DNA-Gadolinium-Gold nanoparticles for in vivo T1 MR imaging of transplanted human neural stem cells

The unambiguous imaging of transplanted cells remains a major challenge to understand their biological function and therapeutic efficacy. In vivo imaging of implanted cells is reliant on tagging these to differentiate them from host tissue, such as the brain. We here characterize a gold nanoparticle conjugate that is functionalized with modified deoxythymidine oligonucleotides bearing Gd(III) chelates and a red fluorescent Cy3 moiety to visualize in vivo transplanted human neural stem cells. This DNA-Gd@Au nanoparticle (DNA-Gd@AuNP) exhibits an improved T1 relaxivity and excellent cell uptake. No significant effects of cell uptake have been found on essential cell functions. Although T1 relaxivity is attenuated within cells, it is sufficiently preserved to afford the in vivo detection of transplanted cells using an optimized voxel size. In vivo MR images were corroborated by a post-mortem histological verification of DNA-Gd@AuNPs in transplanted cells. With 70% of cells being correctly identified using the DNA-Gd-AuNPs indicates an overall reliable detection. Less than 1% of cells were false positive for DNA-Gd@AuNPs, but a significant number of 30% false negatives reveals a dramatic underestimation of transplanted cells using this approach. DNA-Gd@AuNPs therefore offer new opportunities to visualize transplanted cells unequivocally using T1 contrast and use cellular MRI as a tool to derive biologically relevant information that allows us to understand how the survival and location of implanted cells determines therapeutic efficacy.


Introduction
The regional distribution of transplanted neural stem cells (NSCs) influences their sphere of activity and correlates with the degree of therapeutic efficacy [1]. A greater understanding of the positioning of transplanted NSCs could hence improve our targeting of cell injections to areas crucial to their efficacy. However, mapping the in vivo distribution of transplanted cells remains a major challenge [2,3]. In the context of intracerebral transplants, a range of 1,000-400,000 cells needs to be detected at a high in vivo spatial resolution (<64 nL voxel) using an unequivocal multi-voxel signal that predominantly derives from transplanted cells with a low risk of false positives (<5%, type I error) and negatives (<20%, type II error) [2,4]. The selective visualization of transplanted cells by magnetic resonance imaging, nevertheless, requires contrast-inducing particles [3].
Iron oxide (magnetite/maghemite) particles yield a high relaxivity that affords single cell tracking [5]. However, potential blooming artifacts due to air bubbles and small hemorrhage on T 2 -and T 2 *-weighted magnetic resonance images (MRI), in addition to a potential for nanotoxicity in neurons [6], complicates an unequivocal interpretation of in vivo cell distribution in the brain [2]. An unequivocal signal can potentially be produced by T1 agents, such as Mn(II) and Gd(III). Mn(II) agents are easily taken-up into cells by substituting for Ca 2+ ions. Although this affords the T1 detection of labeled cells [7][8][9][10], unchelated manganese which is required for cellular uptake is known to exert cytotoxic effects [11]. 4 Cellular labeling with monomeric gadolinium agents taken-up through endocytosis typically quenches the T1 contrast enhancement due to endosomal sequestration, but electroporation of the agent into the cytoplasm preserves the T1 signal [12]. Several reports indicate the possibility to visualize cells using MRI using this approach [13][14][15][16][17][18]. However, positive identification of a T1 signal in vivo requires evidence that the agent is localized intracellularly in the transplanted cells (as assessed by an independent marker). This is necessary to prevent false identification of contrast agent inadvertently injected, bound to the outside of the cell membrane or exocytosed from the transplanted cells.
The thermodynamic and kinetic stability of chelated Gd(III) compounds is also essential to avoid cytotoxicity, which can be delayed or avoided if Gd(III) can be contained within the chelate. Macrocylic ligands based on 1,4,7,10tetraazacyclododecane-1,4,7,10-teraacetic acid (DOTA) exhibit similar thermodynamic stabilities compared to those of the linear diethylene triamine (DTPA) ligand [19], but are more kinetically stable, and are thus a more favorable chelate for Gd(III)-based agents [20,21]. Immobilization of Gd(III) complexes onto macromolecules or proteins that restrict the motion of Gd(III) chelates can improve the relaxivity compared to monomeric Gd(III)-agents [22]. Intracellular concentration (and hence cellular relaxivity) could be further improved by using this strategy [23].
Achieving a high cellular relaxivity requires an optimized nanoconjugate Gd(III) contrast agent with a high thermostability that affords an efficient cell uptake. In recent work by Song et al [24], Gd(III) labeled DNA gold nanoparticle M A N U S C R I P T A C C E P T E D 5 conjugates (DNA-Gd@AuNP) were shown to be a biocompatible and highly effective platform for cell labeling of HeLa and NIH 3T3 cancer cells [25].
Importantly, the DNA-Gd@AuNP showed an excellent pH stability and inertness towards transmetallation [26], with attached oligonucleotides affording an efficient cell uptake through the CXCR7 scavenger receptor and peri-nuclear localization [27]. Especially in serum-free cultures (as used for human neural stem cells), these oligonucleotides are efficient trans-membrane localization tools. There is further evidence that AuNP in the proximity of Gd(III)-agents serve as "nanoamplifiers" to increase MR relaxivity [28].

ICP-MS analysis of labeled cells
The Gold and Gadolinium content of cells incubated with different NP concentrations for 24 hours was determined using ICP-MS. Cells were collected in 0.1 mL Accutase. A suspension of 5 µl was digested with 120 µL of nitric acid (70%, Sigma) and concentrated hydrochloric acid (37%, Sigma). To assay for Gd(III), samples were diluted to a total volume of 3 mL with Nanopure water to a final combined acid concentration of 2% (v/v) nitric and 2% (v/v) hydrochloric acids. To probe Au(III) content, samples were diluted to a total of 10 mL with Nanopure water with the same final concentrations of acid with Gd(III) samples.
Signal-to-noise (SNR) was calculated by dividing the mean signal from an ROI on the cell pellet by the mean signal outside of the sample, defined as noise.
Bone wax was used to fill the skull defect caused by the drill holes. Animals were sutured and given topical analgesic cream (2.5% Lidocaine and 2.5% Prilocaine, Sandoz) and Buprenex (0.05 mg/kg i.p.; Henry Schein).

Ex-vivo MR Imaging
To determine an appropriate voxel size for in vivo imaging, images for T1 maps were acquired using global Inversion Recovery (IR) with fast spin echo (FSE) readout, whereas T2 maps used a MSME sequence ( nl.

In vivo MR imaging
The day following cell implantation, animals underwent MR imaging (9.4 T). For this, animals were anesthetized using isofluorane (4% induction, 1.5% maintenance) in medical air, while their physiology (temperature, respiratory rate) was monitored using a MP150 Biopac system interfaced with AcqKnowledge v4.1 software (Biopac Systems Inc.). Based on the ex vivo voxel size and cell concentration experiment, T 1 -weighted MRI images were acquired using a Gradient Echo Multi Slice sequence (Table 1).

Statistical analyses
Statistical analysis of cell counts and measurements were performed on mean values using one-or two-way analysis of variance (ANOVA) followed by

Particle loading was assessed by ICP-MS analysis of fully synthesized particles
by observation of the ratio of Gd(III) to Au(III) ratio. By calculating the volume of the nanoparticles using a geometric approximation, based on the observed size by TEM ( Figure 1B), and applying the density of bulk gold, a particle loading of 326±44 Gd(III) molecules per AuNP was determined. This Gd(III) loading corresponds to a particle loading of 65±9 DNA strands per particle.

DNA-Gd@AuNP stability and relaxivity
The stability of the nanoparticles is essential to ensure detection of implanted cells. The relaxivity (r 1 and r 2 ) of DNA-Gd@AuNP nanoparticles was hence measured in solution for 17 days ( Figure 1B). At 1.41 T, r 1 was stable at 14.6 mM -1 s -1 , as was r 2 at 38.5 mM -1 s -1 (measured at 37 °C). Both measurements indicate a stable particle that produces a robust effect on the MR signal. This was further reflected in the minimal loss of Gd(III) from the nanoparticle construct ( Figure 1C). At 4 °C, the cumulative loss of Gd(III) was <0.5%, although a greater loss (<5%) was seen at 37 °C over 17 days. These particles can therefore be safely stored for at least 2 weeks without affecting their stability or relaxivity during shipment at room temperature (21 °C).
As a further quality control procedure, after shipment of particles, relaxivity of nanoparticles in solution was measured again on a 9.4 T MRI at 21 °C and contrasted with commercially available Gd-HPDO3A (ProHance). Both R1 and M A N U S C R I P T 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65   23   R2 were measured for different Gd(III) concentrations to afford a direct comparison as to potential benefits of constructing a gold nanoparticle to improve relaxivity. Indeed, the improvement in relaxivity was clearly evident especially at higher concentrations (Figure 2A). This results in a higher molar relaxivity for the DNA-Gd@AuNPs compared to Gd-HPDO3A, with r 1 values of 6.68 and 3.01 mM -1 s -1 respectively ( Figure 2B). r 2 is also higher, at 53.5 compared to 4.11 mM -1 s -1 ( Figure 2C). It is important to note that this improvement in relaxivity is for each Gd(III) molecule, rather than the polymeric AuNP construct. The r 1 for each DNA-Gd@AuNP is 2548 mM -1 s -1 , a very dramatic ~800 fold improvement compared to monomeric Gd-HPDO3A ( Figure 2D). Nanoparticle complexion therefore improves the relaxation properties of Gd-HPDO3A by increasing the relaxivity of each Gd(III) molecule, as well as by complexion into a polymeric construct that can be used to efficiently label cells.

Labeling of neural stem cells
In order to achieve the most sensitive detection of implanted NSCs, the intracellular concentration of Gd(III) must be maximized without any detrimental effects on cells. Oligonucleotides facilitate an efficient cellular uptake with a dose-dependent uptake being evident with increasing amount of NPs. This was evident by DNA-Gd@AuNPs being visible inside NSCs using the Cy3 moiety at 24 hours ( Figure 3A). Even at 0.02 nM there was clear evidence of DNA-Gd@AuNPs uptake ( Figure 3B), but at 20 nM a far more significant amount of DNA-Gd@AuNPs were present intracellularly ( Figure 3C). Quantification of M A N U S C R I P T 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65 24 uptake was performed using relative Cy3 fluorescence, as well as an absolute measurement using ICP-MS for Gd(III). Both showed the same pattern of a dosedependent uptake ( Figure 3D). A 20 nM incubation resulted in >3 times the uptake of 2 nM, and ~25 times the uptake of 0.02 nM incubation, yielding an intracellular Gd(III) concentration of ~55 μM. A persistent concern in the use of nanotechnology is the potential variability between batches of nanoparticles.

A C C E P T E D ACCEPTED MANUSCRIPT
However, three separately synthesized batches of DNA-Gd@AuNPs resulted in a highly reproducible and consistent uptake of NPs into cells ( Figure 3E). As intracellular Gd(III) concentration is determined by DNA-Gd@AuNPs uptake, batches with lower Gd(III) loading result in lower intracellular concentrations ( Figure 3F). Quality control of DNA-Gd@AuNPs loading for each batch is hence important to ensure an optimal cellular concentration of Gd(III).

Detecting labeled cells using MRI
To assess whether the achieved intracellular concentration of Gd(III) was sufficient for visualization of labeled cells using MRI, cells incubated with 20 nM nanoparticles were imaged at 9.4 T and contrasted with cells incubated with an equivalent concentration of Gd-HPDO3A ( Figure 5A). The R1 for DNA-Gd@AuNP labeled cells was significantly (1.5x) higher than unlabeled cells (p<0.01) and Gd-HPDO3A-labeled cells (p<0.05) ( Figure 5B). Gd-HPDO3A labeling did not result in a significant increase in R1 compared to unlabeled cells.

M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT
Nevertheless, in vivo there is no easily defined ROI for detection and hence a smaller voxel volume is the only option to avoid partial volume effects and detect labeled cells. A clear T1 effect was observed here, although some attenuation was evident due to intracellular incorporation with r 1 relaxivity of DNA-Gd@AuNPs within cells being reduced to 3.87 mM -1 s -1 ( Figure 6D) compared to 6.68 mM -1 s -1 in solution (i.e. a 42% decrease). 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65   27 To establish intracerebral detection of transplanted cells, cell number and voxel size were arrayed and T1 maps were generated to measure signal intensity changes due to labeled cells ( Figure 7A) To achieve a robust in vivo detection, ideally a signal change of at least 25% is required. Using the acquired array, a contour map was generated for the experimental space ( Figure 7B). It is evident that there is some interaction between voxel size and cell number on the T1 signal change. The aim of a 25% change in T1 for a proof-of-principle study can be targeted with a cell injection of  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65   28 to non-labeled cells. To verify that the signal change on the T 1 -weighted images was indeed due to the transplanted cells, immunohistochemical analyses were performed to detect human cells (HNA in green), as well as the Cy3 moiety of the DNA-Gd@AuNPs. Although human NSCs were present in both hemispheres, only those containing DNA-Gd@AuNP were detectable using the T 1 -weighted MR images. A higher magnification image shows the center of a graft, where colocalization of DNA-Gd@AuNP based on the Cy3 moiety can be seen in human cells ( Figure 8B). Aggregates of gold nanoparticles can seen as black spots. In order to assess whether the label may have leaked from transplanted cells, the level of co-localization between HNA and Cy3 was quantified ( Figure 8C).

M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT
The majority of cells (71%) were correctly identified based on the presence of DNA-Gd@AuNP, but there was a significant (29%) portion of false negatives (Type II error). However, there was very little transfer of agent with 0.38% host cells containing DNA-Gd@AuNP (Type I error). DNA-Gd@AuNP therefore afford the in vivo reliable detection of transplanted cells using T 1weighted MR imaging with little risk of a Type I error (<5%), but significantly underestimating the total number of transplanted cells (Type II error >20%).

Discussion
Achieving an unequivocal in vivo detection of transplanted cells remains a major impediment to understand how the distribution of therapeutic cells influences recovery. We here characterized the use of a new class of MR contrast agents M A N U S C R I P T 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65   29 for cellular MRI that uses gold nanoparticles as a platform to assemble multiple components that afford visualization using different modalities, but with added functionality, such as improved cellular uptake. These DNA-Gd@AuNP were efficiently incorporated into cells without the use of a transfection agent or electroporation to produce a high intracellular yield of Gd(III) that preserved T1 relaxivity. At concentrations up to 20 nM nanoparticles, no significant effects on survival, proliferation or differentiation were evident. Batch-to-batch variability was insignificant with good stability and retention of T1 relaxivity. Optimization of MR parameters indicated that voxel size significantly affects the degree of contrast, with smaller voxels providing favorable conditions to detect small numbers of cells. This approach allowed us visualize transplanted cells using T1 contrast in vivo with verification by immunohistochemistry indicating a high level of specificity of the contrast (i.e. low Type I error), although there is a significant underestimation of transplanted cells using this approach (Type II error).

DNA-Gd@AuNP nanoparticle design facilitates T1 detection
The use of gold nanoparticles as platform to assemble multi-functional andmodal nanoconstructs provides several opportunities that improve on the polymeric conjugation of Gd(III)-complexes [42]. Most notably, gold-based constructs can provide a unique opportunity to tune T 1 contrast to achieve optimal relaxivity of each monomer [24,28]. Grafting Gd(III) onto DNA-Au nanoparticles here almost doubles the relaxivity of each Gd-HPDO3A chelate demonstrating a key improvement in performance compared to merely creating a M A N U S C R I P T 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65   30 polymeric compound. It is of note that shape of particles is increasingly emerging as an important design consideration in providing improvements in contrast [34] that potentially will further improve relaxivity yield of each Gd(III)-chelate, but might eventually also allow tuning to multiple independent constructs being visualized using the same contrast mechanism [43]. Conjugation of gadolinium chelates onto the gold nanoparticles further affords an additional capping preventing Gd(III) leaching, hence increasing their stability [44]. High thermostability is essential to avoid ill-effects of Gd(III) ions on cellular health that could compromise long-term detection of transplanted cells [45]. A further key performance characteristic of a putative cellular MRI contrast agent is its reproducibility in manufacturing, as well as relaxivity performance [4]. Indeed, the DNA-Gd@AuNP nanoparticles here provided very consistent performance in physical characteristics, cell uptake, as well as relaxivity further supporting their potential for cellular imaging studies.

A C C E P T E D ACCEPTED MANUSCRIPT
Cell uptake into non-phagocytic cells, such as NSCs, remains a major challenge to yield a sufficient intracellular concentration that affords T1 detection, but also to prevent a quenching of the T1 effect. Since the sensitivity of T1 based MR probes is inherently lower than T2 agents [46], a higher concentration is required for detection. Nonspecific endocytosis or pinocytosis are the most commonly used routes of agent uptake due to their simplicity and efficiency. This typically results in localization of the agent to the endosomes/lysosome pathway, leading to a quenching of the T1 effect above 10 10 Gd/cell [12] and potentially a rapid clearance through exocytosis [47,48]. Instead of quenching of T2 relaxivity, M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65   31 lysosome-entrapped agent creates susceptibility effects that favor T2 detection [49] and can be used to track cells [45,50]. Indeed, the R2 of DNA-Gd@AuNP is exchange (τ m ) and tumbling rate (τ r ) for a given field [51]. As these molecular parameters are highly sensitive to protein binding and water access, relaxivity in solution and inside cells is significantly different [49].
Here, cell incorporation led to a 42% loss of r 1  Gd@AuNP in solution and provide a dramatic improvement in cell detection [34].

Specificity of cell detection
Apart from the intracellular environment influencing the relaxivity of DNA-Gd@AuNP relaxivity, detection of cells is equally dependent on the imaging volume within which these cells are embedded. Partial volume effects are much discussed in the context of diffusion weighted MRI [52], but little attention has been given to them in the context of cell tracking [2,4]. To assess efficacy of potential cell transplant therapies and correlate cell presence and migration with clinically relevant outcomes (such as motor coordination) large animal numbers are required per study [53]. This necessitates shorter imaging times per animal in order to increase throughput. Lower resolution images are significantly faster to acquire and have a better SNR, but are likely to result in a partial volume effect that leads to a significant underestimation of the regional distribution of cells. By arraying voxel size and cell number against T1 signal change, a 12 nL voxel size (229 μm isotropic) here provided the optimal intersection of these parameters for in vivo imaging. Although these arrays indicate which parameters provide appropriate imaging conditions, it does not account for what spatial resolution would be required to yield biologically important information. Unfortunately, at present it remains unknown what information would be relevant biologically to determine a relationship between the distribution of transplanted cells and their therapeutic efficacy. Voxel size must therefore be based on a balance between 33 acceptable cell detection thresholds, imaging time constraints, as well as the biological relevance of the information gained from imaging.
Since there is normally very little contrast in a T 1 -weighted image of the brain, T1 contrast agents are able to generate a very specific signal that is less likely to be confused with naturally occurring phenomena than T2 based agents [4]. However the question of agent specificity to the cells remains an issue. Any agent present outside transplanted cells, whether from the injectate solution or after leakage from labeled cells, will be detected as contrast. Bimodal agents allow validation of MRI contrast using histology, as shown here and elsewhere [41,54,55], in order to assess the specificity of cell detection. One area of particular concern is the fate of the label after cell death, as lingering label could produce erroneous contrast [2,4]. However, there were very few false positives, consistent with reports [16,56] indicating that at least in the short-term there is little risk to overestimate the implanted cells (i.e. Type I error is <.05%). In contrast, there is a significant number of false negatives, approximately 30% which is higher than what is typically acceptable for a Type II error (20%).
Although 100% of cells were labeled prior to injection with a slow proliferation activity in NSCs, the exact source of the underestimation remains unclear. It is conceivable that the fluorescent moiety is detaching and producing an underestimation of labeling using this method of verification. However, there is minimal loss of Gd(III) complexes over two weeks, hence it is unlikely that the fluorescent moiety is detaching this rapidly. Gold nanoparticles are known to 34 eventually undergo a degradation process that leads to their clearance [57], but they have an excellent stability profile [26].

Conclusions
We have shown that DNA-Gd@AuNPs have several advantages for the visualization of cell transplants in the brain. Specifically, these nanoconjugates provide an efficient cellular uptake of large quantities of Gd(III) into nonphagocytic NSCs, while preserving a T1 contrast inside cells that affords an robust in vivo detection using T 1 -weighted MR images. Conjugation of Gd(III) to Au nanoparticles was essential to improve T1 relaxivity of Gd(III) molecules that was attenuated upon intracellular uptake. Further engineering of particles is desirable to potentially further reduce this intracellular attenuation of relaxivity, hence producing an even more efficient contrast agent for cell tracking. Longterm studies with larger group size are nevertheless required to demonstrate the utility of these agents to assess cell survival, as well as graft distribution. Using DNA-Gd@AuNPs offers new opportunities to potentially visualize transplanted cells unequivocally using T1 contrast and use cellular MRI as a tool to derive biologically relevant information that allows us to understand how the survival and location of implanted cells determines therapeutic efficacy.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65 37 M A N U S C R I P T