Structure/Function Relationships of Polyamidoamine/DNA Dendrimers As Gene Delivery Vehicles

Structure/Function Relationships of Polyamidoamine/DNA Dendrimers as Gene Delivery Vehicles

CHAD S. BRAUN,1 JOSEPH A. VETRO,1 DONALD A. TOMALIA,2 GARY S. KOE,3 JANET G. KOE,3 C. RUSSELL MIDDAUGH1 1Department of Pharmaceutical Chemistry, University of Kansas, 2095 Constant Ave., Lawrence, Kansas 66047 2Dendritic Nanotechnologies, Mt. Pleasant, Michigan 48858 3Valentis Inc., Burlingame, California 94010

Received 23 July 2004; revised 16 September 2004; accepted 1 October 2004 Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.20251

ABSTRACT: PAMAM dendrimers are members of a class of polyamine polymers that demonstrate significant gene delivery ability. In this study, a selection of PAMAM dendrimers, spanning a range of sizes (generations 2, 4, 7, and 9) and transfection efficiencies, are characterized by various biophysical methods to search for structural properties that correlate with transfection. Measurements of colloidal properties (size and zeta potential) as a function of charge ratio reveal that highly transfecting dendrimer/DNA complexes have size/zeta potential values between 4 and 8. Circular dichroism (CD) and FTIR spectroscopy of complexes confirm the DNA component remains in B form when associated with all dendrimer generations up to a 5:1 charge ratio (). Isothermal titration calorimetry and differential scanning calorimetry detect changes that are related to polymer structure and charge ratio but do not directly correlate with transfection efficiency. Despite DNA structural and stability changes detected by CD, FTIR, DSC, and ITC that are similar to those seen with other cationic delivery vehicles [e.g., cationic lipids, peptoids/lipitoids, peptides, polyethyleneimines (PEIs), etc.], clear correlations with transfection activity are not readily apparent. This may be due, at least in part, to the heterogeneity of the complexes. ß 2004 Wiley-Liss, Inc. and the American Pharmacists Association J Pharm Sci 94:423–436, 2005 Keywords: dendrimers; DNA; structure/function analysis; gene delivery

INTRODUCTION ferent species formed with transfection ability. In a further extension of this approach, the nature of The mechanism by which nonviral gene delivery the polymer itself can be varied and structure/ vehicles transfect cells remains poorly under- function correlations probed. To this end, we stood.1 This is especially problematic because have systematically studied complexes formed it makes the rational design of more effective between plasmid DNA with various cationic vectors very difficult. One approach to this pro- lipids,2–9 different molecular weight polyethyle- blem is to systematically vary the structure of neimines (PEIs),10 and peptoids with altered side such complexes by altering the ratio of delivery chains11,12 as a function of charge ratios to seek polymer to DNA and attempt to correlate the dif- significant correlations with in vitro transfection efficiency. The result of these efforts has so far been somewhat disappointing. This may reflect the wide differences among the various delivery Correspondence to: C. Russell Middaugh (Telephone: 785- 864-5813; Fax: 785-864-5814; E-mail: [email protected]) vehicles as well as the intrinsic heterogeneity of Journal of Pharmaceutical Sciences, Vol. 94, 423–436 (2005) the complexes so far studied. To obviate some of ß 2004 Wiley-Liss, Inc. and the American Pharmacists Association these difficulties, we have now extended this

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 423 424 BRAUN ET AL. approach to the study of polyamidoamine The representative PAMAM dendrimers inves- (PAMAM) dendrimer complexes. tigated here (G2, 4, 7, and 9) have well-defined The PAMAM dendrimers are polymers contain- chemistry and molecular weights in contrast to the ing both tertiary amines at branch points as well as cationic component of nonviral complexes formed primary amines at the termini (Fig. 1). PAMAM from cationic lipids, peptoids/lipitoids, and PEI, dendrimers (Fig. 1) are formed by exhaustive providing a system in which transfection efficiency Michael addition of ethylene diamine with methyl varies, but with less complication from hetero- acrylate followed by addition of the resulting ester geneity and variable chemistry.10,12 Therefore, core to an excess of ethylene diamine. This syn- we have conducted an extensive biophysical char- thesis gives a high yield of first-generation acterization of PAMAM dendrimer/DNA com- PAMAM dendrimers, with higher generations plexes (PDDCs) with the goal of establishing arising from repetition of the previous two steps. a correlation between their physical properties Generations up to 10 can be formed before fur- and in vitro transfection efficiency. To this end, we ther assembly is limited by the ‘‘de Gennes have investigated the hydrodynamic size and zeta dense packing’’ phenomenon.13 The dendrimers potential of complexes by dynamic light scat- so prepared have a consistent size, structure, and tering and electrophoretic light scattering. The charge characteristic of their generation. Trans- thermodynamics of the complexation process fection efficiency mediated by PAMAM dendri- and the stability of complexes were examined by mers appears to be dependent on dendrimer isothermal titration calorimetry (ITC) and differ- generation, with larger sizes providing higher ential scanning calorimetry (DSC), respectively. efficiency, and the charge ratio of the complexes, The relative affinities of the DNA/dendrimer in- in which a net positive charge is optimal.14 In teractions were determined by ethidium bromide addition to the simple complexation of DNA with dye displacement. The secondary structure of intact dendrimers, further transfection enhance- the DNA component when complexed to dendr- ment by the use of fractured (hydrolysis degraded) imer was investigated using circular dichroism dendrimers15 or addition of cationic excipients (CD) and Fourier transform infrared (FTIR) such as DEAE–dextran have also been reported.14 spectroscopy.

Figure 1. Chemical structure of a generation 1 PAMAM dendrimer.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 STRUCTURE/ACTIVITY RELATIONSHIP FOR PAMAM DENDRIMER/DNA COMPLEXES 425

These same methods were utilized in the In Vitro Transfection Efficiency characterization of a number of other nonviral The transfection efficiency of dendrimer/DNA gene delivery preparations studied previously in complexes was determined by luciferase expres- our laboratory, as indicated above. The current sion in CHO-K1 cells. The CHO-K1 cells were analysis of the dendrimer/DNA delivery system obtained from the American Type Culture Collec- was compared to these previous studies. tion (Rockville, MD). Cultures were maintained in 2 75 cm flasks at 378C and 5% CO2 and grown in MATERIALS AND METHODS Ham’s F-12 media containing L-glutamine and 10% FBS. Cells were split every 4 days using PAMAM dendrimers of generations 2, 4, 7, and 9 standard procedures, and accutase was employed were a generous gift from Dendritic Nanotechnol- for cell lifting. Plates were prepared by adding ogies (Mt. Pleasant, MI). Plasmid DNA pMB 290 8000 cells (e.g., 16% confluence) to each well of (4.9 kbp) and pMB 401 (5.9 kbp, encoding firefly a 96-well plate and incubating for 18–20 h at 378C luciferase), both >95% supercoiled, were provided and 5% CO2 before transfection. by Valentis, Inc. (Burlingame, CA). The former Approximately 1 mg of PAMAM dendrimer of (pMB 290) was used for all the characterization each generation was dried under vacuum with studies while the latter (pMB 401) was used for silica gel desiccant to remove methyl alcohol until the in vitro transfection studies. The luciferase a stable weight was achieved. The pMB 401 and marker gene was under the control of the CMV 10 mM Tris buffer (pH 7.4) used for reconstitution promoter. Ethidium bromide was purchased from of the dendrimers was filtered through a 0.22-mm Molecular Probes (Eugene, OR). Dulbecco’s mod- filter. Appropriate volumes of polymer and DNA ified Eagle media was acquired from Cambrex were mixed to prepare complexes at a 25 mgmL1 (Rutherford, NJ). Fetal bovine serum was pur- DNA concentration. Complexes at various charge chased from Atlanta Biologicals (Nocross, GA). ratios were diluted to a final transfection DNA Accutase was obtained from Innovative Cell concentration of 2 mgmL1 with Opti-MEM. Technologies (San Diego, CA). Opti-Mem was Immediately before transfection, the cells were acquired from Invitrogen (Carlsband, CA). 1,2- washed once with PBS followed by addition of Dioleoyl-3-trimethylammonium propane (DOTAP) 100 mL of solution containing complexes (200 ng of was obtained from Avanti Polar Lipids (Alabaster, pMB 401) to each well. Cells were incubated with AL). Tris and phosphate buffer salts were pur- complexes for 5 h. The 100 mL of transfection solu- chased from Sigma (St. Louis, MO). Nano-purified tion was then removed and replaced by 100 mLof water was used for all buffer preparation. culture media. The cells were incubated for a further 48 h at 378C and 5% CO2. Preparation of Complexes Luciferase expression was assayed using the Luciferase Assay SystemTM from Promega The dendritic polymers were received at high (Madison, WI) following the supplied protocol. The concentration (>10% w/w) in methyl alcohol, and cells were lysed with Reporter Lysis Buffer for were diluted directly into buffer for analysis 30 min. Expression was analyzed in 96-well plates unless otherwise stated. The plasmids were with a FluostarTM Galaxy microtiter plate reader diluted from a stock solution. The concentration (BMG, Offenburg, Germany). Automated addition of DNA solutions was determined using their UV of substrate buffer was employed and the lumines- absorbance at 260 nm (A ) and a molar extinc- 260 cence was then measured. The amount of protein tion coefficient of 0.02 (mg1 cm1 mL). Individual in each well was determined by Coomassie Blue dilutions were performed for DNA and dendrimer assay (Promega). The data were normalized to the at each charge ratio, calculated as the mol of total protein content and reported as relative light dendrimer primary amines per DNA phospha- units per milligram of protein (n ¼ 3). tes. PAMAM dendrimer/DNA complexes (PDDCs) were formed by addition of equal volumes of DNA and dendrimer while stirring. The order of Dynamic and Electrophoretic Light Scattering addition was selected to avoid passing through charge neutrality. These complexes were allow- All samples intended for light scattering analyses ed to equilibrate for 20 min prior to analysis. were prepared using 10 mM Tris buffer, pH 7.4, Complexes were prepared in 10 mM Tris buffer, which was filtered with a 0.22-mm filter to remove pH 7.4 unless otherwise noted. any trace particulates. Complexes were prepared

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 426 BRAUN ET AL. at a 100 mgmL1 DNA concentration and function (7 pts) and a fast Fourier transform at various charge ratios between 0.25 and 5. method. Peak position determination (0.3 nm) Particle sizes were measured by dynamic light and plotting were performed using Origin 6.0 scattering (DLS) using a Brookhaven (Holtsville, software (Microcal Software, North Hamptom, NY) instrument equipped with a 9000AT autocor- MA). relator, a 50-mW HeNe laser operating at 532 nm (JDS Uniphase), an EMI 9863 photomultiplier Fourier-Transform Infrared (FTIR) Spectroscopy tube, and a BI-200M goniometer. The light scat- tered at 908 from the incident light was fit to FTIR spectra were obtained with a Nicolet Magna- an autocorrelation function using the method of IR 560 spectrometer (Madison, WI) configured cumulants. The resultant diffusion coefficient was with a mercury–cadnium–telluride (MCT) de- then converted to a mean hydrodynamic diameter tector. Samples were examined at a DNA concen- by the Stokes-Einstein equation. tration of 1 mg mL1 in a zinc selenide (ZnSe) Zeta potential measurements were obtained by ATR trough cell (Thermo Spectra-Tech, Madison, phase analysis light scattering (PALS) using a WI) with an effective pathlength of 12 mm. Spectra Brookhaven Zeta PALS instrument. The electro- were obtained as 256 scan interferograms under phoretic mobility of the DLS samples was deter- dry air purge using Happ-Genzel apodization mined from the average of 10 cycles of an applied with no zero filling producing an effective resolu- electric field. The zeta potential of complexes was tion of 2 cm1 as described previously.2 Data determined from the electrophoretic mobility by analyses including buffer subtraction, baseline means of the Smoluchowski approximation. adjustment, maximum entropy smoothing (Razor Tools, Spectrum Square Associates, Ithaca, NY), and peak position determination were conducted Ethidium Bromide Displacement Assay in Grams AI (Thermo Galactic, Salem, NH). Peak Ethidium bromide was bound to DNA at a 4:1 positions were determined from the mean of five ratio (DNA bp to ethidium molecule).4 The labeled replicate independent experiments with the stan- DNA was complexed to dendrimers at various dard error plotted as error bars. charge ratios from 0.25 to 5, while the DNA concentration was maintained at 10 mgmL1. All Isothermal Titration Calorimetry (ITC) samples were prepared in triplicate and the relative fluorescence was quantitated by subtract- Calorimetric titrations were conducted at 258C ing the fluorescence intensity of ethidium bromide using a Calorimetry Sciences (CSC) model 4200 in water and normalizing to DNA alone. Samples instrument. The DNA and dried dendrimer (pre- were placed in a 96-well plate and fluorescence pared as described above) were buffer matched by was measured using a Galaxy Fluorstar plate dialysis and drying with subsequent reconstitu- reader at an excitation wavelength of 530 nm and tion in the same buffer. Both solutions were then emission wavelength of 590 nm. The fluorescence degassed prior to analysis. The titration program intensity of the dendrimer alone was similar to consisted of a series of 25 injections of 10 mLof that of ethidium bromide in water. dendrimer into DNA with a 5-min equilibra- tion interval between injections. Binding heats were integrated by Bind Works 3.0 software Circular Dichroism (CD) Spectroscopy (Calorimetry Sciences, Provo, UT). Apparent CD spectra of complexes (at a DNA concentration molar binding enthalpies were calculated from of 100 mgmL1) at various charge ratios were the average of the first six consistent injections acquired with a Jasco J-720 spectropolarimeter (typically two to seven), divided by the molar (Easton, MD). Reported spectra are an average of amount on a charge basis of dendrimer per in- three accumulations over a spectral range of 190– jection as described previously.8 350 nm obtained at a scan rate of 20 nm min1 with samples maintained at 208C in a 1-mm Differential Scanning Calorimetry (DSC) pathlength quartz cuvette. Spectra were buffer subtracted and converted to molar ellipticity DSC thermograms were acquired with a Microcal based on the molar concentration of the DNA capillary VP-DSC (Northampton, MA) instru- using the instrument software. Further proces- ment configured with an autosampler. All com- sing including smoothing with a Savitsky-Golay plexes were prepared in 5 mM phosphate buffer

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 STRUCTURE/ACTIVITY RELATIONSHIP FOR PAMAM DENDRIMER/DNA COMPLEXES 427 either through dialysis or drying and reconstitu- seen at the same charge ratio of 10:1. This general tion in buffer. A phosphate buffer was used trend of the highest transfection efficiency seen in these experiments because of its low tempera- with complexes of moderate generation size (4–7) ture coefficient compared to Tris. Samples were and in the presence of a significant excess of degassed and distributed in a 96-well plate held dendrimer has been previously reported.14–16 The at 48C. Thermal scans at a pressure of 40 psi, expression produced by G4 PDDCs at modest consisted of a single up scan from 20 to 1258C charge ratios (1–4) contrasts sharply with that of at a rate of 18C min1.9 Data analysis including the other generations. The transfection studies subtraction of buffer thermograms, conversion of reported here are in contrast to previous dendri- differential heat to molar heat capacity, and de- mer transfections because the dendrimers were termination of supercoil Tm values was conducted not fractured and additional transfection agents in Origin 7.0. (i.e., dextran sulfate) were not employed. This permits a more rigorous structural analysis of the complexes as described below. RESULTS Particle Sizes and Zeta Potentials In Vitro Transfection of CHO-K1 Cells The mean hydrodynamic diameter and zeta Complexes of DNA with certain polyamidoamine potential of the complexes were determined by dendrimers substantially enhanced luciferase DLS and PALS, respectively (Fig. 3a and b). activity in CHO-K1 cells. The results presented Regions in which large aggregates (1 mm) in Figure 2 demonstrate that expression levels are were formed are excluded from this analysis. In dependent on both the charge ratio of the complex general, particle sizes of complexes at high charge and the dendrimer generation employed. The G2 ratios (4–5) range over a narrow distribution of complexes produced negligible expression com- 100 to 175 nm in diameter. In the charge ratio pared to higher (G4, 7, and 9) generation PDDCs, range where complexes have significant biological with the highest generation considerably less activity (1–10), the G2 PDDCs have the largest effective than the middle two sizes. In the latter hydrodynamic diameters, while G4 PDDCs pos- three cases, expression is generally enhanced at sess the smallest dimensions. Particle sizes were higher charge ratios with maximal expression only characterized up to a charge ratio of 5; however, all complexes that show significant transfection efficiency have a minimal size at this ratio (average of 122 nm). The zeta potential of the complexes is negative at charge ratios less than unity. In general, dendrimer/DNA complexes display poor colloidal stability near neutrality (0.75–1.75), resulting in aggregation and precipitation, except complexes formed from G4 dendrimers. The zeta potential of complexes prepared with charge ratios in excess of 3 appear to form particles with a stable positive surface charge, the value of which is dependent on generation size. Generation 4 PDDCs show a high positive zeta potential (>15 mV) at all charge ratios greater than and equal to 1:1. The surface charge appears related to dendrimer generation with G2 and 4 showing a similar positive zeta potential near Figure 2. Transfection of CHO-K1 cells with PDDCs theoretical charge neutrality while G7 and 9 containing the luciferase gene plotted as relative light maintain a negative value. units per mg of protein (RLU mg1) versus charge ratio. Luciferase activity was measured 48 h posttransfection. Ethidium Bromide Displacement Luciferase activity was normalized to the overall protein content. The data represent the average of three inde- Interactions between PAMAM dendrimers and pendent measurements. Error bars (SEM) are present DNA were evaluated from the ability of dendri- but are difficult to see because of their low magnitude. mers to displace the intercalating dye, ethidium

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 428 BRAUN ET AL.

Figure 4. Ethidium bromide dye displacement by PDDCs. The relative fluorescence intensity of complexes prepared from DNA (10 mgmL1) with ethidium inter- calated at a 4:1 ratio (DNA bp to ethidium) is reported as a function of charge ratio. The data is normalized to the fluorescence intensity of unliganded DNA. The bars represent the mean of three separate measurements with error bars reflecting the standard error of the individual values.

that modify base stacking and charge repulsion between the dye and polyamine polymers.4 It appears that in this case, the mechanism at least involves exclusion of the dye from its DNA binding site into bulk solution, because the dye’s fluores- Figure 3. Mean hydrodynamic diameters (A) and cence can be enhanced by the addition of unli- zeta potential (B) of dendrimer/DNA complexes pre- ganded DNA to high charge ratio complexes (data sented as function of charge ratio. Complexes were not shown). prepared at a DNA concentration of 100 mgmL1 in 10 mM Tris buffer (pH 7.4). The data are the mean of at least three separate measurements with error bars Circular Dichroism Spectroscopy representing the standard error. CD spectroscopy was used to characterize the helical structure of the DNA within PAMAM bromide, from the DNA.4 The smallest dendrimer, dendrimer/DNA complexes. The spectrum of un- G2, predictably displaces the least ethidium liganded plasmid (Fig. 5A) shows the character- bromide, a modest 57% at the highest charge istics of the native B form conformation, a positive ratio examined (5) (Fig. 4). The largest dendrimer, band centered near 275 nm, and a negative band G9, appears to have the greatest relative affinity near 245 nm. None of the dendrimers manifested for DNA, displacing 81% of the dye at charge any optical activity. The spectra of individual ratios in excess of 2. Thus, the relative binding complexes undergo a charge ratio-dependent de- affinity of PAMAM dendrimers appears to crease in intensity of the 275 nm CD bands and a correlate with a generation size in the order red shifting of the peak position of both the 275 G9 G7 > G4 > G2. and 245 nm signals. At a charge ratio of 0.5, the This indirect means of determining binding 275 nm peak is almost completely lost and the affinity may be misleading, however, because associated 245 nm trough is dramatically intensi- multiple mechanisms for the displacement have fied. Both bands are significantly red shifted but been suggested, including conformational changes not to the same degree. As the positive charge

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 STRUCTURE/ACTIVITY RELATIONSHIP FOR PAMAM DENDRIMER/DNA COMPLEXES 429

Figure 5. Representative CD spectra of G4 dendrimer/DNA complexes normalized as molar ellipticity values (A). Plots of the intensity of the 245 nm band (B), the position of 245 nm band (C), the intensity of 275 nm band (D) and the position of 275 nm band (E) versus dendrimer/DNA charge ratio. Samples were prepared at a DNA concentration of 100 mgmL1 in 10 mM Tris buffer (pH 7.4). Individual points are the mean of at least three measurements with error bars representing the standard error.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 430 BRAUN ET AL. increases (charge ratios 3 and 5), spectra continue strong COH stretch from the incomplete to manifest altered shapes in comparison to removal of methyl alcohol used as the dendrimer unliganded DNA, but with intensities and posi- solvent. In spectra of complexes, the dendrimer tions shifted to a lesser degree than seen at lower bands were poorly resolved, and could not be re- charge ratios. Both intensity and peak position liably monitored. The position of the five DNA are summarized for each dendrimer generation as bands, three conformationally sensitive signals a function of charge ratio in Figure 5B–E. The (1715, 1224, and 970 cm1) as well as two DNA gaps in the data near charge neutrality are due to backbone vibrations considered insensitive to the exclusion of data due to sample precipitation. conformation (the symmetric phosphate stretch In general, the maximal spectral perturbation at 1086 cm1 and the antisymmetric CO stretch from the unliganded DNA spectra occurs in of furanose at 1053 cm1), are plotted as a func- complexes prepared near charge neutrality. tion of complex charge ratio in Figure 6B–E. The Complexes with a charge excess of dendrimer conformationally sensitive DNA backbone band exhibit spectra with intensities and peak posi- (970 cm1) and G/T carbonyl stretch (1715 cm1) tions still significantly altered from the native both increase in frequency upon dendrimer bind- plasmid but give nearly identical spectra at ing up to a 1:1 charge ratio while the asym- charge ratios equal to 3. The spectra of DNA in metric phosphate stretch (1224 cm1) decreases in complexes vary in the intensity of the B form frequency over the same range. Although these marker bands. The 245-nm trough increases in conformationally sensitive peaks shift signifi- intensity and the 275-nm peak decreases, with cantly when DNA is bound to the dendrimers, the degree of perturbation from unliganded DNA the positions remain indicative of only the B following the trend G2 > G4 > G9 G7. Accompa- conformation. nying the change in rotational strength of com- The shifts in position of the vibrations arising plexes is a red shift in peak position. The red from both backbone and bases suggest some direct shifts follow the same trend as the intensity dendrimer/DNA interaction occurs outside of the changes with respect to dendrimer generation. expected phosphate–amine electrostatic interac- The red shifts in the predominant bands vary tions. The position of a DNA ribose sugar vibration considerably, with the 245 nm trough shifting as (1053 cm1) shifts significantly to lower frequency much as 7 nm while the 275 nm peak exhibits shifts of nearly 18 nm (G2). This dramatic peak shifting appears to be related to generation size with smaller dendrimers (2 and 4) near charge neutrality manifesting the largest red shift.

Fourier-Transform Infrared (FTIR) Spectroscopy FTIR spectroscopy was employed as a com- plimentary technique to further investigate the secondary structure of the DNA component of complexes. The FTIR spectra of DNA, several G4 dendrimer/DNA complexes and G4 dendrimer alone are compared in Figure 6A. The DNA spectrum in the absence of polymer is indicative of B form with conformationally sensitive vibrations arising from guanine/thymine carbonyl stretching (1715 cm1), coupled sugar-base thymine vibrations (1328 cm1 and 1281 cm1), anti- Figure 6. Representative FTIR absorbance spectra of symmetric phosphate stretching (1224 cm1), G4 dendrimer, DNA, and complexes (A). Plots of selected coupled phosphodiester/deoxyribose backbone 1 peak positions of G2 PDDCs (B), G4 PDDCs (C), G7 stretching (970 cm ), and ribose ring vibrations PDDCs (D) and G9 PDDCs (E) as a function of charge 1 2,10 (897 cm ). The dendrimer spectra have three ratio. All spectra are of complexes prepared at a DNA major absorption bands, two broad peaks arising concentration of 1 mg mL1. The peak position values from amine deformation (1635 cm1) and amide and error bars (SEM) reﬂect an average of ﬁve CNH stretching (1560 cm1)17 as well as a measurements.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 STRUCTURE/ACTIVITY RELATIONSHIP FOR PAMAM DENDRIMER/DNA COMPLEXES 431

Figure 6. (Continued)

(2.3 cm1) in both G7 and 9 PDDCs. Further- The decrease seen in the frequency of the anti- more, as indicated above, the increases in fre- symmetric phosphate stretch (1223 cm1) with quency of the base carbonyl vibrations suggest polymer addition up to a 1:1 ratio also suggests altered hydrogen bonding occurs within the bases. (as expected) this moiety is directly involved in

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 432 BRAUN ET AL. the dendrimer/DNA interaction,10 presumably through a direct electrostatic interaction with dendrimer amino groups.

Isothermal Titration Calorimetry (ITC) To investigate the thermodynamics of binding of dendrimers to DNA, we conducted ITC analysis of each dendrimer generation at 258C and low ionic strength (<0.01 M). Unfortunately, due to aggregation prior to saturation of binding (at charge neutrality), the number of binding sites or an affinity constant could not be determined. An apparent binding enthalpy (DHapp), however, could be estimated by averaging the observed constant binding heats prior to aggregation. These results are summarized in Table 1. Den- drimer generation appears to have little effect Figure 7. The DSC melting temperature of the super- on the binding energetics. The average binding 1 coiled form of DNA (scDNA) is plotted as a function of enthalpy of 7.6 kJ mol observed for the four charge ratio. Individual samples at 0.5 mg mL1 DNA complexes is consistent with electrostatic interac- concentration were prepared at selected charge ratios in 8,18–21 tions dominating the complexation process. phosphate buffer. The plot is representative of a single measurement at each dendrimer/DNA charge ratio but in general the Tms vary by less than 0.48C. Differential Scanning Calorimetry (DSC) DSC was used to evaluate the thermal stability of 9 show similar destabilization at low charge ratio DNA complexed to PAMAM dendrimers. Thermo- and stabilization at high ratios. Maximal tem- grams of DNA display two major melting transi- perature stabilization occurs at a two to three tions. The first originates from linear/open charge ratio regardless of the dendrimer genera- circular forms and occurs between 60 and 708C, tion employed. The extent to which the SC Tm is while the second arises from supercoiled DNA altered does not obviously correlate with the and appears near 1098C (data not shown).9 transfection efficiency of the various dendrimer The thermograms of PAMAM dendrimers alone generations. display no thermal transitions. The melting temperature of the supercoiled plasmid complexed to each generation of dendrimer is plotted versus DISCUSSION charge ratio in Figure 7. The supercoiled form is thermally destabilized when complexed This study attempts to establish a structure– to dendrimers below charge neutrality (0.5). function relationship between the biophysical Complexes prepared with a significant positive properties of dendrimer/DNA complexes and their charge excess (>2), however, show thermal stabi- in vitro transfection efficiencies. A number of lization of the supercoiled form. Dendrimer/DNA biophysical measurements including size and zeta complexes prepared from generations 2, 4, 7, and potential measurement, dye (ethidium bromide) displacement, DSC, CD, and FTIR show significant alterations as the ratio of dendrimer to Table 1. Binding Enthalpies of PAMAM Dendrimers DNA is varied. These changes appear to be great- to Plasmid DNA at 258C Using ITC est up to the point the amount of dendrimer Ligand DH (kJ/mol) becomes equal to that of DNA on a charge basis. These results suggest structural rearrangements Generation 2 Dendrimer 7.6 0.5 in both DNA and the complex itself as charge Generation 4 Dendrimer 9.3 1.3 neutrality is reached. By comparison, PDDCs that Generation 7 Dendrimer 7.4 0.1 show the greatest biological activity (G4 at a 1–10 Generation 9 Dendrimer 6.1 0.9 charge ratio and G7 at 2–10 charge ratios) exhibit

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 STRUCTURE/ACTIVITY RELATIONSHIP FOR PAMAM DENDRIMER/DNA COMPLEXES 433 their maximum activity at charge ratios above hexagonal phase)24 and manifest both low rela- charge neutrality. All four dendrimer generations tive affinity for DNA4 and reduced transfection examined also show optimal structural stability efficiency when prepared with a cationic charge when the dendrimer is in charge excess (2). A excess.5 In contrast, cationic vehicles with a number of studies of various nonviral gene moderate to high relative affinity seem to form delivery vehicles suggest that complexes formed complexes with similar overall morphologies at in significant charge excess reach a structurally least with respect to the properties discussed stable state that differs depending on the chem- above. Peptoids demonstrate a correlation similar istry of the vehicle. Thus, the total of the evidence to G4 and G7 dendrimers with a high transfec- suggests that structural features of the complexes tion efficiency and moderate relative affinity.12 A may be related in some way to their individual number of cationic vehicles have high relative transfection efficiencies, but the details of these affinities but differ in their transfection efficien- relationships are currently obscure. cies. Both PEI (750 kDa) and G9 dendrimer have The colloidal properties (hydrodynamic dia- high relative affinity4 but lower transfection meter and zeta potential) of PDDCs do not efficiencies.10 This observation suggests an upper independently appear to correlate with transfec- threshold may exist in which high affinity reduces tion efficiency. Further analysis of the size and transfection, perhaps by inhibiting DNA release. zeta potential of complexes at charge ratios of It has been previously shown that G9 dendr- 2–5 where G4 and G7 complexes differ in their imer complexes inhibit luciferase expression to a transgene expression, do suggest a correlation in greater extent than lower generations (G5 and 7) which low (4–8) size to charge ratios show high in an in vitro gene expression system.25 An ex- transfection efficiency. Although this correlation ception to the latter may exist, however, in that exists for G4 and 7 PDDCs, the existence of G2 many cationic lipids display both high dye binding and 9 PDDCs with size/charge ratios within the affinity and high transfection activity.4,5,12 The postulated range but without significant transfec- difference in the chemistry between lipid and tion activity suggests that a narrow range of size/ polymer vehicles at least leaves open the possibi- charge ratios may be necessary but insufficient lity that a different DNA release mechanism to explain relative transfection abilities. Previous between the two cations may account for this studies of nonviral vectors composed of cationic discrepancy. The observation that some fraction of lipids, peptoids/lipitoids, or PEI show contrasting the ethidium is excluded suggests that displace- results with regard to any size/charge correlation. ment may be mediated by condensation of the Polyethylenimine complexes formed at N/P ratios DNA and dendrimer to a more compact state. This with a positive zeta potential (4) manifest no mechanism is consistent with the trend in which evidence of a size/charge correlation with transfec- the largest dendrimers (highest generations) tion efficiency.10 Characterization of cationic lipid apparently displaced the greatest fraction of dye. and peptoid/lipitoid transfection active complexes It should be remembered, however, that neither find size/charge ratios within the postulated range ITC nor ethidium bromide exclusion provide an but the limited number of observations preclude unambiguous measure of the affinity of the further refinement of any correlation from this cationic vehicle for DNA. data.2,5,12 A similar charge density correlation has The characterization of PAMAM dendrimer/ recently been proposed for cationic lipid complexes DNA complexes reveals no other obvious correla- that adopt a lamellar organization.22 tions with transfection efficiency. Importantly, The two generations of dendrimer that show however, biophysical methods do reveal a number the greatest biological activity (G4 and 7) both of common features among a wide variety of non- have an intermediate relative affinity for the DNA viral vectors as well as a few measurements as determined by ethidium bromide displace- that may sensitively reflect differences in vector ment. The lowest affinity dendrimer (G2) has chemistry. It appears that the DNA conformation poor transfection efficiency, possibly related to a within nonviral gene delivery complexes of a complex morphology that is different from G4– wide variety of classes (cationic lipids, peptoids/ G9 complexes as suggested by its abnormal gel lipitoids, PEI, and dendrimers) remains in an mobility14 and greater fluorescent dye accessibil- overall B-form helix, although localized altera- ity in direct addition experiments.23 Cationic tions in DNA structure may occur within com- lipids incorporating the helper lipid DOPE simi- plexes as reflected by FTIR position shifts and larly exist in an altered morphology (inverted altered CD spectra (see extensive discussion

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 434 BRAUN ET AL. within ref. 7). These alterations show a strong transcriptional activity for both dendrimer/DNA dependence on the individual chemistries of the and PEI/DNA complexes.25,26 In fact, studies cationic vehicle. The thermal stability of the that monitor the individual events of the gene scDNA melting transition also shows a depen- delivery process (e.g., confocal microscopy and dence on the nature of the cationic component. flow cytometry approaches) suggest that endo- Both dendrimers and cationic lipids stabilize somal release, nuclear localization, or transcrip- supercoiled DNA when complexes are prepared tion are the rate limiting processes rather than in positive charge excess,9 while PEI complexes at cellular uptake.27–32 Thus, attempts to establish charge excess either stabilize the scDNA transi- correlation between the individual stages of the tion (750 and 25 K branched forms) or destabilize (2 transfection process and the physical properties of K branched and 25 K linear) depending on the nonviral gene delivery complexes would seem to geometry and/or molecular weight of the poly- have a higher probability of success than attempts mer.10 to correlate such properties with gross measures of Biophysical characterization of PAMAM den- gene expression. drimer/DNA complexes does suggest possible Fractionating complexes based on relative correlations between transfection efficiency and affinity, size, or density could isolate active sub- particle size/charge ratio and the relative affinity species that manifest stronger correlations. Early of the PAMAM dendrimer and DNA components. attempts to separate cationic lipid/DNA complexes These correlations are not apparent, however, by density-gradient centrifugation, however, have between chemically dissimilar nonviral vehicles. met with only moderate success.33–36 It might The lack of any all-inclusive correlation may be also be possible to establish correlations by focus- due to the different distribution of amine binding on the individual events of the transfection ing moities among differing vehicles or could be process (e.g., cellular entry, endosomal release, a result of complexes consisting of a number of and nuclear translocation) as discussed above subpopulations varying in size, density, and rather than the more complex end point of cellular charge state due to the nonspecific electrostatic transfection. Some progress has been reported interactions involved in their assembly. Although in this regard. A cholesterol-dependent pathway we attempted to reduce cell-type variability in our has been identified for the cellular uptake of den- studies by using the same cell line (CHO-K1) in drimer/DNA complexes,37 and a potential path- a series of studies with different gene delivery way by which dendrimer/DNA complexes may systems,5,10,12 we cannot rule out cell-type varia- escape endosomes has been detected in an anionic bility as another factor in our inability to establish liposome model.38 a comprehensive correlation.14 Our primary goal was to establish structure/ function relationships that are independent of the ACKNOWLEDGMENTS individual steps of the transfection process. It appears likely, however, that efficient transfection The authors wish to thank both Valentis, Inc. requires balance exist among various structural and Dendritic Nanotechnologies for their sup- properties and their role in transgressing the port of this work. C.S.B. acknowledges financial multiple barriers to gene delivery. Thus, we can support from The United States Pharmacopeia speculate that an optimal size/charge ratio, con- Fellowship, and JAV financial support from an sisting of a small size (<150 nm dia.) and positively American Heart Association Postdoctoral Fellow- charged surface, may be important for the cell ship (0225425Z). association and endocytosis processes. In fact, an optimal size/charge ratio for overall transfection efficiency has been reported for at least one cationic lipid/DNA gene delivery system.22 The REFERENCES dye displacement studies reported provide some 1. Wiethoff CM, Middaugh CR. 2003. Barriers to non- indication of the extent of compaction of DNA viral gene delivery. J Pharm Sci 92:203–217. when bound to PAMAM dendrimers. Compaction 2. Choosakoonkriang S, Wiethoff CM, Anchordoquy seems to favor gene delivery by providing small TJ, Koe GS, Smith JG, Middaugh CR. 2001. In- stable particles that protect the DNA from nucle- frared spectroscopic characterization of the inter- ase degradation. Condensation to compact struc- action of cationic lipids with plasmid DNA. J Biol tures, however, has also been shown to inhibit Chem 276:8037–8043.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 STRUCTURE/ACTIVITY RELATIONSHIP FOR PAMAM DENDRIMER/DNA COMPLEXES 435

3. Wiethoff CM, Gill ML, Koe GS, Koe JG, Middaugh 16. Haensler J, Szoka FC Jr. 1993. Polyamidoamine CR. 2002. The structural organization of cationic cascade polymers mediate efficient transfection of lipid–DNA complexes. J Biol Chem 277:44980– cells in culture. Bioconjug Chem 4:372–379. 44987. 17. Kolhe P, Misra E, Kannan RM, Kannan S, 4. Wiethoff CM, Gill ML, Koe GS, Koe JG, Middaugh Lieh-Lai M. 2003. Drug complexation, in vitro CR. 2003. A fluorescence study of the structure release and cellular entry of dendrimers and and accessibility of plasmid DNA condensed with hyperbranched polymers. Int J Pharm 259:143– cationic gene delivery vehicles. J Pharm Sci 92: 160. 1272–1285. 18. Lobo BA, Koe GS, Koe JG, Middaugh CR. 2003. 5. Wiethoff CM, Koe JG, Koe GS, Middaugh CR. 2004. Thermodynamic analysis of binding and protona- Compositional effects of cationic lipid/DNA delivery tion in DOTAP/DOPE (1:1): DNA complexes using systems on transgene expression in cell culture. isothermal titration calorimetry. Biophys Chem J Pharm Sci 93:108–123. 104:67–78. 6. Wiethoff CM, Smith JG, Koe GS, Middaugh CR. 19. Matulis D, Rouzina I, Bloomfield VA. 2002. Ther- 2001. The potential role of proteoglycans in modynamics of cationic lipid binding to DNA cationic lipid-mediated gene delivery. Studies of and DNA condensation: Roles of electrostatics the interaction of cationic lipid–DNA complexes and hydrophobicity. J Am Chem Soc 124:7331– with model glycosaminoglycans. J Biol Chem 7342. 276:32806–32813. 20. Matulis D, Rouzina I, Bloomfield VA. 2000. Ther- 7. Braun CS, Jas GS, Choosakoonkriang S, Koe GS, modynamics of DNA binding and condensation: Smith JG, Middaugh CR. 2003. The structure of Isothermal titration calorimetry and electrostatic DNA within cationic lipid/DNA complexes. Biophys mechanism. J Mol Biol 296:1053–1063. J 84(2 Pt 1):1114–1123. 21. Ross PD, Subramanian S. 1981. Thermodynamics 8. Lobo BA, Davis A, Koe G, Smith JG, Middaugh of protein association reactions: Forces contribut- CR. 2001. Isothermal titration calorimetric analy- ing to stability. Biochemistry 20:3096–3102. sis of the interaction between cationic lipids and 22. Lin AJ, Slack NL, Ahmad A, George CX, Samuel plasmid DNA. Arch Biochem Biophys 386:95– CE, Safinya CR. 2003. Three-dimensional imaging 105. of lipid gene-carriers: Membrane charge density 9. Lobo BA, Rogers SA, Choosakoonkriang S, Smith controls universal transfection behavior in lamellar JG, Koe G, Middaugh CR. 2002. Differential scan- cationic liposome–DNA complexes. Biophys J 84: ning calorimetric studies of the thermal stability of 3307–3316. plasmid DNA complexed with cationic lipids and 23. Chen W, Turro NJ, Tomalia DA. 2000. Using polymers. J Pharm Sci 91:454–466. ethidium bromide to probe the interactions be- 10. Choosakoonkriang S, Lobo BA, Koe GS, Koe JG, tween DNA and dendrimers. Langmuir 16:15–19. Middaugh CR. 2003. Biophysical characterization 24. Koltover I, Salditt T, Radler JO, Safinya CR. 1998. of PEI/DNA complexes. J Pharm Sci 92:1710– An inverted hexagonal phase of cationic liposome– 1722. DNA complexes related to DNA release and deli- 11. Huang CY, Uno T, Murphy JE, Lee S, Hamer JD, very. Science 281:78–81. Escobedo JA, Cohen FE, Radhakrishnan R, Dwarki 25. Bielinska AU, Kukowska-Latallo JF, Baker JR Jr. V, Zuckermann RN. 1998. Lipitoids—Novel catio- 1997. The interaction of plasmid DNA with poly- nic lipids for cellular delivery of plasmid DNA in amidoamine dendrimers: Mechanism of complex vitro. Chem Biol 5:345–354. formation and analysis of alterations induced in 12. Lobo BA, Vetro JA, Suich DM, Zuckermann RN, nuclease sensitivity and transcriptional activity of Middaugh CR. 2003. Structure/function analysis the complexed DNA. Biochim Biophys Acta 1353: of peptoid/lipitoid:DNA complexes. J Pharm Sci 92: 180–190. 1905–1918. 26. Bieber T, Meissner W, Kostin S, Niemann A, 13. Ruiz J, Lafuente G, Marcen S, Ornelas C, Lazare S, Elsasser HP. 2002. Intracellular route and tran- Cloutet E, Blais JC, Astruc D. 2003. Construction of scriptional competence of polyethylenimine–DNA giant dendrimers using a tripodal building block. complexes. J Controlled Release 82:441–454. J Am Chem Soc 125:7250–7257. 27. Godbey WT, Wu KK, Mikos AG. 1999. Tracking the 14. Kukowska-Latallo JF, Bielinska AU, Johnson J, intracellular path of poly(ethylenimine)/DNA com- Spindler R, Tomalia DA, Baker JR Jr. 1996. plexes for gene delivery. Proc Natl Acad Sci USA Efficient transfer of genetic material into mamma- 96:5177–5181. lian cells using Starburst polyamidoamine dendri- 28. Tseng WC, Haselton FR, Giorgio TD. 1997. Trans- mers. Proc Natl Acad Sci USA 93:4897–4902. fection by cationic liposomes using simultaneous 15. Tang MX, Redemann CT, Szoka FC Jr. 1996. In single cell measurements of plasmid delivery and vitro gene delivery by degraded polyamidoamine transgene expression. J Biol Chem 272:25641– dendrimers. Bioconjug Chem 7:703–714. 25647.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005 436 BRAUN ET AL.

29. James MB, Giorgio TD. 2000. Nuclear-associated 34. Xu Y, Hui SW, Frederik P, Szoka FC Jr. 1999. plasmid, but not cell-associated plasmid, is corre- Physicochemical characterization and purification lated with transgene expression in cultured mam- of cationic lipoplexes. Biophys J 77:341–353. malian cells. Mol Ther 1:339–346. 35. Hofland HE, Shephard L, Sullivan SM. 1996. 30. Remy-Kristensen A, Clamme JP, Vuilleumier C, Formation of stable cationic lipid/DNA complexes Kuhry JG, Mely Y. 2001. Role of endocytosis in the for gene transfer. Proc Natl Acad Sci USA 93:7305– transfection of L929 fibroblasts by polyethyleni- 7309. mine/DNA complexes. Biochim Biophys Acta 1514: 36. Smith JG, Wedeking T, Vernachio JH, Way H, 21–32. Niven RW. 1998. Characterization and in vivo test- 31. Banks GA, Roselli RJ, Chen R, Giorgio TD. 2003. A ing of a heterogeneous cationic lipid–DNA formu- model for the analysis of nonviral gene therapy. lation. Pharm Res 15:1356–1363. Gene Ther 10:1766–1775. 37. Manunta M, Tan PH, Sagoo P, Kashefi K, George 32. Zaric V, Weltin D, Erbacher P, Remy JS, Behr JP, AJ. 2004. Gene delivery by dendrimers operates Stephan D. 2004. Effective polyethylenimine- via a cholesterol dependent pathway. Nucleic Acids mediated gene transfer into human endothelial Res 32:2730–2739. cells. J Gene Med 6:176–184. 38. Zhang ZY, Smith BD. 2000. High-generation poly- 33. Mok KW, Cullis PR. 1997. Structural and fusogenic cationic dendrimers are unusually effective at dis- properties of cationic liposomes in the presence of rupting anionic vesicles: Membrane bending model. plasmid DNA. Biophys J 73:2534–2545. Bioconjug Chem 11:805–814.

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 94, NO. 2, FEBRUARY 2005