Magnetic Resonance - Based Evaluation of Small Molecule ... · Magnetic Resonance-based Evaluation of Small Molecule Release from a Thermosensitive Drug Delivery System Amanda Aleong
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Magnetic Resonance - Based Evaluation of Small Molecule Release from a Thermosensitive Drug Delivery
System
by
Amanda Aleong
A thesis submitted in conformity with the requirements for the degree of Master of Applied Science
Institute of Biomaterials and Biomedical Engineering University of Toronto
sections through the agar phantom: The catheter used for remote injection can be seen, as indicated, as
well as the tip of the temperature probe in the gel medium, adjacent to the spherical void.
TSL with a Gd concentration of 1.64 mg/mL and size of 94.3 nm were used for all 1.5%
agar studies. The concentration for free Gad was matched by diluting ProHance® and NTSL with
HBS. NTSL was produced in house as described previously by Zheng et al. with a batch
concentration of 4.24 mg/ml Gd and a mean diameter of 94.0 nm [21]. The similarity in size
allowed NTSL to serve as a suitable negative control. NTSL was diluted in HBS to match the
concentration of Gd in TSL.
24
Steady Temperature Profiles were Maintained During Iimaging
Hyperthermia was achieved using an in-bore recirculating water chamber connected to an
external water bath. Temperature was maintained within 1°C above the target temperature for each
scan. A sample of the profile obtained at each temperature for the duration of the 30-minute scan
is shown in Figure 6. The dip in temperature following injection under hyperthermia may be
attributed to heat transfer from the gel medium to the room temperature agent. This effect is less
evident at body temperature due to the smaller difference in temperature between the agent and
the gel medium.
90° Flip Angle Provided Highest Signal to Baseline Contrast
To achieve high image contrast between the agent and the gel medium, a trade-off was
made in the selection of imaging parameters. Figure 7 shows the signal intensity for agar, free
Gad, NTSL and TSL at 22°C using a series of flip angles between 10 and 90°. Signal intensity was
Figure 6 Representative temperature profiles for the duration of imaging measured via an optical
fibre probe: Temperature for three independent runs is shown each at room temperature (blue), body
temperature (gray) at under hyperthermia (red) with the target temperature indicated in brackets in the
legend.
25
found to increase for free Gad over the range of angle tested. On the other hand, both NTSL and
TSL exhibited a peak in signal intensity below the maximum flip angle. A flip angle of 90°
provided the largest signal contrast relative to baseline intensity in the gel medium for a Gd
concentration of 1.64mg/ml. As such, this flip angle was selected for the DCE-GRE imaging
sequence.
Figure 8 Axial images showing the cross section through the spherical void in the agar phantom
before, immediately after agent injection, and at the end of the imaging period for Free Gad at 22°C: The agar phantom is outlined in yellow and the interface between the agent and the gel at the 30min time
point is shown in red. At t = 1min, no contrast agent has entered the gel while at t = 30min contrast
enhancement is observed as a diffusive aura permeating the gel medium.
Figure 7 Signal intensity vs. flip angle: Signal for air is shown in black, the baseline 1.5% agar gel
in gray, free Gad in blue, NTSL in green and TSL in red. Signal increases with flip angle for free
Gad but shows a peak for liposomal agent. Signal from agar decreases with increasing flip angle and
remains constant for air.
26
Figure 9 Images obtained at t = 30min for free Gad (left panel), non-thermosenstive liposomes
(NTSL; center panel) and thermosensitive liposomes (TSL; right panel) at 22°C, 37°C and 43°C: The
black dotted line in each image indicates the contact surface between the contrast pool and the gel medium
as identified immediately after injection of the agent. Left and bottom scale is shown in pixels and the color
bar indicates raw MR signal intensity in arbitrary units.
Small Molecules Diffuse Visibly Through Agar While Stable Nanoparticles Do Not
Figure 8 shows representative images for free Gad at 22°C at before injection (t = 0 min),
immediately after injection (t = 1 min) and at the end of the imaging period (t = 30 min).
Immediately after injection, a contrast pool formed in the lower half of the spherical void and a
distinct line was seen where the injected contrast agent came into contact with the gel medium.
Over the time-course of imaging, signal enhancement can be seen beyond this contact surface as
contrast agent molecules diffuse into the gel medium. Comparison of the final imaging time frame
(t = 30 min), for each case investigated, reveals a visible difference in the behavior of nanoparticles
and free small molecules (Figure 9). No discernable signal enhancement in the gel was observed
for NTSL at any temperature indicating no diffusion of the intact liposomes away from the
27
injection site. Thus, the signal enhancement measured for TSL at 43°C was confirmed to be the
diffusion of the small molecule Gad into the gel following release from the encapsulating TSL.
Free Gad showed a visible diffusive spread at all three temperatures, comparable to the diffusive
spread observed for TSL at 43°C.
As expected, the signal at the injection site increases with temperature for both NTSL and
TSL. Despite the increased signal at the injection site for NTSL, no diffusion was observed in the
gel medium. This confirmed that the NTSL remained stable under hyperthermia, preventing the
release of their small molecule contrast agent load. These results demonstrated the ability of the
designed phantom to distinguish between MR-based contrast enhancement due to increased proton
exchange across the encapsulating membrane and contrast enhancement due to release of the
contrast agent molecules.
3.2 Aim 2: Quantification of Drug Release in MATLAB
MR Signal Decreases with Increasing Temperature for a Given Gd Concentration
Upon quantification of signal vs. time curves, it was evident that temperature effects were
not negligible. To compare T1-weighted intensity data acquired at various temperatures, it was
necessary to adjust signal data against changes in temperature. The relationship between signal
intensity and temperature at a given Gd concentration is non-linear. As such, signal at a given
temperature can be mapped to its corresponding value at room temperature for a specific contrast
concentration through a simple ratio. Figure 10a shows that the signal vs. concentration for each
temperature can be approximated as linear for up to 0.4mg/ml [Gd] which corresponded to a signal
increase of approximately 600 – 800 AU (arbitrary units) using the same imaging parameters and
conditions as those used in diffusion studies. By fitting the data at each temperature with linear
regression, the slopes of the graphs were obtained and substituted into Equation 6 to map the data
acquired at body temperature and hyperthermia to the scale at room temperature. The result of this
mapping on the calibration data is shown in Figure 10b.
In addition to accounting for temperature effects, it is also important to account for image
to image fluctuation. MR signal intensity is typically measured in arbitrary units where the value
28
corresponding to a specific concentration of contrast agent may vary with respect to time, system
temperature as well as other external factors. The complexity of this noise dependence is well
documented and is the reason that MR remains semi-quantitative in most practical clinical
situations [74], [75]. To reduce this effect image voxels are adjusted to the maximum signal
intensity for a given test case.
Figure 10 Change in signal intensity versus gadolinium concentration before (a) and after adjustment
using Equation 7 (b): Points show signal intensity measured with the T1-weighted DCE-GRE sequence
for agar (solid symbols) and 7.5% BactoTM Agar (open symbols) at 22°C (blue), 37°C (green) and 43°C
(red). Without adjustment, higher temperatures lead to lower signal for each Gd concentration.
Shape of Signal vs. Time Curves Provide Quantitative Evidence of Diffusion
Segmentation of the gel medium into contours radiating outwards from the injection site
resulted in curves with high reproducibility between gel phantoms. Figure 11 shows the signal
versus time curves at select distances from the contrast pool-gel boundary. It should be noted that
while there is an overall loss in signal in the contrast pool (not shown), the signal at the boundary
remains constant for the imaging duration. Progressing through contours at increasing distances,
there is a trend of increasing delay to signal accumulation that is characteristic of diffusion.
Furthermore, contours that are further away from the contrast pool-gel boundary accumulate signal
at a slower rate. This is a consequence of the decreasing concentration gradient across each contour
with increasing distance. The signal vs. time curve at 0.7 mm was selected for further
characterization as it provided the highest signal while ensuring minimal convective contributions.
29
In addition, curves that were far enough from the boundary exhibited linear behavior, allowing
approximation of the diffusion coefficient as a ratio of the linear slope to the spatial Laplacian of
the corresponding voxels.
Signal Profiles Confirm Stability of TSL Up to 37°C and Complete Destabilization at 43°C
Adjusted signal intensity vs. time curves at a distance of 0.7 mm from the injection site
provided quantitative evidence that stable nanoparticles do not diffuse into the gel medium within
the imaging period (Figure 13). NTSL at 22°C, 37°C and 43°C showed no significant signal
accumulation at this distance. On the other hand, free Gad, used as a positive control representing
100% release, showed significant signal enhancement at all three temperatures. TSL at 22°C and
37°C show only a slight increase in signal, comparable to the behavior of NTSL. The signal vs.
time curve for TSL at 43°C matches closely with the free Gad curve at 43°C. This suggests
complete destabilization of the liposomal membrane and free diffusion of the small molecule
contrast agent following release.
Characterization of key curve features revealed increasing rate constants with respect to
temperature. This was expected as diffusion relies on the kinetic energy of the diffusing particle.
Increasing kinetic energy increases the frequency of collisions and therefore the overall net transfer
of molecules. This is similarly reflected in the decreasing time taken for contrast molecules to
Figure 11 Mean signal intensity vs time with increasing distance from the edge of the spherical
void for free Gad in agar at 22°C.
30
diffuse to the selected distance of 0.7 mm from the contrast pool-gel boundary. Slower diffusion
experienced by molecules at a lower temperature results in longer time delays to detectable signal
accumulation. Both K and t0 exhibit statistically significant dependence with respect to temperature
with p-values less than 0.0001. A Student’s t-test comparing data sets in agar to the positive control
free Gad at 43°C revealed that the time delay for TSL at 43°C was statistically different from that
of free Gad at 43°C (p < 0.001). These results suggest that there is a time delay associated with
release of small molecules following destabilization of the TSL. In spite of this delay complete
release of Gad was achieved as shown by the similarity in rate constants for free Gad and TSL at
43°C.
Diffusion Coefficient of Gad Released from TSL is Equivalent to That of Free Gad in Agar at 43°C
Small molecule kinetics quantified as diffusion coefficients confirmed complete
destabilization of the heated TSL liposomes Figure 16. In fact, the diffusion coefficient measured
for TSL at 43°C was (2.90 ± 0.52) ×10-4 mm2/s, which is not statistically different from that
measured for the free Gad in the gel phantom (2.72 ± 0.87) ×10-4 mm2/s. As expected, diffusion
coefficients measured for NTSL were two orders of magnitude lower. However, the accuracy of
this measurement was limited by the low signal accumulation (less than 50 AU), which could not
reliably be distinguished from image noise. Diffusion coefficients measured for TSL at 22°C and
37°C are statistically equivalent to NTSL, supporting the stability of the liposomes at these
temperatures, but are similarly affected by limited signal.
31
3.3 Aim 3: Evaluation of the Physiological Relevance of the Gel Phantom
Ex Vivo Tissue Specimen Fixed in Agar Enabled Evaluation of Small Molecule Diffusion in Tissue
Figure 12 shows representative scout images of chicken muscle and tumor tissue (SKOV-
3) fixed in 1.5% agar, adjacent to a spherical void. From the scout images, a film of agar can be
seen separating the tissue from the injection site. This is more apparent for tumor tissue than for
muscle, where the uneven surface of the tumor disproves the assumption of isotropic diffusion
from the injection site. Time lapse images over the course of the scan period highlight the
inhomogeneity of diffusion in the tumor tissue (example provided in Appendix A). To overcome
this limitation, an alternative segmentation algorithm was developed as described in Section 2.3.2.
Figure 12 Sample scout images for muscle (top panel) and SKOV-3 tumor (lower panel): Tissue can
be clearly identified from the support material, agar.
Inclusion of a spherical void ensured a constant concentration at the boundary of the tissue.
Furthermore, the spherical void served an injection site rather than direct injection into the tissue
which would result in variable distribution of the contrast agent and potential damage to the tissue
(see Appendix B).
32
Figure 13 Adjusted signal vs. time at 0.7mm from the edge of the spherical void for runs performed in agar (a) and muscle (b): Free Gad are is
indicated in blue, NTSL in green and TSL in red. Open symbols denote runs performed at 22°C, half-filled denote 37°C and filled denote 43 °C. NTSL
at all temperatures as well as TSL at 22 and 37°C show minimal signal enhancement while TSL at 43°C matches closely to free small molecule Gad
under hyperthermia. Mean and standard deviation across runs are plotted for each time point with n=3 for all cases.
33
Signal vs. Time Curves Reveal Slower Diffusion in Muscle Compared to 1.5% Agar
Signal vs. time curves measured in ex vivo chicken muscle showed slower diffusion of free
small molecules (Gd = 1.64 mg/ml; n = 3 at all three temperatures) than in agar. Figure 13 shows
the mean signal along contours at 0.7 mm from the contrast pool gel-boundary for each test case.
Signal curves did not plateau during the 30-minute scan period, unlike in 1.5% agar. In addition,
the higher variability between the muscle samples led to slightly larger standard deviations for
triplicate data sets. Despite this, the signal versus time curves were highly reproducible owing to
the homogeneity of the muscle tissue used for this study. Signal data for muscle was adjusted based
on calibration curves obtained in a diffusivity matched 7.5% BactoTM Agar tissue mimic.
Following signal adjustment, dependence on temperature remains apparent for free small molecule
diffusion. Due to batch to batch variabilities, the TSL agent administered to muscle was at a lower
concentration of 1.3mg/ml. As a result, there was a notable difference in diffusion for TSL at 43°C
in muscle compared to the free small molecule control.
Figure 14 Rate constant and time delay for agar and muscle as determined by curve fitting with
Equation 7: Rate constant (left) and time delay (right) are shown for cases where a good fit with R2 >0.9
was achieved: Free Gad (blue) and TSL at 43°C (red). × indicates a significance of p < 0.0001 for muscle
to the corresponding agar group, * indicates p < 0.0001 significance of a group relative to free Gad at 43°C
in the same medium as determined by independent t-tests.
Characterization of curve features using Equation 7 allowed quantitative comparison
between data sets. As with agar, increasing temperature decreases the time taken for Gad molecules
to accumulate at 0.7 mm in the tissue. Similarly, the rate of signal accumulation increases with
34
temperature in muscle, supporting the temperature dependence of diffusion through agar as a tissue
mimicking property. The values derived from curve fitting of TSL at 43°C were found to be
significantly different from those of small molecules at 43°C based on a Student’s t-test performed
on the fitted parameters in muscle. Furthermore, the difference was larger than that observed in
agar. Follow-up tests were performed, to better explain the difference in the case of TSL at 43°C.
Figure 15 Signal vs. time for two concentrations of Gd at 43°C: 1.64 mg/ml Gd as a reference point for
other studies (blue solid circles) and 1.30 mg/ml Gd (blue dotted circles). TSL at 1.3mg/ml injected into a
phantom heated to 43°C is denoted by red dotted squares while TSL pre-heated to 43°C and subsequently
injected into a phantom maintained at 43°C is denoted by black dotted squares. Mean and standard deviation
across runs are plotted for each time point with n=3 for all cases.
Signal Profiles Reveal Partial Release from TSL in Muscle at 43°C
Signal vs. time profiles for TSL at 43°C in muscle was lower than that observed for small
molecules. Subsequent analysis of TSL pre-heated to 43°C prior to injection in the spherical void
supported the hypothesis of partial release. Pre-heating of TSL for 20 min, prior to injection, was
assumed to achieve and maintain 100% release from the liposomes. As expected, pre-heated TSL
mimicked the free small molecule diffusion at the corresponding injected concentration in muscle.
Evaluation of small molecule diffusion at a lower injected concentration (1.3 mg/ml) revealed that
signal profiles were heavily dependent on the concentration at the boundary of diffusion.
Comparison of signal vs. time profiles demonstrated lower plateau values in both agar and muscle
35
tissue for lower injected concentrations. The sensitivity of the signal profiles to differences in
concentration provided a means to quantify partial release from TSL.
Figure 16 Diffusion coefficients calculated in agar (shaded) and muscle (solid) for free small
molecules (blue), NTSL (green) and TSL (red): × indicates p < 0.0001 for muscle compared to the
corresponding agar group. * indicates a significance of p < 0.0001 for TSL compared to free Gad at the
same temperature and + indicates p < 0.0001 for TSL against NTSL at the same temperature.
Diffusion Coefficients in Agar Mimic Trends Observed in Muscle
Calculation of diffusion coefficients for free Gad, NTSL and TSL in muscle allowed
further quantitative comparison showing slower small molecule diffusion in muscle compared to
agar. The temperature dependence observed in agar is also seen for muscle. NTSL at all three
temperatures and TSL at 22°C and 37°C did not diffuse through muscle within the imaging time
frame, again limiting the accuracy with which the diffusion coefficient can be estimated for these
groups. In these cases, no significant difference was found between values estimated in muscle
and those in agar when a multiple comparisons test was performed comparing data in muscle to
the corresponding group in agar. At 43°C, however, the diffusion coefficient of (1.06 ± 0.18) ×10-
4 mm2/s for TSL was significantly to different that calculated for both NTSL, (4.61 ± 3.50) ×10-6
mm2/s (p<0.0001) and free small molecules in muscle, (1.54 ± 0.09) ×10-4 mm2/s (p=0.007) based
on independent t-tests.
36
Diffusion Through Tumor Tissue is Slower than in Muscle and Agar
The result of the k-means clustering algorithm is illustrated in Figure 17, where regions
are color-coded based on the shape of the signal vs. time curve. The signal vs. time curves obtained
in SKOV-3 could not be directly compared to the signal curves obtained for agar and muscle due
to the difference in segmentation algorithm. Applying the contouring algorithm to tumor data
resulted in low signal which may be attributed to (1) the heterogeneity of the tissue along the line
segment and (2) the roughness of the tissue surface, both leading to an inhomogeneous diffusion
pattern (data not shown). Instead, k-means clustering delineated regions with similar contrast
uptake patterns. Inspection of the signal vs. time profiles associated with each region revealed a
similarity to that observed in the homogeneous cases of agar and muscle. Through this method, a
region with an approximately linear signal vs. time trend may be used to estimate a diffusion
coefficient for the tissue. An example of such a curve is shown in Figure 17, indicated by the black
arrow.
Figure 17 Showing clusters generated by K-means segmentation in an ROI indicated by the white
dotted line and the corresponding signal vs. time curves: Clusters are color coded to correspond with
the mean signal vs. time graph on the right. Scale is shown in pixels and color bar indicates voxels assigned
to the same cluster.
37
Small Molecule Diffusion in 1.5% Agar is Significantly Higher than that in Tissue
Small molecule diffusion coefficients calculated for 1.5% agar were consistently higher
than that found in muscle and SKOV-3 tumors for all temperatures tested (Figure 18). The
estimated diffusion coefficient for MDA-MB-231 tumors at room temperature is also shown,
however due to limitations in the signal observed over the imaging period, further tests were not
performed at higher temperatures. The trend of increasing diffusion coefficient with increasing
temperature that was observed in agar was maintained for muscle but not for tumor tissue. In
addition, the diffusion coefficient of (7.13 ± 0.77) ×10-5 mm2/s found for 7.5% BactoTM Agar at
22°C was statistically equivalent to the diffusion coefficient of (5.96 ± 0.73) ×10-5 mm2/s measured
in muscle at 22°C and that measured in SKOV-3, (3.62 ± 0.40) ×10-5 mm2/s. While this
equivalence was similarly maintained at both 37°C and 43°C for muscle, diffusion in 7.5%
BactoTM Agar was found to be significantly higher than in SKOV-3 tumor tissue at 37°C and 43°C,
with p-values of 0.02 and 0.003 respectively determined by a multiple comparisons test comparing
pairs values within each subgroup.
Figure 18 Diffusion coefficients at 22°C, 37°C and 43°C for 1.5% Agar, 7.5% BactoTM Agar, Muscle
and SKOV-3. × denotes significance to all other data sets with p < 0.05. * indicates significance (p < 0.05)
between two data sets as shown.
38
BactoTM Agar 7.5% Provides Diffusion Coefficients Closer to the Physiological Range
To compare signal curves obtained at different temperatures in muscle it was necessary to
confirm linearity of Gd-induced signal. Calculation of diffusion coefficients in various gel media
revealed that 7.5% BactoTM Agar provided diffusion coefficients that were not significantly
different from muscle tissue (results obtained at higher concentrations provided in Appendix D).
Based on the similarity in small molecule diffusion coefficient, 7.5% BactoTM Agar was used as a
tissue mimic to assess signal linearity of Gd relaxivity and generate calibration curves (Figure 10).
Serial dilution of Gad in 7.5% BactoTM Agar showed that linearity of signal intensity was
maintained within the range of concentrations employed for calculation of diffusion coefficients.
Results confirmed the suitability of the agar phantom for assessment of temperature-
dependent small molecule kinetics in MR. The current approach enables MR-based quantification
of small molecule diffusion. This method may be used in conjunction with complementary
techniques such as MR thermometry to achieve comprehensive spatio-temporal in situ evaluation
of non-invasive heat-induced (e.g. HIFU, RF) drug release from thermosensitive carriers. In
summary, the MR-based quantification platform described here enabled effective determination
of macromolecule retention and small molecule diffusion in agar gel phantom and ex vivo
biological tissues at physiological and hyperthermia temperatures. The next chapter will discuss
the benefits and limitations of this platform in relation to current practices in both preclinical and
clinical settings.
39
Chapter 4
Discussion
4.1 Aim 1: Visualization of Drug Release in a Gel Phantom
Tumor type, heterogeneity and vascularity, timing and flow rate of injection, timing and
duration of hyperthermia are just a few examples of the factors that affect the chemotherapeutic
dose delivered to the target site. While it is not possible to directly control the tumor type, several
strategies exist to control the other factors. In fact, it is now also possible to stratify patients for
tumor types that are more likely to respond. Point-based heat applicators add an extra layer of
complexity due to the need to control the heat deposition to ensure conformal heat delivery and
subsequent drug release. Imageable drug delivery systems have been used to answer many of these
questions at the in vivo stage of assessment. However, it is well-established that the optimal
parameters for preclinical studies do not translate directly to the clinic. While, in some cases, it
may be possible to apply adjustment factors to provide an educated guess at the optimal parameters
for patients this is far from ideal. FDA approval of drug delivery systems incorporating contrast
agents is a long and costly process. In the interim, pre-treatment assessment platforms that utilize
non-invasive imaging systems offer a means to determine optimal working parameters for each
institution to meet quality of care standards.
The standard approach for assessing thermosensitive drug delivery systems in phantoms
employs the difference in relaxivity between encapsulated and released small molecule imaging
agents [37], [47]. Typically, this involves doping a low gelling temperature hydrogel such as agar
with the liposomal formulation under investigation. The phantom is subsequently imaged before
and after hyperthermic heating and the corresponding T1 relaxivity maps are used to illustrate drug
release. While this method is sufficient to demonstrate that some form of small molecule release
has been achieved, it does not offer any insight into the percentage dose released or the spatial
response to the hyperthermia applied. It is possible that release is only achieved in the central
region of heating where the temperature is closest to the transition temperature of the liposomes
and the width of signal observed is actually due to subsequent diffusion of the small molecule from
that central region. This leads to the question of whether this matching of the heating zone with
the observed signal will be reflected in tissue or tumor specimen. In fact, our studies show that low
40
percentages of agar (1.5%) do not match the diffusion profiles of most tissues and that low
temperatures observed in the periphery are insufficient to result in drug release. We expect that the
true region of drug release will be smaller in tumor tissue based on the observation that diffusion
is far slower in tissue than in 1.5% agar. This may explain why ensuring uniform heat distribution
above the liposomal transition temperature is essential for positive patient outcomes. It is important
to heat the regions of the tumor with the highest vascularity and to deliver a high enough dose so
that the diffused drug is over the threshold needed for therapeutic efficacy.
Agar and agarose have served as a base phantom material for many applications
investigating hyperthermia and drug release [76]–[78]. Agar’s major advantage over alternative
phantom materials lies in the simplicity of preparation. Typical preparation times for a batch of
pure agar phantoms, in this study, falls under 30 minutes. Incorporation of a tissue specimen raises
that time to approximately 1 hour and 30 minutes to allow time for the agar to cool to temperatures
in the mild hyperthermia range, in order to guard against thermal damage to the tissue. In
comparison, polymer based tissue mimics can take on the order of hours to days to manufacture a
batch of phantoms and typically involves more expensive equipment and starting materials or by-
products that pose a greater risk to the user[56].
In our setup, a spherical void was incorporated in the body of the gel phantom to allow
injection of the agent during scanning. This is the first example of a phantom system for assessment
of hyperthermia-induced drug delivery that models an input function during imaging. Studies have
shown that temperature-sensitive drug delivery systems typically achieve the greatest efficacy
through intra-vascular release of the therapeutic [76], [77]. Pre-heating of the target region allows
drug to be released intravascularly, maintaining a positive and relatively constant concentration
gradient to the extracellular space. Previous phantom studies which mix the liposomal agent into
the body of the gel are unable to model the effect of pre-heating the tumor and therefore do not
provide a representative time response of release in such cases [18], [81]. In addition to the benefit
of observing the rapid release in response to pre-heating, the spherical void incorporated in our
phantom retained the stable nanoparticles over the imaging duration. As a result, we can
definitively distinguish released small molecules which are able to diffuse freely through the gel
and provide a visible contrast gradient in the gel.
41
A recirculating water chamber was incorporated to maintain a steady temperature
throughout imaging. Coupled with a real time temperature feedback, the use of this system enabled
manual temperature control. Temperatures were maintained within 1°C above the target value of
interest. In the case of hyperthermia, this meant that complete destabilization of the liposomal
membrane was expected for TSL. To heat to 43°C in a closed bore magnet, it was necessary to
insulate the recirculating water chamber to protect the RF coil from overheating. Exclusion of the
insulating layer led to poor image quality due to frequency shifts in the coil or triggering of fail-
safes which caused the scan to fail and all data for that imaging set to be lost. The recirculating
water-chamber is assumed to provide a spatially uniform temperature throughout the phantom.
The method used for heating in this study provides a means to characterize the temperature
response of MR signal intensity. Investigating drug release under these conditions provides the
groundwork for quantitative assessments of drug release in response to more complex spatio-
temporal heating patterns. Currently, MR-based strategies for simultaneously guiding
hyperthermia and monitoring of drug release are under investigation. To fully harness the power
of such a tool, understanding the kinetics of drug transport and release under hyperthermia and
optimizing the hyperthermia protocol to achieve the highest concentration of the drug continues to
be a primary concern within the field of temperature-sensitive drug delivery.
Imaging parameters were optimized to achieve the largest dynamic range for the diffusing
contrast agent. To achieve this, a larger flip angle of 90° was used. While it is unusual to use such
a large flip angle for dynamic T1-weighted imaging, it was necessary to minimize the signal
contribution from the gel medium. While the selection of the Ernst angle would have increased the
signal from the contrast agent, it would similarly have increased the signal from the gel, resulting
in a net lower dynamic range. As the selected temporal resolution (15s) was higher than the time
for a single scan (7s), the selection of a larger flip angle did not have adverse effects. The minimum
TE and TR were used for the desired field of view and the spatial resolution was minimized to aid
in discrete analysis of the diffusion coefficient. Further reducing the special resolution would lead
to increased image noise making it harder to calculate the diffusion coefficient from linear regions
where signal tented to be low (150 – 300 AU).
The phantom described in this thesis enabled visual confirmation of hyperthermia-induced
small molecule release from TSL, seen as a diffusive spread of signal enhancement in the gel
medium. Comparison to control cases confirmed that stable nanoparticles did not diffuse visibly
42
through the gel. While it is clear that release is achieved, based on the diffusive spread seen for
TSL at 43°C, the phantom alone does not provide insight into the extent of release achieved in
response to hyperthermia. Hence, a software package was developed to analyze spatio-temporal
signal data and alow quantitative comparison between TSL and control data sets.
4.2 Aim 2: Quantification of Thermosensitive Drug Release
It is a generally accepted fact that MR-based imaging techniques are, at best, semi-
quantitative. Inherent fluctuations in the measured parameters prevent exact estimation of the
number of contrast molecules in a single voxel or direct comparison of the number of released vs.
encapsulated molecules. Previous approaches use the mean over a number of voxels to observe a
rapid rise in signal intensity attributed to the release of the small molecule contrast agent. However,
several other factors may contribute to this net rise in signal. For example, increased blood flow
under hyperthermia, temperature effects on T1, and inherent inhomogeneity of baseline T1 values
all affect the measured signal intensity. Together, these factors limit the accuracy of estimation of
dose distribution in the tumor under hyperthermia conditions. As such, quantification defined in
this thesis refers not to MR’s ability to measure a specific concentration of contrast agent but rather
to enumerate the extent of small molecule release via molecular kinetics that are independent of
the imaging parameters. The diffusion coefficient was selected as a metric to assess the
physiological relevance of the agar phantom because imaging parameters do not directly affect it.
To assess hyperthermia-induced release from TSL using MR, it was necessary to maintain
a constant temperature throughout imaging. A major limitation associated with MR-based
chemodosimetry lies in the temperature sensitivity of the signal measurement [45]. As such,
previous studies opted to perform imaging before and after heating as a work-around [38]. This
study demonstrated that by maintaining a steady temperature during imaging, relevant scaling
factors can be applied during analysis to enable comparison of data sets at different temperatures.
This approach provides several advantages as it allows real time visualization of small molecule
release under hyperthermia conditions, rather than a final dose distribution. Furthermore, this
method simplifies analysis by directly assessing drug release based on temperature adjusted T1-
weighted signal intensity without the need for back-calculation of T1 or concentration maps.
43
As expected, increasing temperature led to increased relaxation times, reflected as
decreased signal. Linear regression of the signal vs. concentration data points resulted in an
adjusted R2 over 0.95 for all three temperatures, supporting this method as a sufficient
approximation for subsequent calculations. The slopes derived from the linear fit provided a means
to map data sets to a chosen dynamic range. To retain as much information as possible, the largest
dynamic range among the data set was selected, which was free Gad at 22°C in 1.5% agar. A
similar technique has been previously investigated by Collewett et al. to normalize image
intensities across data sets by assuming purely multiplicative changes between images [70]. In our
case, the multiplicative change arises from the difference in temperature of the specimen being
imaged. As seen in Figure 10b the mapping provides an excellent correlation with the target
dynamic range when applied to calibration curves at other temperatures.
NTSL encapsulating small molecule contrast agents have been well characterized over the
decades since their development. They have also played a significant role in the advancement of
thermosensitive drug delivery systems by providing a baseline for assessment of liposomal drug
delivery, in many cases acting as a negative control. In comparative studies between NTSL with
and without hyperthermia, increased extravasation has been observed for cases with hyperthermia.
Paramagnetic liposomes investigated by Fossheim et al. showed an increase in relaxivity with
temperature that was associated with increased fluidity of the liposomal membrane [82]. However,
this increase in relaxivity is considered negligible in comparison to the difference observed for
TSL under hyperthermia. While both NTSL and TSL show a moderate increase in relaxivity below
the transition temperature of the liposomal membrane, TSL is higher [48]. The greater permeability
of the thermosensitive lipid membrane allows higher exchange of water molecules across the
membrane surface and therefore greater relaxivity. This effect may be seen in the contrast pool
that remains in the spherical void for NTSL and TSL. This contrasts with the decreasing relaxivity
of the free contrast agent.
Typically, contrast enhancement in the clinic is assessed visually or as the mean/area under
the curve over a specified ROI. Studies reporting on the efficacy of drug delivery using
thermosensitive drug carriers use a wide variety of metrics to measure drug release that are often
not directly comparable. To expand the use of these drug delivery systems across institutions which
may use different instrumentation to achieve activated release, it may be beneficial to establish a
standard protocol via pre-treatment quality assurance. Phantom-based quantification of spatio-
44
temporal drug release provides a reproducible means to ensure that the platform is functioning as
expected and can help raise the bar to a new gold standard for activatable drug delivery.
Furthermore, with continuous improvements in the MR resolution in space and time there is a need
for better metrics which make use of this new information. To fully exploit spatial data, our
approach sections the region of interest i.e. the phantom or tissue into curved line segments that
are equidistant from the edge of the diffusive medium. The geometric symmetry of the phantom
ensures that the mean signal will have a lower error associated with it than using single voxels
along a radial path to represent signal versus time over distance (sample figure shown in Appendix
C).
Comparison of signal vs. time along a contour at a given distance from the injection site
provides a suitable means to quantitatively evaluate differences between imaging data sets. The
signal-time profile at a distance 0.7 mm from the gel-contrast boundary was selected as an optimal
distance for assessing whether or not small molecule release had been achieved. This distance was
far enough from the boundary to reduce the risk of measuring convective flow into the gel, as seen
by the delay in signal accumulation following injection at that distance. In addition, it was close
enough to provide sufficient signal to reliably observe trends between data sets. Using this strategy,
the small differences between free Gad diffusion as well as between TSL and NTSL can be seen,
which was not immediately apparent upon visual inspection. In particular, it was interesting to see
the similarity in TSL at 43°C and free Gad at 43°C which shows an almost perfect overlap in signal
over time in agar. The similarity in the signal curves supports an instantaneous burst release of
small molecules upon thermo-activation. Following release, the small molecules are able to diffuse
at a rate comparable to the positive control at that temperature. This is further confirmed by curve
fitting with Equation 7 which showed no significant difference between TSL at 43°C and free
Gad at 43°C for both time delay and rate constant in agar.
Although the comparison of signal data showed proved to be a viable means of quantifying
the extent of release, it was desirable to provide a metric which would not be dependent on the
units of the selected MR system. The diffusion coefficient was employed to provide a quantitative
measure of the efficacy of release. The suitability of the diffusion coefficient for this purpose relies
on (1) the difference in diffusion between small molecules and stable nanoparticles, (2) the
sensitivity to differences in concentration for dense media and (3) the intrinsic nature of the
45
parameter to the system under investigation. The reproducibility of the diffusion coefficients
measured in this study (n = 3 for all cases) further supports its use as a relevant metric.
As expected, due to the difference in molecular size, diffusivity of stable nanoparticles is
significantly lower than that of free small molecules [51], [52]. Both NTSL, which are not expected
to provide a burst release at any of the three tested temperatures, and TSL, which are expected to
be stable at 22°C and 37°C, demonstrate a diffusivity two orders of magnitude lower than that of
free small molecules. While diffusion coefficients of the 100nm nanoparticles are on the order of
those found in literature, these values cannot be reliably estimated within the time period used for
scanning [60]. The maximum signal accumulation observed is typically less than 3 standard
deviations from the image baseline and is therefore statistically indistinguishable from noise.
Importantly, a significant difference in diffusion coefficient of free small molecules was observed
with respect to temperature in agar. As such, it is necessary to account for the difference in small
molecule kinetics under hyperthermia when assessing the success of release from TSL under
hyperthermia. It is suspected that the diffusion coefficient of small molecules is a major limiting
factor in the transport of small molecules within the tumor environment when delivery is achieved
via thermosensitive drug delivery systems. Therefore, a means to quantify physiologically relevant
diffusion coefficients for each TSL batch would be highly beneficial.
4.3 Aim 3: Evaluation of Physiological Relevance of the Gel Phantom
Investigation of diffusion in the context of drug transport has very different meaning
depending on the size and chemistry of the diffusing molecule or particle. Traditional free drug
therapies exhibit a rapid wash-in and wash-out from the tumor environment, with concentrations
typically lower than the desired threshold for maximal efficacy. For nanoparticles, diffusion based
transport has been shown to dominate following extravasation at the tumor site. The size of the
nanoparticle plays a significant role in the diffusion through the extracellular space [83]. In the
case of TSL, the timing of activation can drastically affect the transport of drug through the tumor.
Manzoor et al. showed the importance of intravascular release on maintaining the concentration of
drug over time [80]. This confirmed a concentration gradient driven movement of drug molecules
through the extra-vascular space. Our approach allows diffusion of small molecules to be observed
46
through ex vivo muscle and tumor by fixing a section of tissue an agar based support material. The
tissue was placed adjacent to a spherical void, to enable diffusion analysis in tissue under
reproducible conditions. Direct injection into the tissue was also investigated but the uncontrolled
distribution and consequent dilution of contrast agent proved to complicate analysis extensively
(see Appendix B). By fixing the position of the tissue adjacent to a spherical void we can apply a
similar strategy for analysis of diffusion as that used for agar.
Compartmental models, such as two-compartment or Modified Toft’s models, are
commonly used to extract kinetic parameters that characterize transport in a physiological system
[84]. Often, this type of modeling requires a complete conversion from signal intensity to contrast
agent concentration which has yet to be achieved for high temporal resolution dynamic imaging.
Furthermore, introduction of hyperthermia, greatly complicates this calculation as the relaxivity
of the contrast agent decreases, the relaxivity of the paramagnetic carriers increases and
temperature dependent susceptibility effects are non-negligible [45]. In addition, such kinetic
parameters provide an overall measure across tumors and are insufficient to describe the
heterogeneity observed within the tumor. To simplify the analysis of diffusion in the controlled
phantom environment, we performed calculations directly on the signal intensity data generated
by the MR. The validity of this approach is based on the demonstrated fact that the signal varies
linearly for low concentration of contrast agents, and that this linearity holds under the
temperatures investigated herein. To compare signal intensities in agar versus tissue, it was
important to further account for the difference in proton density of the aqueous agar media and
tissue. This was achieved by evaluating signal vs. Gd concentration in 7.5% BactoTM Agar as a
tissue mimic which was selected based on similarity in free Gad diffusivity.
In this thesis, diffusion of small molecules was used to indicate successful release from
TSL. The temperature dependence of the diffusion coefficient then plays an important role in
evaluating small molecule kinetics following heat induced destabilization of the nanocarriers. In
aqueous media, the relationship is commonly assumed to be governed by the Stokes-Einstein
equation. In general, it is expected that diffusivity will increase with increasing temperature.
However, the extent to which this occurs in tissue has yet to be fully characterized due to
conglomeration with other effects such as increased blood flow. The trend of increasing
concentration with increasing temperature in agar showed a positive correlation to that seen in
muscle tissue.
47
After adjusting for MR-induced differences in signal, comparison of signal versus time
data obtained in 1.5% agar to those in muscle reveal overall higher small molecule diffusion in the
agar phantom. Muscle showed strong temperature dependence for small molecule diffusion which
supported the use of agar as a tissue mimic with temperature dependent small molecule diffusion
properties. As expected nanoparticles do not diffuse into the tissue within the duration of imaging
due to the larger size which is agreement with previous studies investigating nanoparticle transport
in the tumor [85]. Dreher et al. showed that it can take hours to days for nanoparticles to extravasate
a few millimeters depending on the size of the nanoparticle [83]. This may be true for the
nanoparticles as well but differences observed were not found to be significantly different.
The lower signal accumulation observed in muscle for TSL at 43°C compared to free Gad
at 43°C was likely due to the lower concentration of Gd in that batch of TSL. However, other
possible contributors included partial activation of TSL, excess lipid hindering diffusion or delayed
release. To further investigate the reason for the noticeable difference, a concentration matched
run was performed using free Gad at 1.3mg/ml [Gd]. These runs (TSL and free Gad; 43°C;
1.3mg/ml Gd) were also repeated in 1.5% agar to determine if the difference was specific to
muscle. Results showed that while TSL at 43°C again matched closely with the free Gad at the
same concentration, the TSL in muscle remained lower than the corresponding free Gad in muscle.
Subsequently, TSL was heated to 43°C prior to injection for 20 minutes to ensure that complete
release and diffusion through muscle was observed. This workflow resulted in a signal vs. time
profile that matched free Gad at 43°C in muscle at the concentration. Two conclusions can be
drawn from this: (1) the presence of lipids in solution, following destabilization of the liposomal
membrane did not hinder the diffusion of free small molecules and (2) the lower signal observed
in muscle was most likely due to partial release immediately after injection and therefore a lower
concentration of free small molecules.
A possible explanation for partial release lies in the fact that the temperature probe was
placed adjacent to the tissue sample at the edge of the spherical void rather than in the body of the
tissue. Placement of the probe in the tissue would have caused the tissue to separate from the
supporting gel medium and therefore was not feasible with this setup. Given the difference in
specific heat capacity between agar and tissue it is possible that the tissue was not at the precise
temperature needed to induce complete small molecule release from TSL. Ultimately, this was
similar to the problem suspected in the clinic where there is a difficulty achieving the target
48
temperature in all regions of the tumor. Overall, this emphasizes the need for spatial temperature
maps to guide hyperthermia regimens in the clinic, especially given that diffusion is a limiting
factor and it is desirable to maximize drug delivery to the target site.
A comparison of the diffusion coefficient obtained for small molecule Gad in 1.5% agar at
room temperature, (1.23 ± 0.08) ×10-4 mm2/s, to values found in literature shows a similarity to
the diffusion coefficient of free Gad observed in calf cartilage which has been quoted at (1.55 ±
0.22) ×10-4 mm2/s [57], [59]. Cartilage provides a relevant reference value for agar as it is primarily
a matrix of collagen fibers with no cellular barriers. In comparison, diffusivity of Gad in muscle
at room temperature, (5.96 ± 0.73) ×10-5 mm2/s, is significantly lower than that observed in
cartilage. This was in accordance with observations made by Djelveh et al. who showed that
diffusion across muscle fibers experience significant tortuosity effects leading to a Deff/D0 of 0.13
for glucose in bovine muscle while our study yielded 0.15 for Gad in galline muscle [58].
K-means clustering was employed as an alternative strategy for generation of signal versus
time curves in tumor tissue. By grouping voxels with similar uptake curves, lines of iso-
concentration in space and time may be identified (assuming one to one relationship between
signal and concentration). This approach was similarly employed by Koh et al. to segment tumor
tissue based on the shape of the contrast enhancement curve over time [55], [65]. In this work, the
number of clusters was increased until the program could not identify any new regions within the
gel and began segmenting the background noise. The final number of clusters used was seven in
total which showed a maximum of two overlapping signal curves at background. Clusters revealed
signal profiles with similar shapes to those seen using the contour method for muscle tissue. Voxels
in the tumor tissue were identified with similar profiles to that expected for diffusion. The jagged
nature of the clustered voxels more closely followed the surface of the tissue that can be identified
in scout images suggesting that it is a better estimate of signal uptake in tissue than that measured
using the contouring algorithm which instead relies on the edge of the contrast pool in contact with
the surface layer of gel. In addition, the drop signal drop-off with increasing distance was very
rapid, supporting the relatively low values measured for the diffusion coefficients.
Subsequent evaluation of the physiological relevance of the release kinetics against muscle
and tumor tissue showed that the 1.5% agar phantom overestimates the rate at which the small
molecules diffuse in the tissue environment. However, key features such as temperature
49
dependence of small molecule diffusion and the difference in diffusion coefficient between small
molecules and nanoparticles was preserved. As a result, a higher concentration of 7.5% w/v
BactoTM Agar was investigated which provided a diffusion coefficient closer to the physiological
range observed. A slight temperature dependence was observed as for muscle but a multiple
comparisons test of subgroups determined by temperature showed that the diffusion coefficient
measured was statistically higher than that observed for either tumor model. While this was
expected for MDA-MB-231 tumors, the diffusion coefficient for SKOV-3 was expected to be in a
similar range or even higher than that observed in muscle.
The apparent diffusion coefficient of protons measured using diffusion weighted imaging
has been investigated as a metric for staging cancer due to the observed differences during tumor
progression. The MDA-MB-231 murine breast cancer model used in this study exhibits one of the
lowest recorded ADC values at an average value of 0.49×10-3 mm2/s [86]. Based on the values
found in literature for apparent diffusion coefficient, it was expected that diffusion would show an
increase in the following order: MDA-MB-231, muscle, SKOV-3, with 1.5% agar showing the
highest diffusivity [86]–[89]. While diffusion coefficients in MDA-MB-231 was observed to be
the slowest, SKOV-3 showed no significant difference compared to muscle except at 43°C. Given
that the ADC ratio predicts a diffusion coefficient of approximately 8×10-5 mm2/s for SKOV-3
compared to 6×10-5 mm2/s in muscle, a higher number of samples may be needed to gain the
statistical power to distinguish such a small difference. This is further supported by the
heterogeneity observed for SKOV-3 diffusion, which compromises the accuracy of estimation in
the case of the tumor. In addition, subcutaneous tumors are typically characterized by lower
apparent diffusion coefficients than orthotopic models and as such may underestimate the small
molecule diffusion coefficient [57], [90]. The acquisition of proton density maps may provide
greater insight into the heterogeneity of the agar vs. muscle and tumor tissue. This will allow
modelling of the expected heat distribution and give a directional sense for expected ADC and
small molecule diffusion.
50
4.4 Summary and Future Directions
The platform presented in this thesis enables quantification of hyperthermia induced small
molecule release from temperature-sensitive drug delivery systems using MR. The platform has
been developed as a strategy for pre-treatment evaluation of drug release in response to non-
invasive heating systems such as RF and HIFU. Advances in the spatio-temporal capabilities of
MR has enabled real-time assessment of drug release via imaging of a small molecule drug
surrogate encapsulated in the delivery system. An agar phantom was developed which allowed
diffusion-based separation of released small molecules from stable nanoparticles. Diffusion of
small molecules, following destabilization of the liposomal carriers, was found to be consistent
with the diffusion of small molecules in the absence of liposomal carriers. Quantitative comparison
of signal versus time curves provided concrete evidence of a complete and immediate burst release
of the imaging drug surrogate at the liposomal transition temperature, resulting in profiles that
perfectly mimicked that of free small molecules under hyperthermia. Overall, the developed agar
phantom was shown to be a suitable strategy for detecting and quantifying thermosensitive release
kinetics in response to hyperthermia using MR. Assessment of the physiological relevance of the
phantom revealed that small molecule diffusion in muscle and tumor tissue was significantly
slower than in 1.5% agar. As an alternative, 7.5% agar maintained the benefits of the agar phantom
reproducibility while providing a medium with more tissue relevant diffusion rates.
This platform has the potential to be used in conjunction with complementary techniques
such as MR thermometry to provide comprehensive spatio-temporal feedback on release achieved
using non-invasive heating platforms such as HIFU and RF. One such application may be as pre-
treatment quality assurance toolset in a clinical setting. Prior to implementation in the clinic,
further work is needed to integrate the current platform for quantification of thermosensitive drug
release with MR thermometry techniques. The accuracy of current techniques for evaluating
hyperthermia with MR thermometry is severely compromised by contrast agent-induced
susceptibility effects, thus limiting its application in the context of imaging drug release [45]. As
such, decoupling the temperature effect on magnitude and the susceptibility effect on temperature
measurements would facilitate the quantification of drug release in response to guided
hyperthermia applications. The current platform may be combined with spatial temperature maps
to correlate the drug release achieved in response to in situ hyperthermia and thus enable treatment
planning for temperature sensitive drug delivery systems.
51
References
[1] Y. N. Dou et al., “Heat-activated thermosensitive liposomal cisplatin (HTLC) results in
effective growth delay of cervical carcinoma in mice,” J. Control. Release, vol. 178, no. 1,
pp. 69–78, 2014.
[2] L. Li et al., “Mild hyperthermia triggered doxorubicin release from optimized stealth
thermosensitive liposomes improves intratumoral drug delivery and efficacy,” J. Control.
Release, vol. 168, no. 2, pp. 142–150, 2013.
[3] T. Tagami, M. J. Ernsting, and S. D. Li, “Efficient tumor regression by a single and low
dose treatment with a novel and enhanced formulation of thermosensitive liposomal
doxorubicin,” J. Control. Release, vol. 152, no. 2, pp. 303–309, 2011.
[4] P. S. Yarmolenko et al., “Comparative effects of thermosensitive doxorubicin-containing
liposomes and hyperthermia in human and murine tumours,” Int. J. Hyperth., vol. 26, no.
5, pp. 485–498, 2010.
[5] A. M. Ponce, Z. Vujaskovic, F. Yuan, D. Needham, and M. W. Dewhirst, “Hyperthermia
mediated liposomal drug delivery,” Int. J. Hyperth., vol. 22, no. 3, pp. 205–213, 2006.
[6] D. Needham and M. Dewhirst, “The development and testing of a new temperature-
sensitive drug delivery system for the treatment of solid tumors,” Adv. Drug Deliv. Rev.,
vol. 53, no. 3, pp. 285–305, 2001.
[7] K. Park, “Lessons learned from thermosensitive liposomes for improved chemotherapy,”
J. Control. Release, vol. 174, no. 1, p. 219, 2014.
[8] D. Needham, “Reverse engineering of the low temperature-sensitive liposome (LTSL) for
treating cancer,” in Biomaterials for Cancer Therapeutics, Elsevier, 2013, p. 270–353e.
[9] A. Kim, “A Phase I Study of Lyso-thermosensitive Liposomal Doxorubicin and MR-
HIFU for Pediatric Refractory Solid Tumors,” clinicaltrials.gov, 2015. [Online].