Some 3 years ago, one of our editors (Virgil Percec) suggested that I write a highlight about my current work. I accepted his invitation in principle … but we did not set a deadline. During the intervening months, I experienced his gentle prodding to hunker down and write science. I appreciated the nudging because otherwise I would have done more pleasant things, such as playing with my precocious grandchildren. Then, an invitation to give a lecture at a university gave me the needed activation energy. As I was mulling over ideas for a lecture to a sophisticated polymer audience, it occurred to me that my scientific metamorphosis, which led from living cationic polymerizations to thermoplastic elastomers (TPEs) and thence to biomaterials, would be a rather interesting tale to tell. Then it hit me: this was exactly what my editor wanted from me. Thus was born this highlight.
Brief Background with a Fast Forward to the Present
So where should I begin? I think I will start where I left off late in 1998, when I completed my earlier highlight, “Living Cationic Polymerization of Olefins: How Did the Discovery Come About?”, a they-said-it-cannot-be-done-therefore-we-did-it kind of a piece.1 The experimentation for that discovery was carried out in 1984, and the discovery was complete in 1985, but we delayed publication because of wrangling with the U.S. Patent Office (more on this in ref.2). Ultimately, this breakthrough led to the synthesis of a new TPE, poly(styrene-b-isobutylene-b-styrene), by my group, which included Gabor Kaszas, Judit Puskas, and Bill Hager, whose Ph.D. thesis was on this subject.3, 4
Toward the end of my highlight, in the section “A Glimpse into the Future”, among the many possibilities that I foresaw for living cationic polymerizations, I predicted a bright future for polyisobutylene (PIB)-based TPEs. In view of their unique combination of properties, such as good mechanical properties, softness, outstanding chemical resistance, and barrier properties, combined with excellent processing characteristics and recyclability, I expected these TPEs to be suitable for new hot-melt adhesives, sealants, blending agents, and so forth. I was, therefore, not surprised when I learnt that Kaneka, Inc., of Japan had very recently commercialized a series of brand-new TPEs (named SIBS for styrene–isobutylene–styrene) based on our triblock polymer. I was, however, astonished when I read that at the same international trade show, K2004 in Düsseldorf, Germany, where Kaneka had brought SIBS to market, that BASF, the German chemical giant, had also introduced an essentially identical PIB-based triblock under the trade name Oppanol IBS.
Let me also highlight a surprising important new application of these triblocks. Briefly, in March 2004, the U.S. Food and Drug Administration (FDA) approved the use of these triblocks as drug-eluting coatings of coronary stents, and Boston Scientific Co. has started marketing these devices in the United States under the trade name Taxus Express Paclitaxel-Eluting Stents. These drug-eluting stents were available in Europe a year earlier.
Stents are small, expandable, tubular metal scaffolds that, when inserted into dangerously occluded coronary arteries, pry them open and restore blood flow. It was observed, however, that after approximately 6 months of stent placement, life-threatening restenosis could occur in approximately 30% of patients. Therefore, stents coated with a polymer carrying a restenosis-preventing drug (e.g., paclitaxel) were a significant innovation. The PIB-based triblock is eminently suitable as a drug-eluting stent coating and satisfies the many requirements of such a demanding application, including bio- and hemocompatibility, controlled drug release, sterilization, confluent stent coating, and satisfactory mechanical properties of the coating that withstand the stresses during stent insertion and expansion without the integrity of the coating being compromised. Importantly, stents can be placed with relative ease, and this obviates the risks of major invasive bypass surgery. It is no wonder that stents, particularly drug-eluting stents, are revolutionizing coronary surgery, and because of their use, the number of bypass operations has plummeted some 85%. Currently, over 1 million stents are being implanted yearly worldwide.
Start of the Journey
With living cationic polymerizations firmly in our hands in 1985, we wondered: Which direction(s) among the many exciting possibilities toward heretofore unattainable polymeric architectures should we take? Where lay fame and fortune?
I saw only two choices: polymers for electronic applications or biomaterials, that is, areas in which function was critical and cost was relatively unimportant. I remember discussing these things with my group of students, postdocs, and visiting scientists during group meetings and individually in 1985 and 1986. In hindsight, it was an easy call for me because of my fascination with things biological; at heart, I remained a biochemist (Ph.D. in enzymology, University of Vienna, 1954). Thus, I decided to apply our new precision synthesis technique for the creation of polymeric biomaterials, that is, polymers that can be implanted into a living organism to assist or replace tissues or organs.
In the late 1980s, biomaterials were developed mostly by physicians (!) frustrated by the lack of suitable materials for clinical applications. These adventurous M.D. innovators developed the first polymeric biomaterials from polymers that they found on the shelf put there by industrial scientists who used them in more mundane industrial applications; thus arose surgical sutures from nylon yarns developed for ladies' hose, implantable silicones from electrical insulators, bone cements and ophthalmic materials (lenses, etc.) from Plexiglas [poly(methyl methacrylate) (PMMA)], indwelling polyurethane tubes and catheters from furniture upholstery, vascular grafts and surgical meshes from Dacron polyester no-iron fabric, orthopedic implants from polyethylene electrical insulating cables, and so forth.
I thought living cationic polymerizations could provide an excellent route to new molecular architectures expressly designed for biomaterials. However, I had to find the right need and match it with the right polymer. So I started to buttonhole biologists and all kinds of medical professionals willing to talk to with me, a polymer scientist searching for clinical needs that could be satisfied with polymers made with his newfangled technique of living cationic polymerization.
In the late 1980s, I thus became a biomaterialist.
Toughened Bone Cements
Among our first forays into designed biomaterials was the synthesis of toughened bone cements. Contemporary bone cements are essentially PMMA formulations designed to secure orthopedic prostheses, such as hip joints. The cements are prepared in the operating room by the mixing of the contents of two vials, one containing powdery PMMA, a radioopacifier (BaSO4), and aninitiator (benzoyl peroxide), and the other containing liquid methyl methacrylate (MMA) monomer, in which are dissolved an accelerator (N,N-dimethyl-p-toluidine) and a stabilizer (hydroquinone). Upon the mixing of these ingredients, the monomer dissolves the PMMA, and a paste is formed; at the same time, the initiator starts the complex polymerization/grafting of MMA. While this ill-defined reaction is in progress and the paste is still workable, the surgeon positions the paste around the prosthesis in the patient. The considerable heat of in situ MMA polymerization/grafting raises the temperature of the surrounding tissue to 75–80 °C, and the heat kills adjacent living tissue (no wonder the patient is under general anesthesia). Even more troubling, approximately 10% of the MMA remains unreacted and is released into the patient, and this may cause adverse reactions. (I doubt that the FDA would have allowed PMMA in orthopedic surgery as it is practiced today, had the FDA been in existence at the time bone cements came to be used, but this is another story.) After the polymerization is over, the paste solidifies into a heterogeneous, porous, brittle glass, which fixes the metal prosthesis to the bone. The residual MMA may also act as a softener of the solid PMMA that is formed. The polymer scientist immediately understands that in view of the nature of the three materials and two interfaces involved (bone/PMMA/metal), true chemical bonding cannot occur because PMMA is incompatible with both the bone and the metal. Rather, I think fixation is due to mechanical interlocking of the soft MMA/PMMA paste in the microscopic cavities of the bone and metal surfaces. (Essentially the same mechanism is at work when icicles form on the roof.)
Other major shortcomings of PMMA bone cements are brittleness and consequent debris formation. Revision surgery is needed in approximately 20% of the cases 10 years after implantation,5 and cement fracture is also frequently observed. Hip repair frequently fails because of prosthesis loosening at the bone/cement interface; because PMMA is a glass [glass-transition temperature (Tg) ∼ 105 °C], bone cements will sooner or later crack under a constant and dynamic load.
The toughening of brittle plastics has been thoroughly studied by polymer engineers, and remedies for this problem exist. One of the obvious things that a polymer scientist can do to combat the brittleness of PMMA is to covalently incorporate into the glassy matrix a dispersed elastomer phase. We hypothesized that the toughening of PMMA could be achieved by crosslinking with a triarm star methacrylate (MA)–telechelic PIB (Tg ∼ −70 °C) crosslinker. By this copolymerization, the PIB rubber becomes covalently bound to the continuous glassy PMMA phase, and because this nonpolar rubber is incompatible with the polar PMMA, the rubber will form a desirable dispersed phase in the continuous glassy matrix. Figure 1 illuminates the concept and shows the structure of the key ingredient, the rubbery Φ-(PIB–MA)3. This crosslinking agent can be obtained only by living cationic polymerization.6
We synthesized Φ-(PIB–MA)3's of various molecular weights and added them in various proportions to a commercial bone cement formulation. Extensive synthesis and characterization showed that cements containing 9.2% PIB with a number-average molecular weight (Mn) of 18,000 g/mol exhibited particularly desirable overall properties.7–10 Engineering tests indicated improved flexural strength, maximum deflection, fracture toughness, and fatigue crack propagation rate with respect to a commercial product (Zimmer Regular Bone Cement). An enterprising orthopedic surgeon tried this cement in dogs, but he died and the experiment was discontinued before data could be collected.
I do not know why this lead was not further developed by industrial researchers.
Cyanoacrylate (CNA)-Tipped PIB for Intervertebral Disc
The pain from a herniated (slipped) intervertebral disc can be excruciating. To remove the source of the trauma, the orthopedic surgeon excises the offending disc tissue (by laminectomy, chemonucleolysis, etc.); however, the tissue is not replaced, and the loads that were distributed by the excised disc must be taken over by other tissues.
We hypothesized that this rather unsatisfactory situation could be remedied by the replacement of the excised tissue with an elastomer whose viscoelastic characteristics are similar to those of the disc. It appeared feasible to inject a liquid prepolymer, which would rapidly polymerize to a rubber with the needed viscoelastic properties, into the cavity left behind by surgery. I thought we had the right replacement material: PIB carrying CNA groups. I knew we could prepare linear or star-shaped PIB prepolymers and fit them with reactive CNA groups. Figure 2 shows the structures envisioned, together with Super Glue (in which R is a small substituent, i.e., methyl, ethyl, or butyl). We speculated that prepolymers of suitable molecular weights and compositions could be injected where they were needed by a prefilled sterilized syringe containing the prepolymers. The principal constituent, PIB, a hydrophobic, inert, biocompatible, biostable, and oxidatively stable rubber, fitted with CNA end groups, would rapidly polymerize upon contact with living tissue (nucleic acids, proteins with NH2, OH groups). Because the polymerization would be induced by nucleophilic functional groups of the surrounding tissue, the PIB would be covalently anchored to the tissue, and the polymer would form only in the cavity where the liquid prepolymer was injected. Leakage of the polymer into the surrounding tissue could not occur. Furthermore, the long PIB chains would envelop and sequester the CH2C(CN)(COOPIB) main chain from the hydrolytic and enzymatic degradation that occur with small conventional CNAs.11
We demonstrated that the envisioned prepolymers could be readily synthesized and that they polymerized upon contact with living tissue (blood and egg yolk).11 The polymerization and crosslinking rates could be controlled by the use of various molecular weight products and by the copolymerization of linear and star-shaped prepolymers (see Fig. 2). I still believe that these PIB-based macromonomers would go a long way toward satisfying the requirements for disc replacement or other applications.
Smart Amphiphilic Hydrogels
This project started as a fundamental inquiry. It appeared to me that amphiphilic networks, that is, networks consisting of random hydrophilic and hydrophobic strands, would exhibit some unexpected properties, and an exploration of their synthesis and basic properties would be worthy of modest experimentation. Little did we imagine that this project would spawn drug delivery devices, artificial blood vessels, and immunoisolatory membranes. However, first we should cover some basics.
Amphiphilic networks (aptly named conetworks by B. Ivan) are random, bicontinuous assemblages of hydrophilic/hydrophobic chain segments that swell in both water and hydrocarbons. Because they swell in water, they are hydrogels. Figure 3 helps to visualize amphiphilic networks and emphasizes the conformational changes that these networks undergo upon changes in the medium (environmentally responsive networks). In tetrahydrofuran (an amphiphilic solvent), both chain elements are solvated and expanded, and the entire network swells. In water, the hydrophilic chains are extended, whereas the hydrophobic chains collapse to coils. In hydrocarbons, the opposite occurs: the hydrophilic chains are extended, whereas the hydrophilic chains collapse. In the latter two cases, the collapsed coils want to precipitate, but the covalently bound solvated chains will not let them.12 Surface analytical techniques by contact angle, atomic force microscopy, X-ray photoelectron spectroscopy, and surface atomic ratios (O/C and N/C) have indicated that the surfaces of amphiphilic networks are highly mobile; specifically, dry surfaces rapidly reorganize upon exposure to water.13 These networks are able to adopt different surface conformations in different environments to increase surface accommodation with the milieu and thereby minimize the total free energy of the system. Thus, amphiphilic networks are smart: they change their microstructure (morphology) with the medium. This chameleon-like change may be important for accommodating complex biological systems consisting of all kinds of molecules and constituents (i.e., for biocompatibility).
PIB-Based Amphiphilic Networks: Controlled Release, Artificial Arteries, and the First Generation of Immunoisolatory Membranes
Our first generation of amphiphilic networks was prepared by the free-radical solution copolymerization of hydrophilic monomers [N,N-dimethyl acrylamide (DMAAm), 2-hydroxyethyl methacrylate (HEMA), 2-(dimethylamino)-ethyl methacrylate (DMAEMA), and sulfoethyl methacrylate (SEMA)] with hydrophobic crosslinkers (MA–telechelic PIBs).6, 14–19 Figure 4 shows the formulas. The MA-capped PIB crosslinkers can be prepared only by living isobutylene polymerization followed by end-group functionalization.6 Figure 5 outlines the synthesis of one of the simplest amphiphilic network by the copolymerization of DMAAm with MA–PIB–MA.
By our technique, we can control the overall composition of the networks and the molecular weights between crosslinking points (Mc's). Because these networks have hydrophilic and hydrophobic segments, they exhibit two Mc's: Mc,HI (the molecular weight of the hydrophilic segment between crosslinking points) and Mc,HO (the molecular weight of the hydrophobic segment). Mc,HI controls the pore (or mesh) dimensions of the networks in water (i.e., when implanted in an organism). The pore dimensions of the networks can be regulated by the concentration of the hydrophilic monomer with respect to the hydrophobic crosslinking agent.
Because the chain elements in amphiphilic networks are chemically very different and thermodynamically incompatible, they are two-phase systems with two Tg's.20, 21 Transmission electron microscopy of a typical amphiphilic network has shown 20–50-Å-diameter bicontinuous domains with a salt-and-pepper morphology.
We studied the kinetics of swelling of amphiphilic networks, using both water and n-heptane.15–18, 21 We found that, with an increase in the PIB content, the rate of water uptake decreased, whereas that of n-heptane increased; the opposite was found with an increase in the content of the hydrophilic component (i.e., the water uptake increased). Subsequently, we studied the out-diffusion of select water-soluble model drugs (e.g., folic acid) from drug-loaded amphiphilic networks,14 and we found that the release rate changes with the nature and concentration of the hydrophilic constituent and the molecular weight of the PIB crosslinker. Interestingly, the diffusion coefficient n (obtained from Mt/M∞ = ktn, where Mt is the amount of drug released at time t, M∞ is the amount of drug loaded, and k is a constant), determined for several networks, fell in the 0.7–0.8 range. If n = 0, diffusion is controlled by polymer relaxation (zero-order release), whereas n = 0.5 indicates conventional or Fickian diffusion.22 The experimental values for several of our systems suggest anomalous transport, that is, the presence of another process besides passive diffusion.
A series of amphiphilic networks containing poly(2-sulfoethyl methacrylate) and MA–PIB–MA strands were prepared and characterized by thermal, spectroscopic, mechanical, and swelling experiments18 Their swelling followed non-Fickian kinetics in both water and n-heptane. Networks with higher ionic contents showed rapid and reversible swelling or deswelling upon changes in the pH (in the 2–12 range) of the medium.
The discovery that networks containing approximately 50/50 DMAEMA/PIB (Mn,PIB = 10,000 g/mol) exhibited excellent biocompatibility and biostability in rats14 was of decisive importance for the future course of our research. We found that certain well-defined amphiphilic networks integrated well with tissue and sowed minimal bacterial contamination, no edema, and virtually no fibrosis and adhesion (even less than the polyethylene negative control).14 Indeed some of our materials exhibited better biocompatibility than the negative controls. In cell culture and protein tests, the numbers of cells and the total protein on the amphiphilic networks were similar to those of negative controls (polyethylene, silicone rubber, and glass), and this indicated no toxic response. Cell adhesion and antiadhesion experiments with human monocytes showed monocyte adhesion inhibition for various amphiphilic networks and glass (negative control) with respect to polystyrene (positive control). Further quantitative protein adsorption by radioimmunoassay showed that amphiphilic networks made with DMAAm or HEMA with 50% PIB adsorbed from human plasma less fibrinogen, Hageman factor, and albumin than glass, silicone rubber, or polyethylene.23 These analyses, together with blood cell counts, suggest that select amphiphilic networks are well accepted in vivo; that is, they are biocompatible.
Overall, these observations indicate reduced protein adsorption, with a significant reduction of Hageman factor and fibrinogen adsorption. Together with human monocyte adsorption data, these studies indicate that select amphiphilic networks are biologically compatible at blood-contacting surfaces.
Coronary artery obstruction (stenosis) is life-threatening. A clinical solution to coronary stenosis is major surgery in which the occluded artery is replaced by another of the patient's native blood vessels (typically, the saphenous vein or mammary artery). Unfortunately, 3–5-mm-diameter blood vessels, such as the coronary artery, cannot be replaced by synthetic polymer tubes [e.g., polyurethane, polyethylene, polyester, or poly(tetrafluoroethylene)] because after graft placement they quickly (within a few hours) become occluded by platelet and fibrin deposition.24
In view of the bio- and hemocompatibility of select amphiphilic networks, we set out to explore whether a narrow-caliber tube made of our materials could be used for coronary artery replacement; specifically, we wondered how many minutes per hour a 3-mm-caliber amphiphilic tube would remain free of stenosis with blood circulating through it.
We built an apparatus by which we circulated fresh rat blood (no anticoagulant) at 37 °C through 4–5-cm-long amphiphilic tube sections made by the copolymerization of DMAEMA/MA–PIB–MA in a rotating glass cylinder [the rotational copolymerization technique has been described elsewhere].25 A peristaltic pump was used to simulate the pumping action of the heart; that is, the tube pulsated during blood circulation. Experiments were carried out for 60 min, during which time no trace of platelet deposition in the amphiphilic tube was observed! Under similar conditions, the negative control, a plasticized poly(vinyl chloride) (Tygon) tube, showed significant platelet deposition.26 The results obtained by this dynamic hemocompatibility test corroborated those obtained by protein and cell adsorption experiments (as previously discussed).
Encouraged by these findings, we carried out a more demanding dynamic hemocompatibility test jointly with researchers at the University of Wisconsin.27 By the use of an artificial heart device, these researchers circulated 60 mL of heparinized (3 units) fresh bovine blood containing 111In labeled platelets through various 3-mm-caliber tubes for 120 min. We compared three of our amphiphilic network tubes made by the copolymerization of DMAAm, DMAEMA, or HEMA plus the MA–PIB–MA crosslinker (as previously discussed), with a polyurethane tube as a negative control. Thrombosis was quantitated by γ counting of the platelets. Although the polyurethane control gave more than 80,000 counts per minute (cpm), which indicated relatively high platelet deposition, our amphiphilic tubes showed far less platelet deposition, approximately 1000 cpm, with the tube made with DMAAm being the best (∼200 cpm, average of three experiments).
Jointly with Dutch biomaterials scientists, we studied platelet adhesion and blood coagulation activation of select amphiphilic networks in reference to polyethylene and poly(vinyl chloride) negative controls in vitro, that is, polymers commonly used in blood-contacting applications.28 Networks of DMAAm/PIB exhibited lower thrombogenicity than polyethylene and poly(vinyl chloride) and significantly lower platelet adhesion than poly(vinyl chloride), the reference with lowest thrombogenicity.
Although the results of these bio- and hemocompatibility experiments were encouraging, the mechanical properties of our tubes were insufficient for vascular implantation; microsuturing was deemed impossible because of the poor puncture and tear strengths of our water-swollen tubes. To overcome this problem, we hypothesized that thromboresistant narrow-caliber tubes could be obtained by the coating of 3-mm-diameter expanded poly(tetrafluoroethylene) (ePTFE) tubes with amphiphilic networks. Although large-caliber (>6 mm) ePTFE tubes (Goretex and Impra) are extensively used in vascular surgery, narrow-caliber tubes cannot be used because they are thrombogenic. Thus, in preparation for in vivo implantation experiments, we coated the surfaces of 3-mm-diameter Goretex tubes with one of our promising amphiphilic formulations (50/50 DMAEMA/MA–PIB–MA, with Mn,PIB = 4500 g/mol). Scanning electron microscopy (SEM) showed smooth, featureless surfaces, whereas the surfaces of uncoated Goretex tubes displayed the characteristic striated, fibrillar morphology of Goretex. Interestingly, water wetted the coated tubes even after their coating was manually peeled off. Evidently, the amphiphilic coating penetrated and remained in the interstices of the porous ePTFE and rendered it amphiphilic.
We grafted approximately 2-cm sections of 3-mm-caliber uncoated (control) and amphiphilic-polymer-coated Goretex tubes into the aortas of two rabbits.29 Coating was carried out by the immersion of Goretex tubes into polymerizing amphiphilic charges. Heparin was administered to prevent blood clotting. After 7 days, the animals were sacrificed, and the junctions of the graft and native aortas were examined by SEM. The luminal surface of the amphiphilic network-coated graft showed a featureless confluent coating with no platelet or fibrin deposition or cellular debris; in contrast, the surfaces of the uncoated control were extensively covered with thrombus. Figure 6 shows representative SEM pictures.
Significantly, we noted minor areas of discontinuities in the homogeneous coats, which exhibited the characteristic fibrillar morphology of Goretex (see Fig. 6, right). These minor discontinuities were probably due to imperfect coat deposition (entrapped air bubbles?). The absence of cellular debris over these areas suggests that even an ultrathin, SEM-invisible amphiphilic-polymer layer deposited on the surface of Goretex is sufficient to prevent thrombus formation. Recent follow-up experiments carried out jointly with Dr. K. Ouriel et al. of the Cleveland Clinic Foundation have been focused on improving the coating of ePTFE (Impra) by the use of a rapidly rotating coating device (to be published).
In conclusion, these preliminary studies show that 3.0-mm-diameter amphiphilic-polymer-coated ePTFE can be used as an aortic conduit in the rabbit model, that the amphiphilic-polymer coat does not change the handling characteristics of the ePTFE tube, and that amphiphilic-network-coated ePTFE exhibit significantly less thrombus deposition than uncoated tubes. Efforts are in progress to follow up this lead.
First Generation of Immunoisolatory Membranes
Numerous discussions with clinicians, biomaterial researchers, and immunologists have convinced me that there is a need for novel immunoisolatory membranes (the language barrier between our disciplines sometimes hampered our discussions). Immunoisolatory membranes are used to encapsulate and transplant living tissue from a donor to a host organism (xenotransplantation). Such membranes protect the transplants from being destroyed by the host's immune system, thereby eliminating costly and dangerous immunosuppressive drug therapy.30–32 Immunoisolation of, for instance, living pig pancreatic islets (beta cells) into diabetic humans may correct diabetes.
Islet encapsulation/transplantation has been investigated by many researchers employing many kinds of materials and methods. A thorough analysis of the requirements gleaned from the scientific and patent literature has led us to conclude that an ideal immunoisolatory membrane, clinically useful for a bioartificial pancreas, must simultaneously satisfy all the following biological, chemical, physical, mechanical, surface, and processing properties:
1.Biocompatibility with the host (human) and guest (e.g., porcine islets).
4.Smooth, slippery, nonclogging, nonfouling, avascular, and nonthrombogenic or, in other words, immunologically invisible surfaces.
5.Controlled semipermeability: precisely designed/defined pore dimensions (molecular weight cutoff ranges) that allow the passage of aqueous solutions of nutrients and biologically active molecules (insulin) and the exit of metabolic wastes but exclude antibodies and white blood cells.
6.Physiologically satisfactory bidirectional fluxes of glucose, insulin, nutrients, and metabolites.
7.Thin membrane walls (micrometer range) to minimize diffusion paths.
8.Satisfactory mechanical properties (strength, modulus, elongation, and fatigue) for the implantation and explantation of large numbers (∼8 × 105) of islets.
9.Highest and rapid oxygen and water transport.
10.All the above properties to be maintained for long times (6–12 months).
11.Simple and efficient membrane synthesis.
12.Easily manufactured into sealable and preferentially transparent tubes of pouches of well-defined volumes.
13.Ease of implantation and explantation.
15.All this for a reasonable cost.
Implantable membranes examined by others for the correction of diabetes have several but not all of these characteristics; for example, many researchers use alginates or siloxane gels to microencapsulate individual islets. One of the fundamental disadvantages of microencapsulation by hydrogels is that the immunoprotected tissue cannot be reliably or completely retrieved. The other is that hydrogels have very poor oxygen permeability (water is a barrier Io oxygen diffision).
With this analysis in mind, we set out to synthesize immunoisolatory membranes from our amphiphilic polymers with the required biological and mechanical properties. We thought to achieve semipermeability control (see items 5–8) by regulating the length of Mc,HI and by the overall hydrophilic/hydrophobic composition of the membranes. The molecular weight cutoff range (pore size control) was to be achieved by the regulation of the length of the hydrophilic and hydrophobic segments to allow the rapid countercurrent diffusion of glucose and insulin, but the membranes were to be impermeable to large proteins of the immune system [e.g., immunoglobulin G (IgG)].
Systematic experimentation showed that amphiphilic membranes containing approximately 50/50 poly(N,N-dimethyl acrylamide)/PIB with Mc,HI ∼ 4500 g/mol had semipermeability and diffusion rates suitable for the immunoisolation of pancreatic islets.33 These membranes allowed the countercurrent diffusion of glucose and insulin (Mn = 180 and 5700 g/mol, respectively) but prevented the diffusion of albumin (Mn ∼ 66,000 g/mol), and the diffusion rates (fluxes) of glucose and insulin were deemed appropriate for islet immunoisolation.33 Subsequently, we determined that pig islets placed in such semipermeable amphiphilic-polymer tubules could be kept viable in tissue culture for at least 4 months and that encapsulated islets produced insulin upon glucose challenge.34 Importantly, we also demonstrated that a diabetic rat, fitted subcutaneously with our bioartificial pancreas (i.e., an amphiphilic tubule containing pig islets), started to produce insulin and that the glucose concentration in the blood of the rat decreased markedly. When the bioartificial pancreas was removed, the rat became diabetic again. In these experiments, the rat was its own control! Figure 7 summarizes the results of this experiment.35
Polydimethylsiloxane (PDMS)-Based Amphiphilic Networks: The Second Generation of Immunoisolatory Membranes
There were three compelling reasons that we decided to advance from PIB-based immunoisolatory membranes to PDMS-based immunoisolatory membranes: (1) to simplify membrane synthesis, (2) to enhance oxygen permeability, and (3) to obtain uniform pore dimensions.
Let me explain:
1.Our first generation of amphiphilic membranes obtained by the free-radical copolymerization of acrylates yielded very promising products; however, the synthesis of the PIB-based crosslinking agents needs considerable expertise, and the radical copolymerization mechanism is inherently ill defined. I was therefore constantly searching to find a less demanding synthetic procedure, and I was pleased when we found that excellent amphiphilic membranes could also be made by simple hydrosilation/condensation (discussed later).
2.A thorough search of the scientific and patent literature, including web pages of companies engaged in immunoisolation, indicated that one of the crucial requirements of immunoisolatory membranes was oxygen permeability. Although native pancreatic islets receive oxygen via blood circulation, immunoisolated islets receive oxygen (and eliminate waste) only via passive diffusion through the encapsulating membrane. Obviously, then, to ensure adequate oxygen supply to the islets, the membranes must be as thin as feasible and as friendly to oxygen as possible. We reasoned that the oxygen permeability of PIB-based amphiphilic membranes, that is, membranes through which oxygen transport can occur only via water in the hydrophilic domains (PIB is a barrier to oxygen diffusion), could be vastly enhanced by the substitution of PDMS for PIB. Moreover, oxygen diffusion only via the hydrophilic domains was thought to be inadequate because of the low solubility of oxygen in water (200–300 mg/mL). Thus, we turned to PDMS, the most oxygen-permeable rubber, whose oxygen permeability is far superior to that of water. It is true that PDMS is somewhat weaker than PIB, but we had ideas how to increase the strength of PDMS-based membranes, should this become an issue.
3.Uniform pore dimensions cannot be obtained with networks made by a random free-radical copolymerization of hydrophilic monomers with MA-functionalized PIBs, and Mc,HI will always exhibit a rather broad length distribution [weight-average molecular weight/number-average molecular weight (Mw/Mn) ∼ 2.0]. Such a broad distribution, however, is dangerous because in the presence of even a minute fraction of larger pores, some immunoproteins may traverse the membrane and thus compromise the encapsulated living tissue. Thus, to obtain uniform pore dimensions, we had to redesign our membranes. So how should we proceed?
To obtain the highest oxygen permeability and pore uniformity, we decided to use precisely defined hydrophilic poly(ethylene glycol) (PEG) and hydrophobic PDMS starting materials, both with the narrowest possible molecular weight distributions (Mw/Mn ∼ 1.0). Specifically, we thought that random combinations of precise-length (molecular weight) PEG and PDMS chains would yield amphiphilic networks with uniform, precisely defined, and controllable pore dimensions. However, how can we combine the incompatible PEG and PDMS segments into a network?
At this point, fate came to our aid: Fortuitously, about this time, one of my students, P. Kurian, investigated reactions of cyclosiloxanes for another project and found that pentamethylcyclopentasiloxane (D5H) rapidly polycondenses to poly(pentamethylcyclopentasiloxane) (PD5).36 It occurred to me that PD5 was exactly what we needed because (1) PD5 domains could function as efficient crosslinkers by the cohydrosilation of olefin–ditelechelic PEG and –PDMS segments, (2) PD5 would function as reinforcing domains, and (3) as a bonus the oxyphilic PD5 domains would provide auxiliary oxygen channels (in addition to those provided by PDMS). In other words, the PD5 domains/phases in our networks would perform triple duty by providing crosslinking, reinforcement, and supplemental oxygen channels.36
Figure 8 helps to visualize the synthetic strategy and micromorphology of an amphiphilic network consisting of PEG and PDMS segments crosslinked and reinforced by PD5 domains. The synthesis involves the cohydrosilation of A–PEG–A and V–PDMS–V mixtures by D5H, followed by simultaneous water-mediated oxidation of the excess SiH groups to SiOH groups and in situ polycondensation to PD5 domains.37 A–PEG–A is easily obtained from commercially available HO–PEG–OH,37 and the other two starting materials are inexpensive commercial products. The relative concentrations of the three constituents (PEG/PD5/PDMS) control the overall membrane composition, which in turn controls the overall membrane properties.
The crosslink density and stiffness of the networks increase with the D5H concentration, and the concentration of water controls the rate and extent of crosslinking. It did not take us long to develop suitable cohydrosilation, oxidation, and polycondensation conditions and to obtain membranes with excellent mechanical properties (>5 MPa strength and 500% elongation) appropriate for immunoisolation.38, 39
Having established synthesis simplicity and versatility, we prepared and characterized series of amphiphilic membranes with various Mc's, PEGs (in the 4600–20,000 g/mol range), and various PEG/PD5/PDMS compositions.39 Surface analyses indicated the expected rapid conformational rearrangements. The oxygen permeability increased with the amount of PDMS in the membrane, and the insulin permeability was shown to be dependent on the length of Mc,PEG. The permeability of insulin through a series of membranes with various Mc,PEG's was determined. Molecular weight cutoff studies showed controllable semipermeability, that is, the diffusion of small proteins including insulin and the rejection of larger proteins such as albumin (Mn ∼ 66,000 g/mol). The implantation of representative PEG/PD5/PDMS membranes in rats showed minimal response with respect to inflammation, foreign body reaction, tissue ingrowth, and fibrous capsule formation.39
To further enhance reinforcement, we prepared amphiphilic networks by cocrosslinking PEG/PDMS mixtures with polyhedral oligomeric silsesquioxane (POSS) fitted with eight SiH groups (in lieu of D5H).40
Recently, we demonstrated, by a series of diffusion experiments, the selective permeability of glucose and insulin with the simultaneous exclusion of IgG.39 Subsequently, we also demonstrated that PEG/PD5/PDMS membranes do not foul and remain permeable to both glucose and insulin even after long (1-month) incubation with IgG. Figure 9 provides details and shows our findings.
In sum, we developed a simple and efficient membrane synthesis method; the bicontinuous membrane architecture and the oxyphilic PDMS/PD5 domains provide superior oxygen transport; the precisely defined PEG segments (in terms of the molecular weights and molecular weight distributions) yield uniform pore dimensions (i.e., well-controlled Mc,HI); and the membranes exhibit desirable semipermeability and are biocompatible.
We are on the road to a clinically useful bioartificial pancreas.
Although support has been received from many sources during these investigations, the author is mainly indebted to the National Science Foundation for its continuous financial help (DMR-8920826 and DMR-0243314).
JOSEPH P. KENNEDY
Joseph P. Kennedy started his university career in his native city, Budapest. Just before graduating from the university, he was removed by the communist administration because of his bourgeois origin. He escaped to Vienna, where he received his Ph.D. in biochemistry in 1954, and subsequently he did postdoctoral work in Paris and Montreal (1954–1957). He came to the United States in 1957 and became an industrial polymer researcher, first with Celanese and then with Exxon. In 1961, he received an M.B.A. at Rutgers. He resumed his academic career at the University of Akron in 1970, where he is still carrying out research as a Distinguished Professor of Polymer Science and Chemistry. Kennedy's main interest lies in ionic (particularly cationic) polymerizations and, for the last 15 years, in designed biomaterials. He has written three books and almost 700 publications and has over 90 issued U.S. patents, some of them in commercial production. He has received many awards, including the two premier international polymer awards of the American Chemical Society (Polymer Chemistry and Applied Polymer Science). For obvious reasons, he derives his greatest satisfaction from the honorary doctorate awarded by the best science university in Hungary (1989) and by his election as a member of the Hungarian Academy of Sciences (1993).