Available online at www.sciencedirect.

com

Colloids and Surfaces B: Biointerfaces 61 (2008) 53–60

The effect of pharmaceuticals on the nanoscale structure of

PEO–PPO–PEO micelles
Praveen K. Sharma, Meghan J. Reilly, Deanna N. Jones,
Paul M. Robinson, Surita R. Bhatia ∗
Department of Chemical Engineering, 159 Goessmann Laboratory, University of Massachusetts, Amherst,
MA 01003-9303, United States
Received 6 June 2007; received in revised form 9 July 2007; accepted 12 July 2007
Available online 19 July 2007

Abstract
We present results on the effects of various hydrophobic drugs and additives on the micellar structure of Pluronic F127 solutions. Small-angle
neutron scattering experiments on 5 wt% F127 solutions were used to measure micelle core size (R1 ), micelle corona size (R2 ), intermicellar
interaction distance (Rint ), polydispersity (σ), and aggregation number (Nagg ); dynamic light scattering was used to measure critical micelle
concentration (CMC); and ultraviolet spectroscopy was used to measure drug solubility and apparent micelle–water partition coefficient (Kmw ).
The core and corona size were found to generally increase in the presence of the drugs, as did Rint . Both σ and Nagg were found to decrease in the
presence of most of the drugs, and the CMC was found to vary considerably with no clear correlation. A design of experiments (DOE) approach
was used to analyze the results and build empirical correlations. All of the parameters from the SANS experiments were found to depend strongly
on drug solubility, with a weak dependence on Kmw in most cases. The aggregation number, however, was found to depend strongly on both Kmw
and solubility. The correlations can be used to roughly predict the structural parameters of F127 micelles for other hydrophobic drugs.
© 2007 Elsevier B.V. All rights reserved.

Keywords: SANS; Pluronic; Poloxamers; Paclitaxel; Drug delivery

1. Introduction mulations of F127 with solubilized drugs can be administered

as cold liquids before becoming gels on contact with warm body
Pluronic F127 is a member of a family of triblock copolymers tissue. Controlled and sustained release of the solubilized drugs
that has generated much interest in the field of controlled release is then achieved by slow diffusion through the aqueous chan-
drug delivery due to its ability to form gels in response to changes nels surrounding the micelles. Controlled release of this kind
in temperature [1]. It consists of poly(ethylene oxide) (PEO) has been generally found to be advantageous in drug delivery
and poly(propylene oxide) (PPO) monomers in the arrangement applications that suffer from undesirable side effects or low-drug
PEO100 PPO70 PEO100 . In an aqueous solvent the relative differ- efficacy [3].
ence in hydrophobicity between PPO and PEO allows for the Once a drug–polymer system has been classified as being
formation of self-assembled micellar structures, whereby cores stable, it is of primary importance to know the particle size
of PPO and water are surrounded by coronas consisting of PEO and aggregation number in a delivery system [1]. For example,
and water [1]. At moderate concentrations and ambient tempera- solubilization of a drug in a copolymer formulation is highly
ture, micellar solutions of F127 are found to exist in a disordered dependent upon the structural properties of the micelles, such
liquid state. Upon increasing the temperature, the micellar cores as the size of the micelle cores and number of chains involved
become dehydrated [2] resulting in an entropic change of state in the formation of a micelle (i.e., the aggregation number) [1].
into a partly ordered cubic liquid crystalline gel state [1]. For- Furthermore, information on the micelle size, aggregation num-
ber and distance between micelles could allow for the prediction
of in vitro drug release profiles (with the assumption of a certain
∗ Corresponding author. Tel.: +1 413 545 0096; fax: +1 413 545 1647. transport model), which could be very useful in assessing pos-
E-mail address: sbhatia@ecs.umass.edu (S.R. Bhatia). sible delivery routes and performance characteristics for newly

0927-7765/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.colsurfb.2007.07.002
54 P.K. Sharma et al. / Colloids and Surfaces B: Biointerfaces 61 (2008) 53–60

discovered drugs that are in low supply. It is therefore important previously [4]. In order to calculate Kmw , data from both the
to understand and predict the effects that drugs and additives SANS experiments and DLS experiments were used. Reported
may have on the micellar structure of copolymer solutions. To Kmw values are at 25 ◦ C.
our knowledge, there have been relatively few studies to date
concerning the effects of high molecular weight solutes and 2.3. Dynamic light scattering
pharmaceuticals on the structure of Pluronic F127 micelles. Our
previous work involved the effects of two anti-inflammatory The critical micelle concentration of F127 in water with each
agents, naproxen and indomethacin, on the micellar structure drug was estimated using dynamic light scattering (DLS). For
and gelation behavior of F127 solutions [4]. In the present study, each drug system, a stock solution of 5 wt% F127 in water was
we extend this investigation further to other hydrophobic drugs made. An excess amount of the appropriate drug was then added
in order to examine if there are changes in the structure that can to this solution, before filtering through a 0.45 ␮m Millex-HV
be generalized to gain a better fundamental understanding. syringe driven filter. Several samples at F127 concentrations
We carried out small-angle neutron scattering (SANS) exper- between 0.01 and 1 wt% were made from this stock solution.
iments to measure the structural properties of F127 micelles. Dust was removed from the samples by again filtering slowly
We also carried out dynamic light scattering experiments and through a 0.45 ␮m Millex-HV syringe driven filter (Millipore
ultraviolet spectroscopy experiments to determine the criti- Corp.). The filtrates were put into clean light scattering tubes
cal micelle concentration (CMC), the drug solubility, and the that had been thoroughly washed with nanopure water, dried and
apparent micelle–water partition coefficient (Kmw ). From these cleaned with a Kensington Dust Blaster compressed gas duster.
experiments, changes can be observed in structure and self- The samples were placed into a decahydronaphthalene vat jack-
assembly that may result from the presence of hydrophobic eted by a temperature-controlled water bath with a setpoint of
drug molecules in F127 formulations. The drugs we have 25 ◦ C. Each sample was allowed to equilibrate to the waterbath
studied were from various classes, including local anesthet- temperature for at least 30 min in the BI 200SM Goniometer
ics (dibucaine, lidocaine, and tetracaine), anti-inflammatory setup (Brookhaven Instruments Corp.) before data collection
agents (sulindac), and anticancer agents (paclitaxel and a related was started. Light scattering experiments were carried out using
secondary metabolite, baccatin III). All of these drugs are con- a 514.4 nm Lexel 95 8 W Argon Ion Laser working at a constant
sidered to be relatively hydrophobic in nature, which makes power output of 200 mW. A Brookhaven Instruments photomul-
them difficult to deliver safely and effectively through con- tiplier detector tube (in photon counting mode) was kept at a
ventional means, but potentially good candidates for delivery fixed angle of 90◦ to the incident beam path, with a pinhole aper-
via Pluronic F127 formulations. In addition, we also studied ture size of 200 ␮m. For each sample, time-dependent intensity
the pharmaceutical additives methyl-4-hydroxybenzoate, ethyl- fluctuation data was collected from the instrument and correlated
4-hydroxybenzoate, and propyl-4-hydroxybenzoate (methyl-, with a BI 9000AT digital correlator (Brookhaven Instruments
ethyl-, and propyl paraben), which are sometimes added to phar- Corp.) over a delay range of 25 ns to 100 ms (using 342 channels
maceutical and personal care formulations as preservatives and plus 4 extended channels). Analysis of the autocorrelation func-
antifungal components. tion in terms of particle size distribution was done numerically
using a non-negatively constrained least squares (Regularized
2. Materials and experimental methods CONTIN) method over a particle size range of 1.00–100 nm.
All samples showed a small diameter population corresponding
2.1. Materials to unimers, and samples above the CMC also showed a micel-
lar population of greater diameter. The lowest concentration at
Pluronic F127, dibucaine, lidocaine, tetracaine, paclitaxel, which this micellar population was observed was taken to be the
baccatin III, sulindac, methyl paraben, ethyl paraben, and propyl critical micelle concentration.
paraben were obtained from Sigma–Aldrich Co. and used with-
out further purification. All solutions of F127 were made on 2.4. Small-angle neutron scattering
a mass basis using the methods described by Schmolka [5].
Depending on the experimental technique, solutions of F127 SANS was used to measure the structural properties of the
were either made in deuterium oxide (99.9% D) from Cambridge F127 micelles with each of the drugs dissolved at their saturation
Isotope Laboratories, Inc., or nanopure water obtained from a concentrations. A stock solution of 5 wt% F127 in deuterium
Barnstead NANOpure Infinity UF filtration unit. The solubil- oxide (99.9% D) was made. An excess amount of each drug
ity experiments utilized ethanol (from Pharmaco Products, Inc.) was then added to a 5 ml sample of the stock solution. Each
that was 200 proof, absolute (dehydrated). super-saturated sample was then stirred and allowed to equili-
brate before being filtered slowly through a 0.45 ␮m Millex-HV
2.2. Ultraviolet spectroscopy syringe driven filter. SANS experiments on these samples were
performed on the small-angle diffractometer (SAD) at the
Ultraviolet spectroscopy (UV) was used to measure the sol- intense pulsed neutron source (IPNS) at Argonne National Lab-
ubility and calculate the Kmw of each drug in F127 solutions. oratory (ANL). Quartz sample cells of path length 2.0 mm were
The details of the experimental technique used for these mea- used and spectra were collected at 25 ◦ C for up to 1 h each.
surements and calculations to obtain Kmw have been described Deuterium oxide was used to quantify scattering from the sol-
 P.K. Sharma et al. / Colloids and Surfaces B: Biointerfaces 61 (2008) 53–60 55

vent, which was subsequently subtracted off. For each sample, Table 2
the signal at high q was used to estimate the incoherent back- Critical micelle concentration of Pluronic® solutions
ground scattering that was added to the calculated intensity of System CMC (wt%)
the data-fitting model. Paclitaxel 0.15 ± 0.03
Baccatin III 0.71 ± 0.02
Sulindac 0.45 ± 0.02
3. Results and discussion Dibucaine 0.50 ± 0.01
Tetracaine 0.22 ± 0.04
3.1. Ultraviolet spectroscopy Lidocaine 0.45 ± 0.02
Methyl paraben 0.10 ± 0.02
Ethyl paraben 0.16 ± 0.03
The apparent micelle–water partition coefficients, Kmw , esti-
Propyl paraben 0.20 ± 0.02
mated for all the drugs studied are given in Table 1. In addition, Naproxen 0.33 ± 0.04
Table 1 also shows the solubility measured for each drug in water Indomethacin 0.24 ± 0.01
and 10 wt% F127. The Kmw indicates the affinity of a drug for 5 wt% Pluronic® F127 0.26 ± 0.03
the PPO cores of the micelles as opposed to the PEO–water
The critical micelle concentration (CMC) of Pluronic® F127 solutions in the
regions outside of the cores. This parameter therefore gives an presence of each pharmaceutical, as measured from dynamic light scattering
indication of the general distribution of the drug in the micellar (DLS) experiments. The CMC in the absence of pharmaceutical solutes is also
system. All the drug solutes chosen for this study are gener- given. The errors in CMC are based on the increments in copolymer concentra-
ally considered to be “hydrophobic” drugs. It is interesting to tion for the samples.
note the wide spread (from 20 to 4510) of Kmw observed in
our study. Tetracaine is considered to be a fairly hydrophobic 3.2. Dynamic light scattering
drug; the lipid–water partition coefficient in DMPC vesicles has
been estimated to be about 3160 [6]. However, the apparent The critical micelle concentration of Pluronic® F127 in the
micelle–water partition coefficient in F127 we have estimated absence and presence of pharmaceutical solutes was measured
here is only 162. The difference in absolute values is due to the using dynamic light scattering. The CMC was estimated as
difference in the hydrophobic regions of DMPC vesicles and the concentration at which the average hydrodynamic diame-
F127 micelles. Hence, when using partition coefficient data for ter suddenly increased; it is given for each solute system in
behavior prediction, it is important to carefully consider any Table 2. Table 2 shows the CMC of Pluronic® F127 in water
differences between the actual phases used to measure the data and the absence of any solutes to be about 0.26 ± 0.03 wt%,
and those for which the predictions are required. For exam- which is of the same order as values previously reported by oth-
ple, lipid–water partition coefficients are useful for predicting ers. The present measurement is somewhat close to the value
in vivo absorption of drug molecules [7]; however, they cannot of 0.12 w/v% reported by Wanka et al. [2], but is considerably
be used for the assessment of polymeric drug delivery for- lower than 0.7 w/v% as reported byAlexandridis et al. [8], and
mulations. Micelle–water partition coefficients estimated from 2 wt% as reported by Desai et al. [9]. In comparison to Pluronic®
actual F127 solutions are necessary for the accurate prediction of F127 in water alone, it appears that the presence of some of the
drug distribution (and hence initial concentration for release) in pharmaceuticals caused an enhancement of micellar assembly,
F127 formulations. Major implications for F127 formulations which was exhibited by a decrease in the CMC. Methyl paraben,
include the prediction of in vitro release profiles, the assess- ethyl paraben, and propyl paraben showed a systematic decrease
ment of exposure to clearance and degradation mechanisms, and in the CMC, suggesting that within a series of chemically simi-
a better understanding of the suitability of a drug in an F127 lar solutes, the length of an alkyl chain present on a solute (and
formulation [4]. hence the solute hydrophobicity) may determine the extent to

Table 1
Solubility and partition coefficient of pharmaceuticals

System Csat
0
(wt%) (in water) Csat (wt%) (in 10 wt% copolymer) Kmw

Paclitaxel 0.000028 ± 0.000002 0.00214 ± 0.00011 4510 ± 982

Baccatin III 0.00457 ± 0.00004 0.06362 ± 0.00168 561 ± 179
Sulindac 0.03361 ± 0.00054 0.09091 ± 0.00099 82 ± 8
Dibucaine 0.00417 ± 0.00018 0.20118 ± 0.00256 2490 ± 128
Tetracaine 0.05980 ± 0.00172 0.68613 ± 0.02399 162 ± 18
Lidocaine 0.49070 ± 0.03022 0.98795 ± 0.00748 20 ± 1
Methyl Paraben 0.24636 ± 0.00397 1.5250 ± 0.0196 77 ± 2
Ethyl Paraben 0.03076 ± 0.00160 1.3238 ± 0.0345 560 ± 9
Propyl Paraben 0.00613 ± 0.00042 1.3213 ± 0.0811 2953 ± 61
Naproxen 0.02873 ± 0.00052 0.22046 ± 0.00867 355 ± 64
Indomethacin 0.01463 ± 0.00039 0.07328 ± 0.00128 474 ± 33
0
The solubility in water (Csat ), solubility in 10 wt% Pluronic® F127 (Csat ), and the apparent micelle–water partition coefficient at 25 ◦ C (Kmw ), of each pharmaceutical,
as obtained from ultraviolet spectroscopy (UV) experiments.
56 P.K. Sharma et al. / Colloids and Surfaces B: Biointerfaces 61 (2008) 53–60

which self-assembly is promoted. However, from Table 2, it is cellar interactions in the system. The copolymer concentration
also clear that the presence of some of the other drugs seemed of 5 wt% Pluronic® F127 was used because it was high enough
to inhibit micellar assembly, as evidenced by an increase in the to give some indication of the interactions occurring between
measured CMC. Such systems seemed to affect the CMC to a micelles, without the SANS spectra being dominated by very
much greater degree (i.e., larger absolute changes in the CMC) strong peaks that would be characteristic of a cubic arrangement
in comparison to those that promoted micellization. Cosolvents in the system.
such as NaCl [9], propylene glycol, glycerol, and PEG 400 [10] We fitted the SANS data from 5 wt% F127 solutions using
have been shown to affect the CMC of Pluronic® F127 micelles; a modification of the spherical core–corona model as described
therefore, it is not surprising that drug molecules are also able by Goldmints et al. [13], given in the following equations:
to affect the CMC of Pluronic® F127 solutions.
(ρ)2 F (q)
However, there seems to be no correlation between the CMC
 2
and any of the drug characteristics considered here, such as the 4πR31 (ρ1 − ρ2 ) 3j1 (qR1 ) 4πR32 (ρ2 − ρS ) 3j1 (qR2 )
apparent micelle–water partition coefficient or the solubility (see = +
3 qR1 3 qR2
Table 1). It is likely that another property of the drugs is influ-
encing the CMC to a greater extent than either of these two. The (1)
values of the CMC given in Table 2 were used to fit the data from
the SANS experiments and calculate the partition coefficients (CF127 − CMC)NAV
N= (2)
from the UV data. Nagg
Here, (ρ)2 F(q) is the combined contrast and intraparticle form
3.3. Small-angle neutron scattering
factor; R1 and R2 the micelle core radius and micelle corona
radius, respectively; ρ1 , ρ2 , and ρS the scattering length densi-
Nearly all the scattering data obtained from 5 wt% F127
ties of the core, corona, and solvent, respectively; and j1 is the
solutions (Fig. 1) showed features characteristic of spherical
first order spherical Bessel function. The number density of scat-
micelles in ordered structures. Several studies have shown that
tering centers, N, is given by the F127 concentration, CF127 ; the
the micelles of F127 are spherical in nature at concentrations of
critical micelle concentration, CMC; Avogadro’s number, NAV ;
up to 20 wt% F127, and that non-spherical structures only exist
and the micelle aggregation number, Nagg . The parameters R1 ,
above temperatures of 65 ◦ C [11,12]. It is therefore clear that the
R2 , and Nagg were fit; others were calculated or obtained from
presence of the drugs we studied did not affect the shape of the
experiment.
F127 micelles. Instead, differences in the structure were mani-
It was also necessary to account for the interactions between
fested as changes in the aggregation number and micelle size.
micelles, owing to the strong correlation peak observed at low
For 5 wt% F127 in D2 O, a strong correlation peak was present at
q. The hard-sphere repulsion interaction described by Yang et
a q value of around 0.023 Å−1 . This peak corresponds to intermi-
al. [14] was used to model the structure factor, S(q), introducing
Rint , the intermicellar interaction distance, as another parameter
into the overall model. The volume fraction, φ, used in this model
was calculated using:
4πR3int N
φ= (3)
3
where N was obtained from Eq. (2).
Finally, since polydispersity causes “smearing” of the SANS
spectra, such effects were taken into account through a fifth
parameter, σ, the half-width of a Gaussian distribution of the
micelle size, R2 ; the spherical core–corona model was multiplied
by this distribution and integrated over all positive values of R2 .
In order to keep the number of fitted parameters reasonable, and
since there was not enough resolution in the scattering spectra to
distinguish between polydispersity in the micelle core size and
polydispersity in the overall micelle size, only that of the overall
micelle size was used. The scattering intensity was then given
by
Fig. 1. SANS spectra of 5 wt% Pluronic® solutions. Small-angle neutron scat-  ∞ −(r−R2 )2 /2σ 2
e
tering of 5 wt% Pluronic F127 micelles in the absence and presence of various I(q) = NS(q) √ (ρ)2 F (q)dr (4)
drugs at saturation. For clarity, items in the legend are shown in order of high- 0 σ 2π
est to lowest intensity from the main plot. Data between 0.01 and 1 Å−1 are
where the full equations for S(q) are given elsewhere [14].
shown due to the large uncertainties at very low and high values of q. Inset:
example data fits are shown for the samples that showed the highest and lowest Good fits to the data were obtained using the above model;
intensities—all other data fits fall between these two traces. for clarity, they are not shown in Fig. 1. Instead, two examples
 P.K. Sharma et al. / Colloids and Surfaces B: Biointerfaces 61 (2008) 53–60 57

Table 3
Structural parameters of Pluronic® micelles

System R1 (Å) R2 (Å) Rint (Å) σ (%) Nagg

Paclitaxel 43.45 ± 0.06 67.67 ± 0.10 108.3 ± 0.3 10.78 ± 0.03 79.5 ± 0.3
Baccatin III 41.00 ± 0.07 63.07 ± 0.12 91.3 ± 0.5 10.58 ± 0.04 61.5 ± 0.4
Sulindac 40.79 ± 0.07 62.76 ± 0.11 92.5 ± 0.4 10.44 ± 0.04 60.8 ± 0.3
Dibucaine 43.43 ± 0.06 67.39 ± 0.10 106.6 ± 0.4 11.06 ± 0.03 78.0 ± 0.3
Tetracaine 58.86 ± 0.04 72.16 ± 0.04 111.5 ± 0.1 8.35 ± 0.02 56.7 ± 0.1
Lidocaine 58.47 ± 0.03 71.31 ± 0.01 111.2 ± 0.2 8.02 ± 0.01 57.0 ± 0.1
Methyl paraben 64.43 ± 0.02 79.01 ± 0.03 119.5 ± 0.04 7.20 ± 0.05 71.4 ± 0.4
Ethyl paraben 65.33 ± 0.04 80.63 ± 0.04 121.2 ± 0.2 7.35 ± 0.02 74.5 ± 0.1
Propyl paraben 65.05 ± 0.04 79.95 ± 0.04 121.0 ± 0.1 7.46 ± 0.02 74.0 ± 0.1
5 wt% Pluronic® F127 42.85 ± 0.06 66.45 ± 0.11 101.3 ± 0.4 10.45 ± 0.03 73.4 ± 0.3

The micelle core size (R1 ), corona size (R2 ), intermicellar interaction distance (Rint ), polydispersity (σ), and the aggregation number (Nagg ), of F127 micelles in the
presence of each drug, as obtained from fitting the small-angle neutron scattering data; and the critical micelle concentration (CMC) of F127 solutions in the presence
of each drug, as measured from dynamic light scattering. Data for 5 wt% F127 in the absence of any drugs is also given.

are shown in the inset. The parameters obtained from the fits shown in the appropriate section of the diagram. Using a DOE
are shown in Table 3. The micelle core radius was found to be approach enables the individual effects of the drug partition coef-
42.8 Å for 5 wt% Pluronic® F127 in D2 O, which agrees well ficient and drug solubility to be seen more clearly, along with any
with a previously reported core radius of 44 Å [15]. The cor- coupled effects that may exist between them. This then makes
responding overall micelle radius obtained from the data was it easier to effectively correlate the effects that the drugs have
66.5 Å with a polydispersity of 10.4%. This micelle size is a on the various structural parameters. Note that by choosing the
little larger than the previously reported micelle radius of 57 Å solubility in a 10 wt% Pluronic F127 solution as one of the fac-
[2]. However, the present findings are reasonable, considering tors, rather than simply the solubility in water, the manner in
the magnitude of the polydispersity, and the fact that the micelle which a solute interacts with the copolymer is accounted for
radius observed from SANS is generally smaller than the hydro- to some extent. Thus, the magnitude of this physicochemical
dynamic radii observed from dynamic light scattering studies, parameter indicates the suitability of the solute with Pluronic®
11–15 nm [9,15,16]. The aggregation number that was obtained F127, and this information is retained in any correlations that
for Pluronic® F127 in water was approximately 73, which is are determined between this factor and a response of the system.
slightly higher than expected. Previous studies have reported From Fig. 3 we see that most drugs cause an increase in the
Nagg ranging from 37 [2] to 72 [9], with the variability in findings core size, corona size, and intermicellar interaction distance.
to be most likely due to the differences in experimental methods The exception to this occurs with drugs that have low solubil-
[16]. In general, the results of the SANS analysis agreed well ity and partitioning in F127. All three parameters depend very
with the findings of others, indicating that the model used to fit strongly on drug solubility, with a relatively weak dependence
the data was appropriate and reliable. on partition coefficient. This suggests that the total amount of
drug present affects the size and distance between F127 micelles
3.4. Design of experiments analysis more significantly than does the proportion present in the PPO
cores. However, both factors do have an effect as can be seen
Ideally, to determine the effects of different factors, one by the small increases along the Kmw axis. In order to correlate
would systematically vary one input while holding the others these effects accurately, terms containing both the drug solubil-
constant. However, when the input variables are physicochem- ity and the partition coefficient are required. Eqs. (5)–(7) show
ical properties, such as Kmw , molecular weight, or solubility, such correlations obtained for these three structural parameters,
this is not always possible. Thus, we have used a design of with coefficients of determination of 0.943, 0.956, and 0.902,
experiments (DOE) approach to analyze our results. Details of respectively.
this method of data analysis can be found elsewhere [17]. This
type of approach is becoming increasingly common in analyzing R1 = 17.6Csat + 0.000350Kmw + 40.9 (5)
experimental data when a large number of factors are involved,
R2 = 11.6Csat + 0.00117Kmw + 62.1 (6)
particular when different factors may have confounding effects.
We chose a three-level, two-factor design, with Kmw and the Rint = 17.9Csat + 0.00297Kmw + 93.1 (7)
drug solubility in 10 wt% F127 as the two factors. As shown by
Fig. 2, each drug falls into a separate region of the design space, The relative contributions to these structural parameters from
resulting in all nine of the regions being covered by our exper- the solubility and partition coefficient are clear from their coef-
iments. Figs. 3–5 show various DOE diagrams for the results ficients in these equations. It is not unreasonable to expect the
from both the SANS and DLS experiments. The five responses solubility of a drug to significantly affect the size and dis-
shown are core size (R1 ), corona size (R2 ), intermicellar inter- tance between F127 micelles. For a hydrophobic drug, the drug
action distance (Rint ), aggregation number (Nagg ), and critical molecules would mostly be located within both the micelle core
micelle concentration (CMC). The values for each response are and corona. Their presence could then effectively result in some
58 P.K. Sharma et al. / Colloids and Surfaces B: Biointerfaces 61 (2008) 53–60

Fig. 2. General design of experiments diagram. DOE diagram showing the molecular structure of each pharmaceutical solute and the two physicochemical factors
of apparent micelle–water partition coefficient at 25 ◦ C (Kmw ) and solubility in 10 wt% Pluronic® F127 (Csat ).

sort of swelling of the micelles. This increase in size, along with From Table 3, we see that most of the drugs cause a decrease
the resulting stronger interactions between adjacent micelles, in the polydispersity of F127 micelles, making them slightly
could also cause an increase in the intermicellar interaction dis- more uniform. As interactions between micelles increase in the
tance. An increase in this parameter has important implications presence of drug molecules, they may be forced to conform to a
on the gelation behavior of F127 solutions. Stronger repulsions more uniform size. However, the value of σ is also affected by
between micelles may result in a higher effective micellar vol- instrument resolution, which has not been accounted for in this
ume fraction, which has been previously discussed as being analysis. It is therefore unclear whether the changes observed in
one of the key factors of the gelation process [4]. It is there- this parameter are significant.
fore not unreasonable to expect solutions containing these drugs Of all the structural parameters, it is the aggregation num-
to become gels at lower temperatures and concentrations than ber that shows the most interesting and complex behavior with
neat F127 solutions, and to form slightly stronger gels. Overall, regards to drug solubility and partitioning. The DOE design
the changes in the micelle size and interaction distance could diagram shown in Fig. 4 shows that the effects of these two
have large macroscopic effects on the physical behavior of the factors on the aggregation number are much more complicated.
system. In general, it would appear that most drugs cause either no sig-
nificant change or a decrease in the aggregation number, with the

Fig. 3. DOE diagram of micellar sizes and interaction distances. DOE diagram
showing the micelle core size (R1 ), overall micelle size (R2 ), and intermicel- Fig. 4. DOE diagram of the aggregation number. DOE diagram showing the
lar interaction distance (Rint ), as a response to the two factors of apparent aggregation number (Nagg ), as a response to the two factors of apparent
micelle–water partition coefficient (Kmw ) and solubility in 10 wt% Pluronic® micelle–water partition coefficient (Kmw ) and solubility in 10 wt% Pluronic®
F127 (Csat ). F127 (Csat ).
 P.K. Sharma et al. / Colloids and Surfaces B: Biointerfaces 61 (2008) 53–60 59

partition coefficient (Kmw ). There are no obvious trends in how

the CMC varies with either of these factors; from a linear regres-
sion analysis, no simple correlations could be determined. This
suggests that either the critical micelle concentration is a much
more complicated function of the two factors, or it depends
more strongly upon another factor that has not been considered
here, such as the degree of ionization, extent of conjugation,
or presence/absence of specific functional groups on the solute
molecule. In addition the CMC could not be correlated with any
of the other measured responses of the system that are discussed
later (i.e., the micelle core and overall size, the intermicellar
interaction distance, the aggregation number, and the change
in liquid-to-gel transition temperature), suggesting that it does
Fig. 5. DOE diagram for the critical micelle concentration. DOE diagram show- not play an important role in determining them. Considering
ing the critical micelle concentration (CMC) as a response to the two factors the large variation in reported findings for the CMC, it is also
of apparent micelle–water partition coefficient (Kmw ) and solubility in 10 wt%
Pluronic® F127 (Csat ).
possible that in a few of the cases, the changes observed in the
presence of the solutes may not have been significant enough.
With correlations for the various structural properties of the
exception being drugs that are not very soluble and are highly
F127 micelles, it is possible to predict the structural features
partitioned into F127 micelles; such drugs cause relatively small
using only a few experiments. For example, UV can be used to
increases in the aggregation number. This general finding agrees
measure the solubility in F127 solutions, as well as the apparent
with previous work that has shown the aggregation number to
micelle–water partition coefficient in F127. These measure-
decrease considerably in the presence of the hydrophobic drugs
ments can then be used in Eqs. (5)–(8) to calculate the required
naproxen and indomethacin [4]. A decrease in the aggregation
structural properties, without the need to perform expensive
number leads to an increase in the number density of micelles
scattering experiments.
present, which affects gelation through the micellar volume frac-
tion [4]. From the DOE diagram, there is definitely a dependence
3.5. Implications of structural change
upon drug solubility, but not of a monotonic nature. There is
also a non-linear dependence upon the partition coefficient, as
Changes in the structural properties of F127 micelles have
can be seen from the aggregation number at low drug solubility.
important implications on formulations containing drugs or
With the use of a logarithmic transformation for the partition
additives. For example, an increase in the micelle core size may
coefficient, a linear relationship with the aggregation number is
result in a larger solubilizing capacity of F127 micelles [1]. In
obtained. A good correlation must therefore take these two non-
particular, a sufficiently large micelle core size and high aggre-
linear effects into account. Eq. (8) shows such a relationship,
gation number is required for efficient solubilization [1]. Since
with a coefficient of determination of 0.803.
addition of the alkyl-4-hydroxybenzoate molecules caused the
Nagg = 8.41 Log(Kmw ) + 15.5Csat
2
− 18.4Csat + 46.4 (8) largest increases in the micelle core size without significantly
affecting the aggregation number, it is possible that drugs with
The relative importance of the partition coefficient on the aggre- poor solubility such as paclitaxel could be solubilized to a
gation number is much greater than for the other structural much greater extent by simply adding these compounds to the
parameters; however, the solubility also has a more compli- formulation. It could be argued that most of the drugs them-
cated relationship in this particular case. It is clear that through selves promote their own ‘self-solubilization’ by increasing the
steric factors, the presence of a drug solute could potentially micelle core size considerably, while maintaining a high aggre-
cause a decrease in the aggregation number. The initial effect gation number. Hence, in some cases, additives may not even
of increasing the drug solubility would then be to amplify be required to achieve the concentrations required for clinical
this decrease in the aggregation number. However, as solubility applications.
becomes high, the larger presence of drug molecules may result As previously mentioned, the increase in the intermicellar
in stronger drug–polymer interactions that cause the aggrega- interaction distance observed in the presence of most of the drugs
tion number to begin increasing again. This would explain the indicates that at formulation concentrations of F127, solutions
minimum observed with regards to the drug solubility. Drug may form gels at lower temperatures. This has significant impli-
molecules partitioning strongly into F127 micelles could result cations on the clinical use of F127 formulations, since it means
in the promotion of aggregation, through strong drug–polymer that lower concentrations of F127 could be used to achieve the
interactions occurring in the core, once again offsetting the gen- same gelation behavior. Although F127 is reported to be the least
eral decrease in the aggregation number. This could explain the toxic of the Pluronic block copolymers [5], safety is always an
contribution from the partition coefficient that is seen in Eq. (8). issue with any pharmaceutical product. A lowering of the amount
Fig. 5 shows the DOE diagram obtained when the CMC is of F127 required in pharmaceutical formulations could result in
taken to be the response of the system to the two factors: solu- both reduced toxicity concerns from the FDA and lower produc-
bility in 10 wt% Pluronic® F127 (Csat ), and the micelle–water tion costs for manufacturers. As a result of this, the use of F127
60 P.K. Sharma et al. / Colloids and Surfaces B: Biointerfaces 61 (2008) 53–60

formulations for hydrophobic drug delivery applications could the research investment required by pharmaceutical companies
be promoted significantly in the pharmaceutical industry. for candidates of controlled delivery with F127.
By far, the most useful benefit of understanding the change in
the micellar structure is the ability to predict the release behav- Acknowledgements
ior of a potential formulation. This is particularly important for
hydrophobic drugs that cannot be delivered by a conventional Results shown in this report are derived from work per-
oral route and are either expensive to produce, unavailable in formed at Argonne National Laboratory. Argonne is operated by
large quantities, or simply a candidate drug molecule. From the UChicago Argonne, LLC, for the U.S. Department of Energy
solubility and partition coefficient of a drug (obtained from UV under contract DE-AC02-06CH11357. This material is based
experiments) and measurement of the CMC, our correlations can upon work partially supported by the National Science Founda-
be used to calculate the necessary parameters for the initial con- tion under Grant No. DMI-0531171.
centration in the micellar core. This concentration in the micelle
core sets up the initial concentration gradient for drug release References
and hence the release rate itself. The release profile can then be
predicted by using an appropriate in vitro transport model for [1] M. Malmsten, Surfactants and Polymers in Drug Delivery; Drugs and
drug release from the F127 micelles in the gel state. This will be Pharmaceutical Sciences Series, vol. 122, Marcel Dekker, New York, NY,
2002.
the subject of a future publication. Through these types of pre- [2] G. Wanka, H. Hoffmann, W. Ulbricht, Macromolecules 27 (1994) 4145.
dictions (a fairly straightforward task involving relatively few [3] S. Miyazaki, C. Yokouchi, T. Nakamura, N. Hashiguchi, W.M. Hou, M.
experiments), the suitability and performance of a formulation Takada, Chem. Pharm. Bull. 34 (1986) 1801.
can be assessed with a considerably lower investment in time [4] P.K. Sharma, S.R. Bhatia, Int. J. Pharm. 278 (2004) 361.
and money. [5] I.R. Schmolka, J. Biomed. Mater. Res. 6 (1972) 571.
[6] Y. Kaminoh, T. Inoue, S.M. Ma, I. Ueda, S.H. Lin, Biochem. Biophys. Acta
946 (1988) 337.
4. Conclusions [7] G.L. Patrick, An Introduction to Medicinal Chemistry, Oxford University
Press, New York, NY, 1997.
Changes in the structural parameters of Pluronic F127 [8] P. Alexandridis, J.F. Holzworth, T.A. Hatton, Macromolecules 27 (1994)
micelles caused by the presence of various hydrophobic drugs 2414.
[9] P.R. Desai, N.J. Jain, R.K. Sharma, P. Bahadur, Colloids Surf. A 178 (2001)
and pharmaceutical additives were investigated. In general, the 57.
presence of most of the drugs was found to result in an observed [10] N.K. Pandit, H.J. Mcintyre, Pharm. Dev. Technol. 2 (1997) 181.
increase in the size of the micelle cores and micelle coronas. [11] K. Mortensen, Y. Talmon, Macromolecules 28 (1995) 8829.
The intermicellar interaction distance was also found to increase [12] Y.M. Lam, N. Grigorieff, G. Goldbeck-Wood, Phys. Chem. Chem. Phys. 1
accordingly, however, the aggregation number was found to (1999) 3331.
[13] I. Goldmints, G.E. Yu, C. Booth, K.A. Smith, T.A. Hatton, Langmuir 15
decrease in the presence of most of the drugs. Using a DOE (1999) 1651.
approach, we were able to obtain correlations to describe the [14] L. Yang, P. Alexandridis, D.C. Steytler, K.M. J, H.J. F, Langmuir 16 (2000)
structural parameters of the micelles from the drug solubility and 8555.
apparent micelle–water partition coefficient. Our results imply [15] R.K. Prud’homme, G.W. Wu, D.K. Schneider, Langmuir 12 (1996)
that the solubilization of a drug would be increased merely by 4651.
[16] G.E. Yu, Y.L. Deng, S. Dalton, Q.G. Wang, D. Attwood, C. Price, C. Booth,
its presence, and that formulations may require lower concentra- J. Chem. Soc. Faraday Trans. 88 (1992) 2537.
tions of F127 to achieve the same gelation behavior. In addition, [17] G.E.P. Box, W.G. Hunter, J.S. Hunter, Statistics for Experiments: An Intro-
the structural parameters can be used with a measured CMC duction to Design, Data Analysis, and Model Building, John Wiley & Sons,
value to calculate a predicted release profile, potentially reducing New York, NY, 1978.