, K.; Woessner, J. F., Jr. TIMP-3 binds to sulfated glycosaminoglycans of the extracellular matrix. J. Biol. Chem. 2000, 275 (40), 31226−32. (45) Chu, H.; Gao, J.; Chen, C. W.; Huard, J.; Wang, Y. Injectable fibroblast growth factor-2 coacervate for persistent angiogenesis. Proc. Natl. Acad. Sci. U. S. A. 2011, 108 (33), 13444−9. (46) Karsan, A.; Yee, E.; Poirier, G. G.; Zhou, P.; Craig, R.; Harlan, J. M. Fibroblast growth factor-2 inhibits endothelial cell apoptosis by Bcl-2-dependent and independent mechanisms. Am. J. Pathol. 1997, 151 (6), 1775−84. (47) Tian, H.; Huang, M. L.; Liu, K. Y.; Jia, Z. B.; Sun, L.; Jiang, S. L.; Liu, W.; McDonald Kinkaid, H. Y.; Wu, J.; Li, R. K. Inhibiting matrix metalloproteinase by cell-based timp-3 gene transfer effectively treats acute and chronic ischemic cardiomyopathy. Cell Transplant 2012, 21 (5), 1039−53. (48) Hu, X.; Dai, S.; Wu, W. J.; Tan, W.; Zhu, X.; Mu, J.; Guo, Y.; Bolli, R.; Rokosh, G. Stromal cell derived factor-1 alpha confers protection against myocardial ischemia/reperfusion injury: role of the cardiac stromal cell derived factor-1 alpha CXCR4 axis. Circulation 2007, 116 (6), 654−63.

886

DOI: 10.1021/acsbiomaterials.6b00146 ACS Biomater. Sci. Eng. 2016, 2, 879−886