Methods for an improved detection of the MRI-CEST effect
Citation for published version (APA):
Terreno, E., Stancanello, J., Longo, D., Delli Castelli, D., Milone, L., Sanders, H. M. H. F., Kok, M. B., Uggeri, F.,
& Aime, S. (2009). Methods for an improved detection of the MRI-CEST effect. Contrast Media and Molecular
Imaging, 4(5), 237-247. https://doi.org/10.1002/cmmi.290
DOI:
10.1002/cmmi.290
Document status and date:
Published: 01/01/2009
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be
important differences between the submitted version and the official published version of record. People
interested in the research are advised to contact the author for the final version of the publication, or visit the
DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page
numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne
Take down policy
If you believe that this document breaches copyright please contact us at:
openaccess@tue.nl
Fulvio Uggeri
band Silvio Aime
aCEST imaging is a recently introduced MRI contrast modality based on the use of endogenous or exogenous molecules whose exchangeable proton pools transfer saturated magnetization to bulk water, thus creating negative contrast. One of the critical issues for further development of these agents is represented by their limited sensitivity in vivo. The aim of this work is to improve the detection of CEST agents by exploring new approaches through which the saturation transfer (ST) effect can be enhanced. The performance of the proposed methods has been tested in vitro and in vivo using highly sensitive and highly shifted lipoCEST agents, and the results were compared with the standard ST evaluation mode. The acquired Z-spectra were interpolated locally and voxel-by-voxel by smoothing splines. Besides expressing the ST in the standard mode, we explore two methods, enhanced and integral ST, which better exploit all the information contained in the Z-spectrum. By combining different modes for ST assessment a significant improvement in the detection of the lipoCEST agents, both in vitro and in vivo, has been found. The results obtained from the application of the proposed methods outline the importance of post-processing analysis for highlighting the CEST-MRI contrast. Copyright # 2009 John Wiley & Sons, Ltd.
Keywords: CEST; saturation transfer; liposomes; MRI
1.
Introduction
The chemical exchange saturation transfer (CEST) modality is a recently introduced imaging procedure based on the use of molecules (CEST agents) containing one or more exchangeable proton pools (1–3). Radiofrequency-irradiation of the resonance of such mobile protons results in a saturated magnetization, which is transferred to the water signal by chemical exchange. The contrast in the resulting magnetic resonance imaging (MRI) experiment is then determined by the extent of the saturated magnetization transfer. Thus, differently from conventional MRI agents, the contrast is ‘frequency-encoded’. The decrease of the water signal intensity depends on the interplay between the physico-chemical properties of the CEST agent and instrumental parameters. Among the several advantages offered by CEST-based imaging, we can highlight: (i) the possibility of switching the contrast ‘on’ or ‘off’ at will simply by changing the irradiation frequency offset; (ii) the possibility of simultaneously visualizing many CEST agents in the same region; and (iii) the pre-contrast image can be acquired almost simultaneously to the CEST image just by off-setting the irradiation frequency. Several studies illustrating the high potential of CEST agents have already been reported, including the possibility of using these probes as ‘smart’ agents for pH measurements (1–4), cell labeling experiments (5), the imaging of diagnostic markers such as glucose (6,7), lactate (8) and zinc (9), the visualization of enzyme activity (10), and even monitoring gene expression (11).
CEST-MR images can be generated upon the irradiation of mobile protons belonging to either endogenous or exogenous molecules. The former approach offers the advantage of being fully non-invasive (12,13), but it requires quite a high concen-tration of endogenous proton sites, and it may suffer from the
small frequency offset between the absorption frequency of the exchanging proton pool and the bulk water signal. The resulting small saturation transfer response is therefore particularly challenging when dealing with noisy in vivo applications. Nevertheless, clinically relevant applications in which endogen-ous amide protons are irradiated have been recently reported (14–18). The term exogenous CEST agents refers to both diamagnetic (DIACEST) and paramagnetic (PARACEST) systems. DIACEST probes have the same limitations as those that have already been mentioned for endogenous compounds, but their sensitivity can be significantly improved by increasing the number of labile protons per molecule and optimizing their exchange rate (19,20). On the other hand, it has been demonstrated that
* Correspondence to: E. Terreno, Dipartimento di Chimica IFM and Molecular Imaging Center, University of Torino, Via P. Giuria 7, 10125, Torino, Italy. a E. Terreno, D. Longo, D. D. Castelli, L. Milone, S. Aime
Dipartimento di Chimica IFM and Molecular Imaging Center, University of Torino, Via P. Giuria 7, 10125, Torino, Italy
b J. Stancanello, F. Uggeri
CRM Bracco Imaging S.p.A., Via Ribes 5, Colleretto Giacosa (TO), Italy c H. M. H. F. Sanders, M. Kok
Biomedical NMR, Department of Biomedical Engineering, Eindhoven University of Technology, Postbox 513, 5600 MB, Eindhoven, The Netherlands Contract/grant sponsor: EC-PF6 Project DiMI; contract/grant number: LSHB-CT-2005-512146.
Contract/grant sponsor: EC-PF6 Project MEDITRANS; contract/grant number: NMP4-CT-2006-026668.
Contract/grant sponsor: EC-PF7 project ENCITE; contract/grant number:
PARACEST agents can considerably extend the range of accessible irradiation offset values (Dv), thus allowing the successful saturation transfer of mobile protons (large Dv ! large kex).
Moreover, the large Dv values strongly limit the detrimental effects usually associated with the direct saturation of bulk water as well as the occurrence of artifacts related to the inhomogen-eity of its absorption frequency (21,22). The in vivo evaluation of the CEST response may have several shortcomings. Errors mainly arise from direct saturation (spillover) effects, from endogenous (non-symmetric) magnetization transfer (MT) (23) and from inter-voxel differences in the water resonance frequency offset caused by tissue heterogeneity, non-optimal sample shimming and tuning, radiation damping (24), B0and B1inhomogeneity and
periodical or non-periodical movements and deformations related to organ motions (25). Thus, all these error sources may significantly affect the accuracy of the ST determination, producing a decrease or an erroneous, fictitious increase in its value, according to the combined results of all the previously mentioned effects.
MT effects are not considered in the Bloch equations modified for including exchange terms, and, in addition, many parameters of such a model are unknown. Analytical (26) and numerical (27,28) approaches have been proposed in an attempt to correctly describe the MR-CEST experiment, but their in vivo application still remains quite challenging. In this setting, we recently developed a procedure for determining and locally compensating ‘shift effects’ in the bulk water signal based on the voxel-by-voxel interpolation of Z-spectra by smoothing splines (29). A similar approach based on the acquisition of a Z-spectrum and specifically devoted to the accurate determination of the bulk water resonance for CEST agents with very small frequency offsets was published afterwards (30).
The present work is aimed at investigating the possibility of exploring novel approaches for calculating the CEST effect in order to improve the use of this kind of MR contrast. The proposed modalities were tested, and compared with the standard determination mode, either in vitro or in vivo, using liposome-based CEST (lipoCEST) agents as highly-sensitive CEST probes (31). Additionally, to facilitate the visualization of the agent and promote multiplex visualization, we used the novel generation of osmotically shrunken lipoCEST agents (32), which are characterized by a higher intraliposomal water proton resonance chemical shift.
2.
Results and discussion
2.1. New modes for detecting ST contrast
Usually, the CEST effect is determined by comparing the intensity of the water signal upon the irradiation of two different frequency offsets which are equally spaced from the resonance of the bulk water (fixed to 0), of which one (ION) corresponds to the resonance frequency offset of the mobile protons of the CEST agent, and the other (IOFF) corresponds to the water intensity at the contralateral frequency. This procedure is necessary in order to take into account any symmetric effects on the water signal caused by the application of the irradiation pulse.
Thus, classically, the saturation transfer (ST) efficiency is described by the use of the following equation:
ST¼ 1I ON IOFF ¼ I OFF ION IOFF (1)
The CEST contrast efficiency is usually expressed as a percentage (ST%¼ ST 100). The ST value should be close to zero in all frequency regions, except for the region around the absorbance of the mobile protons of the CEST agent. Equation (1) suggests that ST ranges from 0, when no CEST effect is registered, to 1, when, conversely, the saturation transfer is so efficient that it completely nullifies the water signal.
Other authors (3) calculate the CEST contrast (here indicated as ST0) by using the water signal intensity without irradiation (I0) as a reference:
ST0¼I
OFF ION
I0 (2)
Quantitatively, ST > ST0when I0
>IOFF, and this generally occurs when the saturation pulse leads to non-negligible direct saturation effects (spillover effects) on the bulk water signal.
In both of the above-mentioned cases the CEST effect is obtained by comparing two intensities measured upon irradia-tion at a single frequency offset. Thus, both ST and ST0may be
considered as ‘single frequency’ functions of the saturation transfer efficiency and the approaches for calculating such effects will be referred to, throughout this paper, as punctual modes.
An improvement in the detection of the CEST contrast may be obtained when the ST effect is calculated by increasing the dynamic range of the ST values. An increase in the dynamic range can be obtained by considering a range of frequencies in the neighborhood of the frequency offset classically used for determining the punctual ST.
Through this new approach, the ST is calculated by comparing two signal areas, i.e. by comparing two signal intensities measured as the integral of the signal area calculated in a given range of frequency offset, instead of two punctual intensities. In such a way an integral-based ST may be obtained that is not simply representative of the punctual ST obtained at the offset frequency, but, rather, that is representative of the information contained inside a range of frequencies in the neighbourhood of the offset frequency where the ST is at a maximum. As a result, an increase in the contrast-to-noise ratio (CNR) may be obtained,
Figure 1. Graphic illustration of STINT_BULK. A simulated Z-spectrum obtained for a single CEST agent highlights the two areas used in the calculation of STINT_BULK[eqn (3)]: the area subtended by the Z-spectrum between the bulk water resonance frequency (‘zero offset’) and the frequency where ST reaches its local maximum (dark gray area, on the left), and the contralateral area (light gray area, on the right).
resulting in an improvement in the detection of the CEST contrast. It is expected that this modality will be particularly useful in the case of broad CEST peaks, e.g. those typically observed for paramagnetic CEST agents with fast exchanging protons pools (2).
Two integral-based modalities can be envisaged, termed STINT_BULK and STINT_PEAK, respectively. STINT_BULK refers to a
modality in which the integral ST is calculated by comparing the Z-spectrum area subtended between the bulk water resonance frequency (zero offset) and a frequency that include the CEST peak, hereinafter identified as AON
INT BULK, with the corresponding
contralateral region, hereinafter identified AOFF INT BULK.
As an example, the AON
INT BULK and AOFFINT BULK areas are
repre-sented in Fig. 1 as dark gray area and light gray area, respectively.
In this modality, the ST effect may be calculated using the following equation: STINT BULK¼ 1 AON INT BULK AOFF INT BULK ¼ A OFF
INT BULK AONINT BULK
AOFF INT BULK
(3)
In principle, the STINT_BULKapproach provides enhanced CEST
contrast by adding to the effect that is normally generated by the CEST peak, the contribution associated with deriving by any asymmetry in the direct saturation peak of the bulk water that can be related to the presence of the CEST agent. On this basis, the method might be useful both in vitro, when the CEST probe appears in the Z-spectrum as a shoulder of the bulk water signal, and in vivo, where the conventional magnetization transfer
Figure 2. Graphic illustration of STINT_PEAK. Simulated Z-spectra obtained for a single CEST agent (CA, left, a) and for two CEST agents (CA1 and CA2, right, b) highlight the areas used in the calculation of STINT_PEAK[eqn (4)]: the regions are visualized in dark gray (on the left side) and light gray (on the right side) to distinguish between the area related to the CEST effect and the contralateral area, respectively.
Figure 3. Standard vs enhanced modalities for the punctual ST. The enhanced modality becomes more and more advantageous as the standard ST
contrast makes the visualization of the CEST peak of the agent difficult.
A possible drawback of this method may arise from the fact that, besides the direct saturation of the peak asymmetry, asymmetries coming from the artifacts in the low range fre-quency can also contribute to enhance the STINT_BULKvalue.
This disadvantage can be avoided by using a second integral-based modality, called STINT_PEAK, in which the ST effect
is calculated by comparing the Z-spectrum area subtended in two frequency intervals, symmetrically located with respect to the bulk water, and centered at the frequency where the punctual ST reaches its maximum value (Fig. 2).
This STINT_PEAKmodality can be calculated according to eqn (4):
STINT PEAK¼ 1 AON INT PEAK AOFF INT PEAK ¼ A OFF
INT PEAK AONINT PEAK
AOFF INT PEAK
(4)
The STINT_PEAK method appears particularly useful for the
detection of CEST contrast when the CEST peak is broad, i.e. in the cases of mobile protons approaching coalescence, the dispersion
of chemical shift values in magnetically pseudo-equivalent mobile protons, or short T2values of the exchanging protons.
Another way to increase the dynamic range of the ST effect can be obtained in the punctual modality by using the IONvalue as a
reference instead of IOFF. Analogous effects can be achieved in the integral-based modalities when AON
INT BULKand AONINT PEAKvalues are
used as reference instead of AOFF
INT BULKand AOFFINT PEAK.
This approach leads to three new expressions, hereinafter labeled as enhanced modes, in which the ST effect is calculated as follows: STENH¼ 1 I OFF ION ¼ I OFF ION ION (5) STENH INT BULK¼ 1 AOFF INT BULK AON INT BULK ¼ A OFF
INT BULK AONINT BULK
AON INT BULK
(6)
STINT PEAKENH ¼ 1 A
OFF INT PEAK AON INT PEAK ¼ A OFF
INT PEAK AONINT PEAK
AON INT PEAK
(7)
Differently from the standard modes, the enhanced expressions yield values ranging from 0, when no ST effect is present, to1 when the ST effect is maximum (ON signal is zero). A graphical correlation between standard and enhanced representations is displayed in Fig. 3 for the punctual ST mode.
The plot clearly indicates that the enhanced approach starts to be particularly advantageous when the saturation transfer response is relatively high. As a consequence, the noise, that is usually lower than 0.1, will not be amplified, thus increasing the CEST contrast-to-noise ratio.
Finally, a further representation of the recorded CEST effect may be obtained by modulating the dynamic range of the ST effect through an approach hereinafter identified as complemen-tary. This modality may be applied to all the ST modalities described so far, thus providing a novel set of complementary modalities where the ST effect is calculated by using the following equations: STCOM¼ 1 ð1 I ONÞ ð1 IOFFÞ ¼I OFF ION 1 IOFF (8) STENH COM¼ 1ð1 I OFFÞ ð1 IONÞ ¼I OFF ION 1 ION (9) Figure 4. Classical versus complementary modes for STINT_PEAK. A
simu-lated Z-spectrum highlights the differences between standard and comp-lementary integral values to be used for calculating STINT_PEAK.
Table 1. ST effects for six simulated experiments with different normalized ON and OFF water intensities
ST STENH STCOMP STENH_COMP
Experiment 1 ION¼ 0.8 0.16 0.19 3.0 0.75 IOFF¼ 0.95 Experiment 2 ION¼ 0.6 0.4 0.29 IOFF¼ 0.715 Experiment 3 ION¼ 0.4 0.14 0.125 IOFF¼ 0.475 Experiment 4 ION¼ 0.45 0.5 1.0 4.5 0.81 IOFF¼ 0.9 Experiment 5 ION¼ 0.3 0.75 0.43 IOFF¼ 0.6 Experiment 6 ION¼ 0.1 0.125 0.11 IOFF¼ 0.2
240
STINT BULKCOM ¼ 1 ðAINT BULK A
ON INT BULK
ðAINT BULK AOFFINT BULK
¼A
OFF
INT BULK AONINT BULK
AINT BULK AOFFINT BULK
(10)
STENH COM INT BULK ¼ 1
ðAINT BULK AOFF INT BULKÞ
ðAINT BULK AON INT BULKÞ
¼A
OFF
INT BULK AONINT BULK
AINT BULK AON INT BULK
(11)
STCOM
INT PEAK¼ 1
ðAINT PEAK AON INT PEAKÞ
ðAINT PEAK AOFF INT PEAKÞ
¼A
OFF
INT PEAK AONINT PEAK
AINT PEAK AOFF INT PEAK
(12)
STINT PEAKENH COM¼ 1ðAINT PEAK AOFFINT PEAKÞ
ðAINT PEAK AON INT PEAKÞ
¼A
OFF
INT PEAK AONINT PEAK
AINT PEAK AON INT PEAK
(13)
Figure 5. Set of the simulated Z-spectra used for the in silico testing of the ST modes proposed in this work. Each spectrum reports the punctual standard ST value.
The term complementary, herein used to identify these modes, may be better understood by considering that they are all based on the use of water intensities/areas that are complementary intensities with respect to the values used for the other, standard modalities previously described. For instance, when considering the standard punctual approach, 1 is the maximum intensity value in the normalized Z-spectrum, and the corresponding value in the complementary modality is zero (i.e. the complementary value to 1). In the integral-based approach the values used are those complementary to the area subtended to the whole Z-spectrum region considered. In this case, the reference (AINT_BULK or
AINT_PEAK) is the area of the quadrilateral with height 1 and base
corresponding to the frequency range considered. Figure 4 graphically shows this approach as well as the difference existing between standard and complementary approaches for the INT_PEAK modality.
Table 1 compares the performance of some of the comp-lementary modes with respect to the corresponding standard and enhanced versions for six datasets with different normalized water intensities. The virtual experiments 1–3 have the same, relatively small, standard ST value (0.19), whereas experiments 4–6 have a more efficient saturation transfer (0.5). The results obtained indicate that the performance of the complementary mode improves as the normalized water intensity at the contralateral side (IOFF) increases and it exceeds the standard methods for IOFF larger than 0.5. It is important to note that, differently from the standard mode, the complementary approach does not draw any benefit from the enhanced version. In fact, the term at the denominator (1 – ION) increases as the saturation
transfer gets more efficient, thus reducing the overall saturation transfer. Therefore, we selected eight novel modalities (STENH, STCOM, STINT_BULK, STENHINT_BULK, STCOMINT_BULK, STINT_PEAK,
STENHINT_PEAK and ST COM
INT_PEAK) to be tested and compared
with that of the standard punctual mode in different experimental environments (in silico, in vitro and in vivo). The comparison was carried out by calculating first the CEST CNR as the ratio between the ST effect of a given mode and its standard deviation. Then, the CNR value of the given mode was normalized to the CNR of the standard punctual modality (CNR normalized to ST).
2.2. In silico test
Six Z-spectra, differing in the broadness of their CEST and bulk water peaks, in the frequency separation between the exchan-ging protons, and in the concentration of the mobile protons, were generated using the numerical model developed by Woessner et al. (27) (Fig. 5). Furthermore, a Gaussian error of 10% was applied on the simulated data in order to better mimic real experiments. The standard deviation was calculated for each mode (at the same frequency/frequency interval) by analyzing five Z-spectra simulated without the presence of the mobile protons of the CEST agent.
The histogram displayed in Fig. 6 illustrates the improvement provided by the use of both the punctual enhanced and the complementary modes with respect to the standard one. In particular, when the ST efficiency is high (as in A2, B2, and C2 datasets) the enhanced mode works very well, whereas for the data simulating low-efficiency systems, especially B1 and C1 datasets, the complementary mode exhibited a better perform-ance. In the case of A1 spectrum, the low ST value in combination with the small (IOFF-ION) difference makes the complementary
mode perform less well than the standard one. In addition, it has
to be noted that the standard deviation of the complementary mode is much higher (ca 20-fold) than both the standard and the enhanced versions, thus decreasing the overall performance of this modality. However, it is important to point out that for each dataset there exists a novel punctual mode that works better than the standard method.
In Fig. 7, the comparison between the performance of standard punctual and integral-based methods is reported. In general, the ST effects calculated using the integral-based methods displayed higher normalized CNR values, especially for the peak integral case. As expected, the integral approach provides better results when the CEST peak is broader (B1 and B2 datasets), whereas for narrow peaks (A1 and A2 datasets) a benefit was observed for the peak integral mode only. It is worth noting that both the integral-based modes displayed very small errors, thus improving the CNR value.
Figure 6. In silico comparison between complementary and enhanced ST punctual methods. CNR (normalized to the standard punctual ST) for STCOMPand STENHmodes obtained from the analysis of the simulated Z-spectra reported in Fig. 5.
Figure 7. In silico comparison between STINT_BULKand STINT_PEAKmodes. CNR (normalized to the standard punctual ST) for the integral-based STINT_BULK and STINT_PEAKmodes as calculated from the analysis of the simulated Z-spectra reported in Fig. 5.
Figure 8 illustrates the performance of the complementary and the enhanced versions of the integral peak modality. It can be noted that the proposed modes enhance the normalized CNR values in all the examined samples, especially when the spectrum is characterized by broad (B1 and B2) or low shifted (C1 and C2) CEST peaks. Similar results were observed for the integral bulk modality.
2.3. In vitro experiments
The set of novel ST methods was also tested in vitro in a phantom consisting of six capillaries filled with differently concentrated suspensions of osmotically shrunken lipoCEST agents, loaded with Dy(III)-based shift reagents, whose intraliposomal water proton shift is45 ppm. Z-spectra at different saturation pulse amplitudes (B2equal to 3, 6 and 12 mT) were acquired. Figure 9
shows the Z-spectra obtained at the different intensities for the most concentrated solution (top), and the Z-spectra obtained at 12 mT for the most and the least concentrated solutions (bottom). In all samples, the CEST peak corresponding to the paramagne-tically shifted resonance of the intraliposomal water protons is clearly detectable.
The normalized CNR values were determined by calculating the standard deviation in a solution containing the same buffer in which the lipoCEST agent was suspended. The obtained results are reported in Fig. 10 for the least concentrated solution upon irradiation at 6 mT.
All the novel ST methods proposed in this study worked better than the standard punctual ST. In particular, great benefits were obtained with the complementary integral-based modes. This observation is not unexpected because the Z-spectrum of this sample is characterized by a small, broad and well-shifted CEST peak. Such conditions lead to very large AOFFvalues that, in turn, cause a significant increase of the complementary modes [see eqns (10) and (12)].
Figure 11 shows the results obtained at 12 mT for the sample with the highest (black), the lowest (white), and the intermediate (gray) concentration of the CEST agent. The results obtained indicate once again the peculiar properties of the enhanced and
Figure 9. In vitro Z-spectra of a phantom of lipoCEST agents. Top: Z-spectra of a suspension of a Dy-based lipoCEST agent acquired at three saturation pulse intensities. Bottom: Z-spectra acquired at 12 mT for the most (solid squares) and the least (open squares) concentrated solution of the lipoCEST agent. The lines represent a simple data interpolation to guide the eye.
Figure 10. In vitro performance. CNR (normalized to the standard punc-tual ST) for all the proposed ST modes calculated from the analysis of the in vitro Z-spectrum (saturation amplitude 6 mT) for the least concentrated solution of the Dy-loaded lipoCEST agent.
Figure 8. In silico comparison between standard, complementary and enhanced expressions of STINT_PEAKmode. CNR (normalized to the stan-dard punctual ST) for STINT_PEAK(and its complementary and enhanced versions) calculated from the analysis of the simulated Z-spectra reported in Fig. 5.
complementary modes, with the former that enhance the CNR at high ST efficiency (so at high concentration of CEST agent) and the latter that perform much better when the saturation transfer efficiency decreases. The integral-based methods STINT_BULKand
STINT_PEAKwork better than the standard ST, but did not show any
significant concentration effect. The gain in the CEST agent detection is clearly evident in Fig. 12 where the performance upon a saturation pulse amplitude of 6 mT of standard punctual ST and STCOMINT_PEAKare compared for the phantom made of the
lipoCEST agents.
2.4. Ex vivo and in vivo experiments
The performance of the proposed ST methods were also tested in biological environments that are usually characterized by a very broad peak centered in close proximity to the bulk water resonance and associated with endogenous mobile protons interacting with tissues, which display NMR behavior similar to that of solid-like systems. The experiments consisted of injecting lipoCEST agents into bovine muscle (ex vivo), subcutaneously into a mouse flank (in vivo), and intratumorally into a mouse bearing a xenografted B16 melanoma (in vivo).
Figure 13 shows the normalized CNR values for the proposed ST modes after the injection into a bovine muscle of a Tm-based osmotically shrunken lipoCEST agent whose intraliposomal water protons were shifted by 18 ppm from the bulk water. In this specific case most of the applied methods enhanced the CNR values, but to a much lower extent than the previously described in silico and in vitro experiments. The bottom of Fig. 13 shows two MR-CEST maps calculated according to the standard ST and its enhanced version. In the latter map the injection site of the lipoCEST agent is more easily detectable.
Figure 11. In vitro performance. CNR (normalized to the standard punc-tual ST) for all the proposed ST modes calculated from the analysis of the in vitro Z-spectra (saturation amplitude 12 mT) for the highest (black bars), the lowest (white bars), and the intermediate (gray bars) concentration of the Dy-based lipoCEST agent.
Figure 12. In vitro MR-CEST images of a phantom of a lipoCEST agent. Standard punctual ST (left) and STCOM
INT PEAK(right) maps (saturation amplitude 6 mT)
superimposed to the in vitro MR image of the phantom of lipoCEST agent at different concentrations (the concentration decreases following the order 1–6).
Figure 13. Ex vivo performance. Top: CNR (normalized to the standard punctual ST) for all the proposed ST modes calculated from the analysis of the ex vivo MR-CEST images (saturation amplitude 6 mT) of a bovine muscle injected with a Tm-based lipoCEST agent (saturation frequency offset equal to 18 ppm). Bottom: standard and enhanced punctual ST maps superimposed to the MR image of the bovine muscle injected with the lipoCEST agent.
As far as the in vivo experiments is concerned, in one case two lipoCEST agents with different intraliposomal water proton chemical shift values (Tm-based, offset 7 ppm, and Dy-based, offset17 ppm) were subcutaneously injected into the flanks of a mouse. In a second experiment, a Tm-based lipoCEST (saturation
3.
Conclusions
In this work a set of novel procedures for calculating the MR-CEST contrast have been considered. The proposed methods are designed to exploit at best all the information that characterize the Z-spectrum in order to improve the CEST contrast detection (e.g. linewidth of the bulk water and CEST peaks, frequency offset and depth of the latter signal). The performance of the ST modalities has been tested under different conditions (in silico, in vitro, ex vivo and in vivo) and compared with the modality routinely used for determining the ST efficiency.
The obtained results point to show that the performance of the proposed methods is markedly dependent on the specific characteristics of the Z-spectrum, that are not only sensitive to the properties of the CEST probe, but are also strongly influenced
Figure 14. In vivo performance. CNR (normalized to the standard punc-tual ST) for all the proposed ST modes calculated from the analysis of two in vivo CEST-MR experiments (saturation amplitude 6 mT): (1) two lipoCEST agents differing in the chemical shift of their intraliposomal water protons (Tm-based ! black bars, and Dy-based ! white bars) were subcu-taneously injected in the flank of a mouse; and (2) a Tm-based lipoCEST agent was intratumorally injected in a mouse bearing a xenografted B16 melanoma (gray bars).
by the environment in which the probe is distributed. However, some general indications can be drawn. The enhanced approach works very well and a significant increase of the CEST CNR can be observed, but only when the CEST contribution by the probe is already well delineated (standard ST > 0.3). This is the case encountered in many in vitro studies characterized by high concentration of the exchangeable protons and no interference by the environment. When the CEST peak is distant from the bulk water resonance and broadened (typical for PARACEST agents) the integral-based methods, especially INT_PEAK, appear advan-tageous. In sample with a small ST efficiency or in biological environments where a significant level of endogenous magne-tization transfer contrast is typically observed, the complementary approaches appear to perform better than standard ST determination.
We believe that the methodological achievements reported in this work, in addition to the optimization of all the other steps required for the set-up of a CEST-MR experiment (hardware, pulse sequences, probe design, etc.), may significantly contribute to increasing the potential of this class of MR imaging agents.
4.
Experimental
4.1. Preparation of LIPOCEST agents
Osmotically shrunken LIPOCEST agents, encapsulating a hydro-philic lanthanide-based shift reagent (Ln-HPDO3A) and incorpor-ating an amphiphilic lanthanide-based shift reagent [Ln-1 in Terreno et al. (32), Chart 1], were prepared following the pro-cedure described in Terreno et al. (32). The lanthanide complexes were kindly provided by Bracco Imaging (Colleretto Giacosa, Turin, Italy). For in vitro experiments, two LIPOCEST agents, differing in the nature of the paramagnetic Ln(III) ion, [Tm(III) and Dy(III)] were prepared. The formulation of the liposome membrane was 1,2-dipalmitoyl-sn-glycero-3-phosphocholine– 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(poly-ethylene glycol)-2000]–Ln-1 (75:5:20 in molar ratio, total amount of lipids: 20 mg). The concentration of the Ln-HPDO3A complexes in the aqueous solution used for hydrating the lipidic film (volume 1 ml) was 10 mM. After the extrusion, liposomes were
dialyzed against an isotonic physiological buffer (pH 7.4, NaCl 0.15Mþ HEPES 2.8 mM) in order to remove the non-encapsulated compound and induce the hyperosmotic stress necessary to change the liposome shape. Liposome size was determined by dynamic light scattering (NanoS Zetasizer, Malvern, UK) and was shown to be 146 and 158 nm for Tm- and Dy-based lipoCEST agents respectively (polydispersity index, PDI < 0.1). The chemi-cal shift difference between the intraliposomal water protons and bulk water was 17 and45 ppm for Tm- and Dy-based lipoCEST agents respectively. This was assessed by high-resolution NMR spectra carried out at 14 T and 378C (Bruker Avance 600). In vitro experiments were carried out in a phantom consisting of six capillaries filled with solutions at different liposome concen-trations prepared by diluting the initial suspension with isotonic buffer (dilution factors 0, 1, 2, 4, 8, 16).
The liposome concentrations in the suspension were roughly estimated by using the method described in Terreno et al. (33), and were 52 and 44 nMfor Tm- and Dy-based lipoCEST agents, respectively.
For in vivo experiments, other two osmotically shrunken lipoCEST agents (again Tm- and Dy-based) with higher sensitivity
were prepared. The membrane formulation was 1-palmitoyl-2-oleyl-sn-glycero-3-phosphocholine–cholesterol–DSPE-PEG2000– Ln-1 (50:15:5:30 in molar ratio, total amount of lipids 20 mg). The concentration of the Ln-HPDO3A complexes in the aqueous solution used for hydrating the lipidic film (volume 1 ml) was 40 mM. The liposome size was 240 and 260 nm for Tm- and
Dy-based lipoCESTs, respectively (PDI < 0.1). The chemical shift difference between the intraliposomal water protons and bulk water was 7 and15 ppm for Tm- and Dy-based lipoCEST agents, respectively. In vivo experiments were performed by the subcutaneous injection of the two lipoCEST probes (ca 100 ml) into the flanks of a healthy mouse.
4.2. MRI acquisitions
MR images were acquired on a Bruker Avance 300 spectrometer operating at 7 T and equipped with a microimaging probe. Two different resonators [diameters of 10 and 30 mm used for in vitro and in vivo (or ex vivo) experiments respectively] were used. Z-spectra were acquired in the frequency offset range150 ppm using a Spin Echo RARE sequence (typical setting TR/TE/NEX/ RARE factor¼ 6000 ms/15 ms/1/8) preceded by a saturation transfer pulse (single block pulse, duration 2 s) at three amplitudes (B2equal to 3, 6 and 12 mT). A 64 64 acquisition
matrix was used with a slice thickness of 2 mm and FOV values of 10 10 or 30 30 mm, depending on the resonator. Proton density images were obtained using a standard spin echo pulse sequence (TR/TE/NEX¼ 6000 ms/3 ms/1). All the in vivo acqui-sitions were performed using a standard fat-suppression module.
4.3. Image and data analysis
Raw Z-spectra data were exported from the MRI workstation and automatically elaborated by means of software executed in MATLAB (The Mathworks Inc., Natick, MA, USA). The analysis consisted of several steps [image segmentation, global region of interest (ROI) analysis, local voxel-by-voxel analysis] according to our previously reported method (29). In particular, after the automatic segmentation of the morphological image, an ROI was manually selected. This identification aims at paying particular attention to some global ROI features such as the bulk water resonance frequency and the ST response inside a selected region (for example, investigation can be focused on tumor contrast uptake, blood vessels or edemas). To build Z-spectra, we sampled the signal by using a 1 ppm sampling step for the irradiation frequency values close to the bulk water resonance. When investigating ST effects at larger offset values, the response showed smooth variations (except for noise), supporting the use of a lower sampling rate. Z-spectra were interpolated by using smoothing splines (29).
On the base of the global analysis of the selected ROI, the segmented image can be locally investigated by calculating the ST at any desired frequency offset. Because the global ROI analysis represents an average of the voxel-by-voxel behavior, local effects may be masked when averaged over a large ROI. This implies that global ROI analysis could mask the presence of local, low ST effects. For this reason, a local analysis, accounting for the different physico-chemical properties in each single voxel, is necessary. This voxel-by-voxel analysis consisted of interpolating the samples coming from each voxel of the segmented image in order to obtain the local Z-spectrum. The approach followed for the local analysis is the same as the global ROI analysis, i.e. based
Acknowledgements
Economic and scientific support from local (Nano-IGT and C 130 projects) and national government (FIRB RBNE03PX83_006, FIRB RBIP06293N, and PRIN 2005039914 projects), as well as EU projects DiMI (LSHB-CT-2005-512146), EMIL (LSHC-CT-2004-503569), MEDITRANS (Targeted Delivery of Nanomedicine: NMP4-CT-2006-026668) and ENCITE (201842) is gratefully acknowledged. This work was performed within the COSTD38 action. The authors thank Dr Dale Lawson for the stylistic and linguistic revision of the manuscript.
References
1. Ward KM, Aletras AH, Balaban RS. A new class of contrast agents for MRI based on proton chemical exchange dependent satur-ation transfer (CEST). J. Magn. Reson. 2000; 143: 79–87. 2. Woods M, Woessner DE, Sherry AD. Paramagnetic lanthanide
complexes as PARACEST agents for medical imaging. Chem. Soc. Rev. 2006; 35: 500–511.
3. Zhou J, van Zijl PCM. Chemical exchange saturation transfer imaging and spectroscopy. Prog. Nuclear Mag. Res. Spectrosc. 2006; 48: 109–136.
4. Ward KM, Balaban RS. Determination of pH using water protons and chemical exchange dependent saturation transfer (CEST). Magn. Reson. Med. 2000; 44: 799–802.
5. Aime S, Carrera C, Delli Castelli D, Geninatti Crich S, Terreno E. Tunable imaging of cells labeled with MRI-PARACEST agents. Angew. Chemie Int. Edn 2005; 44: 1813–1815.
6. Zhang S, Trokowski R, Sherry AD. A paramagnetic CEST agent for imaging glucose by MRI. J. Am. Chem. Soc. 2003; 125: 15288–15289.
7. Trokowski R, Zhang S, Sherry AD. Cyclen-based phenylboronate ligands and their Eu3þ complexes for sensing glucose by MRI. Bioconjug. Chem. 2004; 15: 1431–1440.
8. Aime S, Delli Castelli D, Fedeli F, Terreno E. A paramagnetic MRI-CEST agent responsive to lactate concentration. J. Am. Chem. Soc. 2002; 124: 9364–9365.
9. Trokowski R, Ren J, Kalman FK, Sherry AD. Selective sensing of zinc ions with a PARACEST contrast agent. Angew. Chemie Int. Edn 2005; 44: 6920–6923.
10. Yoo B, Pagel MD. A PARACEST MRI contrast agent to detect enzyme activity. J. Am. Chem. Soc. 2006; 128: 14032–14033. 11. Gilad AA, McMahon MT, Walczak P, Winnard PT Jr, Raman V, Bulte
JW, van Zijl PC. Artificial reporter gene providing MRI contrast in vivo based on proton exchange. Nat. Biotechnol. 2007; 25: 217–219.
12. Guivel-Scharen V, Sinnwell T, Wolff SD, Balaban RS. Detection of proton chemical exchange between metabolites and water in biological tissues. J. Magn. Reson. 1998; 133: 36–45.
13. Dagher AP, Aletras A, Choyke P, Balaban RS. Imaging of urea using chemical exchange-dependent saturation transfer at 1.5 T. J. Magn. Reson. 2000; 12: 745–748.
dependent saturation transfer (gagCEST). Proc. Natl Acad. Sci. 2008; 105: 2266–2270.
19. Goffeney N, Bulte JW, Duyn J, Bryant LH Jr, van Zijl PC. Sensitive NMR detection of cationic-polymer based gene delivery systems using saturation transfer via proton exchange. J. Am. Chem. Soc. 2001; 123: 8628–8629.
20. Snoussi K, Bulte JW, Gueron M, van Zijl PC. Sensitive CEST agents based on nucleic acid imino proton exchange: detection of poly(rU) and of a dendrimer-poly(rU) model for nucleic acid delivery and pharmacology. Magn. Reson. Med. 2003; 49: 998–1005.
21. Zhang S, Merritt M, Woessner DE, Lenkinski RE, Sherry AD. PARACEST agents: modulating MRI contrast via water proton exchange. Acc. Chem. Res. 2003; 36: 783–790.
22. Terreno E, Castelli DD, Cravotto G, Milone L, Aime S. Ln(III)-DOTAMGly complexes: a versatile series to assess the determi-nants of the efficacy of paramagnetic chemical exchange satur-ation transfer agents for magnetic resonance imaging applications. Invest. Radiol. 2004; 39: 235–243.
23. Pekar J, Jezzard P, Roberts DA, Leigh JS, Franck JA, McLaughlin AC. Perfusion imaging with compensation for asymmetric magneti-zation transfer effects. Magn. Reson. Med. 1996; 35: 70–79. 24. Williamson DC, Narvainen J, Hubbard PL, Kauppinen RA, Morris
GA. Effects of radiation dumping on Z-spectra. J. Magn. Reson. 2006; 183: 215–224.
25. Sun PZ, Farrar CT, Sorensen AG. Correction for artifacts induced by B0 and B1 field inhomogeneities in pH-sensitive chemical
exchange saturation transfer (CEST) Imag. Magn. Reson. Med. 2007; 58: 1207–1215.
26. Sun PZ, van Zijl PCM, Zhou J. Optimization of the irradiation power in chemical exchange dependent saturation transfer experiments. J. Magn. Reson. 2005; 175: 193–200.
27. Woessner DE, Zhang S, Merritt ME, Sherry AD. Numerical solution of the Bloch equations provides insights into the optimum design of PARACEST agents for MRI. Magn. Reson. Med. 2005; 53: 790–799.
28. Li AX, Hudson RH, Barrett JW, Jones CK, Pasternak SH, Bartha R. Four-pool modeling of proton exchange processes in biological systems in the presence of MRI-paramagnetic chemical exchange saturation transfer (PARACEST) agents. Magn. Reson. Med. 2008; 60: 1197–1206.
29. Stancanello J, Terreno E, Delli Castelli D, Cabella, Uggeri F, Aime S. Development and validation of a smoothing-splines-based cor-rection method for improving the analysis of CEST-MR images. Contrast Media Mol. Imag. 2008; 3: 136–149.
30. Kim M, Gillen J, Landman BA, Zhou J, van Zijl PC. Water saturation shift referencing (WASSR) for chemical exchange saturation transfer (CEST) experiments. Magn. Reson. Med. 2009; 61: 1441–1450.
31. Aime S, Delli Castelli D, Terreno E. Highly sensitive MRI chemical exchange saturation transfer agents using liposomes. Angew. Chemie Int. Edn 2005; 44: 5513–5515.
32. Terreno E, Cabella C, Carrera C, Delli Castelli D, Mazzon R, Rollet S, Stancanello J, Visigalli M, Aime S. From spherical to osmotically shrunken paramagnetic liposomes: an improved generation of LIPOCEST MRI agents with highly shifted water protons. Angew. Chemie Int. Edn 2007; 46: 966–968.
33. Terreno E, Delli Castelli D, Milone L, Rollet S, Stancanello J, Violante E, Aime S. First ex-vivo MRI co-localization of two LIPOCEST agents. Contr. Media Mol. Imag. 2008; 3: 38–43.
