|Home | About | Journals | Submit | Contact Us | Français|
Traditional studies of neuroanatomical connections require injection of tracer compounds into living brains, then histology of the post-mortem tissue. Here we describe and validate a new compound that reveals neuronal connections in vivo, using MRI. The classic anatomical tracer CTB (cholera-toxin subunit-B) was conjugated with a gadolinium-chelate to form GdDOTA-CTB. GdDOTA-CTB was injected into the primary somatosensory cortex (S1) or the olfactory pathway of rats. High-resolution MR images were collected at a range of time points at 11.7T and 7T. The transported GdDOTA-CTB was visible for at least 1 month post-injection, clearing within two months. Control injections of non-conjugated GdDOTA into S1 were not transported, and cleared within 1–2 days. Control injections of Gd-Albumin were not transported either, clearing within 7 days. These MR results were verified by classic immunohistochemical staining for CTB, in the same animals. The GdDOTA-CTB neuronal transport was target-specific, monosynaptic, stable for several weeks, and reproducible.
Defining the anatomical connections of the brain is crucial for understanding its functions. In addition to revealing normal circuitry, studies of anatomical connections can show rewiring and outgrowth during development, or degeneration following brain injury.
To reveal anatomical circuits, conventional approaches require injection of tracers in vivo, followed by sacrifice after a specific and limited survival time, followed by histological processing of the ex vivo tissue. This approach requires a large sample size in order to compare the results across animals, and it is not suitable for chronic or longitudinal experiments.
Recent developments in MR neuroimaging have begun to reveal information on brain connections in vivo. Diffusion tensor imaging (DTI) has been widely used to study major white matter bundles (‘tractography’) in humans (LeBihan et al., 2001). However, even when based on sophisticated data acquisition and mathematical calculations, DTI estimates large fiber tracts rather than revealing connections per se. Therefore it cannot reveal whether or not one brain region is connected with another, nor can it determine whether the fiber tract projects in the efferent or the afferent direction (Tuch et al., 2005; Owen et al., 2007; Fonteijn et al., 2008). Moreover, DTI performs poorly in regions where fibers merge or diverge (Mukherjee et al., 2008; Peled et al., 2006; Ciccarelli et al., 2008), or when fibers turn sharply (Wedeen et al., 2008). Finally, diseased or aged brains often have altered DTI parameters that can affect the tractography results (Clark et al., 2001; Beaulieu, 2002; Salat et al., 2005; Camchong et al., 2009; Brubaker et al., 2009; Bava et al., 2009).
In animals, in which invasive experiments are possible, an alternative approach is to trace connections using MRI, following injection of the contrast agent manganese chloride. This manganese-enhanced MRI (MEMRI) tracing approach can reveal multisynaptic circuits, and it has found widespread use in a number of animal models, including rodents, birds, and non-human primates (Pautler et al., 1998; Saleem et al., 2002; Van der Linden et al., 2002; Wu et al., 2006; Simmons et al., 2008; Chuang and Koretsky, 2009). However the interpretation of manganese transport and anatomy can be complicated because manganese is transported multisynaptically. This uncertainty has been partially overcome using precise timing to define the numbers of synaptic steps along a given axonal pathway (Tucciarone et al., 2009; Chuang and Koretsky, 2009). However, the uptake and transynaptic transport of manganese can reflect neuronal activity (Lin et al., 1997; Aoki et al., 2002; Yu et al., 2005; Silva et al., 2008; Eschenko et al., 2010a), which can further complicate interpretations of the anatomical projections. Additional complications arise from a non-neuronal systemic diffusion of manganese through the CSF or blood stream (Chuang and Koretsky, 2009). Also, manganese is toxic above a specific dose (Wu et al., 2006; Simmons et al., 2008; Eschenko et al., 2010a,b).
Here, our goal was to develop and test a new contrast agent for MRI-based anatomical tracing that reveals monosynaptic connections between specific brain areas, easily and reliably. The new compound is a conventional tracer (cholera-toxin subunit-B; CTB), made visible for MRI by conjugation with Gadolinium-chelates (GdDOTA, gadolinium-tetraazacyclododecanetetraacetic acid). Gadolinium-chelates are widely used as MRI contrast agents in clinical studies (for reviews, see Griaf and Steiner, 1986; Sosnovik, 2008; Port et al., 2008). Combining these two properties, we hypothesized that the CTB component of the conjugate would be taken up by neurons and transported axonally, and that the GdDOTA component of the conjugate would reveal the location of the transported compound based on MRI.
We tested this compound in the well-defined circuitry of the rat brain. In the first experiment, GdDOTA-CTB was injected into primary somatosensory cortex (S1) to test for local MR signal enhancement in well-known thalamic targets of S1: the ventral posterolateral thalamic nucleus (VPL), posterior thalamic nuclear group (Po), and reticular thalamic nucleus (Rt) (Koralek et al., 1988, Kaas and Ebner, 1998; Liu and Jones, 1999; Paxinos, 2004). MRI scans were performed at systematically varied time points to measure the neuronal uptake and transport dynamics of GdDOTA-CTB. In a second experiment, we validated the above results by comparison with the immunohistochemical staining of CTB in the same animals that previously received GdDOTA-CTB injections and MRI. In addition, we evaluated the extent of possible tissue disruption at the GdDOTA-CTB injection sites using histology. Third, we demonstrated additional MR enhancement in sites expected in/near the injection site, including the gray matter, the underlying white matter and local intrinsic connections. In addition, we also found patchy enhancement in the caudate/putaman (CPu) in the regions known to have connections with S1 (Gerfen, 1989; Kincaid and Wilson, 1996; Hoover et al., 2003). The second and third experiments also investigated the direction of transport, and whether or not the GdDOTA-CTB traces connections monosynaptically vs. multisynaptically. In a fourth experiment, we compared the intra-neuronal transport rate of GdDOTA-CTB with the extracellular diffusion rate of GdDOTA alone, by injecting the two compounds into comparable locations in S1. To confirm that the tract-tracing properties of GdDOTA-CTB are mediated by active uptake and axonal transport mechanisms, we also performed control experiments using Gd-Albumin, a gadolinium-conjugated serum protein. This compound has a molecular weight comparable to that of GdDOTA-CTB, but without any known tract-tracing properties (Nagaraja et al., 2006; Astary et al., 2010). Fifth, we compared the transport properties of GdDOTA-CTB with those of the MEMRI, in an otherwise-matched experiment. Finally, we investigated whether GdDOTA-CTB can reveal neuronal tracts in other regions of the brain, by testing it in the olfactory pathway of rats.
Following unilateral GdDOTA-CTB injections into S1, we found target-specific enhancement in the main thalamic nuclei known to be connected with S1, namely VPL, Po and Rt (Fig. 1). This presumptive transport was observed when using multiple types of MR sequences: 2D and 3D T1-weighted (T1-W) and 3D T1- inversion recovery (T1-IR) (see Methods and Supplemental Information).
Depending on the MR sequence used, different brain regions (e.g. white matter vs. gray matter) were often enhanced differentially in the raw MR images, because each type of brain tissue has a different relaxation time. Fig. 2 shows such image variations across a rostrocaudal series through the thalamus (Fig. 2A, B), and the subtraction analysis (e.g. Fig. 2C), which separated the experimentally-induced MR enhancements from this intrinsic background variation. Figs. 2A–C show results after extensive signal averaging. Fig. 2A was acquired during a 14-hr scan using the T1-IR sequence, which yielded the highest image contrast; this was the single ex vivo experiment that we performed. Images in Figs. 2B and C show the average from 9 scans over 3 scan sessions, from a single in vivo case. At a threshold of p < 0.002 (uncorrected), the subtraction images (Fig. 2C) confirmed enhanced MR signals (presumptive transport) in thalamic targets VPL, Po, and VM (i.e. the ventromedial thalamic nucleus), consistent with known connections (Koralek et al., 1988, Kaas and Ebner, 1998; Liu and Jones, 1999; Paxinos, 2004; MacLeod and James, 1984; Desbois and Villanueva, 2001). Additional enhancement was apparent in the raw images (e.g. Rt, in Fig. 2B) but it did not reach statistical significance at p < 0.002, given this level of signal averaging. The lack of significance in Rt (Fig. 2C) may also reflect the small size of the nucleus, relative to the limits of brain co-registration processes.
A second, simpler strategy for isolating enhancement was to measure MR levels in mirror-symmetric locations in each hemisphere from a common slice, then to use the contralateral hemisphere as a control for that in the injected hemisphere. For example, Fig. 3 shows enhancements ipsilateral to the S1 injection site in 4 slices centered on VPL, based on both T1-W (Fig. 3A, B) and T1-IR (Fig. 3C, D) sequences. In Figs. 3A and B, the slice planes included putative Rt. In the T1-W images, enhancement in VPL was typically 10–20%. As expected, the background suppression sequence (T1-IR) yielded higher contrast enhancement; in VPL, this amounted to 70–90%.
Our subsequent analyses focused on VPL, because VPL is the largest of S1’s thalamic-recipient nuclei, and it includes somatotopic map variations large enough to be resolved with MRI. Of the 24 animals injected with GdDOTA-CTB into the forepaw region of S1, all showed MR enhancements in the corresponding forepaw representation of VPL.
To resolve the time course of this presumptive transport, we re-scanned animals at a range of time points following the GdDOTA-CTB injections: days 1–7, 1 week, 3 weeks, 4 weeks, and 8 weeks. Fig. 4 shows the level of MR enhancement over time in VPL, in group-averaged data (n=8). The mean signal remained near baseline through day 2 post-injection. In this dataset, the signal increase became statistically significant on day 5 (p = 0.034), and reached a plateau near day 7, approximately 10% above baseline in these T1-W images. The enhancement remained stable for at least 4 weeks post-injection, and then returned to baseline levels by 8 weeks. The time course of the peak MR enhancement (Fig. 4) was consistent with the survival times required for optimal CTB transport in conventional histological studies (e.g. Ericson and Blomqvist, 1988; Bruce and Grofova, 1992; Sakai et al., 1998; Angelucci et al., 1996).
To verify the thalamic targets of the MR results, CTB immunohistochemical staining was conducted in animals that had received GdDOTA-CTB injections into S1 followed by MRI scans.
The connections of S1 with VPL are known to be reciprocal: S1 projects to VPL, and S1 receives projections from VPL. In contrast, S1 connections with Rt are unidirectional: S1 projects to Rt, but does not receive projections from Rt (Kaas and Ebner, 1998; Liu and Jones, 1999; see also reviews Alitto and Usrey, 2003, Jones, 2007). If the GdDOTA-CTB were operating as a classic neuronal tracer, injections of this compound into S1 should confirm these and related predictions based on known anatomical features revealed by the CTB histology. For instance: a) CTB injections into S1 should label cell bodies and presynaptic terminals in VPL, as localized by the MRI in the same animals; b) such CTB-labeled cell bodies should be absent in thalamic regions immediately surrounding VPL, since those regions do not project to S1; c) presynaptic terminals (from S1) should be labeled by CTB in Rt; d) all the CTB labeling should be confined to the ipsilateral thalamus; and e) CTB labeling should be confined to the somatotopic subfield of VPL that corresponds to the injected region in S1 (i.e. the forepaw representation of VPL).
To test these predictions, brain slices from the thalamus and S1 were stained using standard immunohistochemical procedures (Bruce and Grofova, 1992; Angelucci et al., 1996; Sakai et al., 1998, 2000; Wu and Kaas, 2000), from the same animals in which MR images had been collected (see Fig. 5). The locations and boundaries of VPL and Rt were localized independently, based on known cytoarchitectonic differences between thalamic nuclei (for review, see Jones 2007, Paxinos, 2004, see also Fig. 1B, CO-stained brain section).
All the above predictions were confirmed: a) CTB-containing cell bodies and terminals were found within VPL (Fig. 5B–D); b) such CTB-labeled cell bodies were absent in thalamic regions surrounding VPL (see Fig. 5D); c) Rt showed the typical ‘dusty’ appearance of labeled presynaptic terminals (Fig. 5C), relative to the non-specific background staining (e.g. Bruce and Grofova, 1992; Sakai et al., 2000); d) all labeling was confined to the ipsilateral thalamus; and e) the label in VPL was confined to the somatotopically appropriate segment (i.e. the crescent moon-shaped subfield of VPL, in Fig. 5A–B), consistent with the highly topographic patterns of anatomical connections known from previous studies (Fabri and Burton, 1991a, Hoover et al., 2003; see Jones, 2007 for review).
Although CTB was not toxic in previous studies (see Ishitsuka and Kobayashi, 2008), it could be argued that transport properties of the GdDOTA-CTB were complicated by increased osmolarity or chemical toxicity, at much higher concentration at the injection site. To test this, we did a control experiment of injecting comparable volumes (1 l) of saline or GdDOTA-CTB (50%) into S1 in 4 additional animals, followed by sacrifice 5 days post-injection. Subsequent histology of the injection sites revealed that GdDOTA-CTB injections produced tissue disruption comparable to that in the saline control (see Fig. S1). Thus, the GdDOTA-CTB was not obviously toxic at the injection sites, at the present concentration.
The S1 injections also produced MR enhancement in the most dorsolateral region of the caudate/putamen (CPu) (Fig. S2). This region is known to receive direct inputs from the forepaw representation of S1 (Hoover et al., 2003) and to show forepaw responses physiologically (West et al., 1990; Brown, 1992; Brown and Sharp, 1995). The MR enhancement was discontinuous and restricted to patches, approximately 200–400 m in diameter. The size and location of these patches suggests that the GdDOTA-CTB projects into striasomes, as reported based on conventional tracers and immunocytochemical staining (Graybiel and Moratalla, 1989; Gerfen, 1989; Schoen and Grabiel, 1993; Kincaid and Wilson, 1996; Hoover et al., 2003).
As early as 4–5 days after GdDOTA-CTB injections, enhancement could be clearly detected in the white matter just beneath the injection sites. Such enhancement could be traced along the rostrocaudal direction in the horizontal plane (Fig. S3A), and along the mediolateral direction in the coronal plane (Fig. S3B–G). Note that white matter enhancement only appeared after a few days post-injection. Thus, presumably the enhancements resulted from active transport in the white matter tract, rather than from contamination of the white matter at the time of injection. Immunohistochemical staining confirmed the presence of CTB-labeled axons in the corresponding location of the white matter (Fig. S3C compare to Fig. S3D). The MR enhancement could be seen both in the raw images (Fig. S3F) and in the quantitative subtraction (Fig. S3G) from the same animal.
In some cases, we found that cortical injections of GdDOTA-CTB produced a band of horizontally-oriented, elongated enhancement in the middle layer(s) of cortex, running parallel to the brain surface (arrows in Figs. 6A–C). This result was especially prominent when the injection core involved the superficial cortical layers. This evidence suggests intrinsic transport of the GdDOTA-CTB, a common finding in studies using classic neural tracers.
To test this interpretation, Fig. 6 shows the CTB immunohistochemical staining at higher magnification (Fig. 6D–F). Clearly, CTB-labeled pyramidal neurons were found in layers II, III and V, with axonal processes and extensive apical dendrites. It is well known that the apical dendrites from the cortical pyramidal cells extend to (plexiform) layer I (for review, see Cauller, 1995; Rubio-Garrido, 2009; see also Fig. 2 in Wang et al., 2009). Thus our results suggest that the bands of MR enhancement within S1 reflect (at least in part) the labeled pyramidal neurons in layers II, III and V, with their apical dendrites extending to the superficial cortical layers.
Injections into the forepaw representation of S1 also enhanced MR signal in the adjacent M1 cortex (Fig. 6C). Based on the location previously reported from microstimulation mapping experiments and the standard brain atlas, the enhancement we observed was restricted to the forepaw representation of M1 (Donoghue and Wise, 1982; Neafsey et al., 1986; Pixinos and Watson, 2004). This is consistent with the extensive and topographically organized inter-cortical connections between S1 and M1 described in earlier studies (Akers and Killackey, 1978; Donoghue and Parham, 1983; Fabri and Burton, 1991b; Colechio and Alloway, 2009; Izraeli and Porter, 1995). The M1 enhancements took the form of a thick band-like pattern concentrated in the middle part of the cortex. Judging from its laminar location and thickness, and our CTB histology, this band-like enhancement appears to include layers III through V.
In many cortical systems, one would expect to find callosal connections to corresponding cortical areas in the contralateral hemisphere. For instance, such ‘homotypical’ callosal connections have been demonstrated between S1 representations of the jaw, whisker barrels, and midline body. However evidence from classical tracers (Akers and Killackey, 1978; Killackey and Fleming, 1985; Iwamura 2000; Hayama and Ogawa, 1997; Manzoni et al., 1989) has shown that such callosal connections are extremely weak or absent between S1 forepaw representation (i.e., the site injected here). Therefore our finding a lack of callosal enhancement is consistent with previous negative results.
Supplemental Fig. S4 shows the time course of the MR enhancements at the injection site throughout the week following S1 injections. Prominent signal increases were observed immediately (1–2 hrs) after the injections of GdDOTA-CTB. In the injection site cores, the enhancement was relatively weaker, presumably reflecting the known shortening of T2 signals at high gadolinium concentrations (Fig. S4A, 1–2 hrs). After day 3, the enhancement in the injection cores contracted slightly, and the borders were sharpened - but otherwise the size and shape of the injection site remained stable for a month thereafter (Fig. 7A, D, day 5 and week 3; Fig. S4A, day 5 and 7; Fig. S6A, C, 90 hr and 170 hr). Group statistical measurements (n=11) confirmed that the level of signal enhancement in the injection sites also reached asymptote on day 3, and remained relatively stable for a month after the injection. Enhancement at the injection site was largely cleared by 2 months post-injection (Fig. S4C, red line), similar to the time course of the transported compound in VPL (Fig. 4B).
Thus, in both the injection site and the transport targets, the time course of GdDOTA-CTB enhancement was fully consistent with neuronal uptake of the tracer CTB; presumably extracellular diffusion of GdDOTA alone would clear much faster. To confirm the latter, we directly measured the rate of extracellular diffusion by injecting GdDOTA alone into S1 (n=4). GdDOTA injections immediately produced a strong signal enhancement throughout a large region of the cortex, as one would expect from rapid extracellular diffusion (Fig. S4B, blue line in S4C). Moreover, the enhancement due to GdDOTA alone cleared extremely rapidly: it peaked at the first data point, immediately after injection, and it cleared completely within 24 hours. Thus the GdDOTA was completely cleared well before the enhancement due to neuronal transport (i.e., from the GdDOTA-CTB) peaked (days 5–7, cf. red vs. blue lines in Fig. S4C).
An additional control experiment was designed to further rule out the possibility that GdDOTA-CTB transport can be mediated by non-selective, passive uptake and diffusion. To test that hypothesis, we injected another contrast compound, Gd-Albumin, in which the gadolinium was conjugated with bovine serum albumin (a protein that has a molecular weight similar to CTB). Although the injected gadolinium concentration from Gd-Albumin was comparable to the GdDOTA-CTB (i.e. 65–75 mM), and the protein concentration from Gd-Albumin was also very high (30% protein), we found that the signal intensity at the Gd-Albumin injection sites was much weaker (compare Fig. S4 red line and Fig. S5 solid black line) and the enhancement was much smaller, indicating that Gd-Albumin was rapidly cleared immediately after injection. The signal intensity at the injection core continued to decrease on day 4, returning to baseline values by day 7 (Fig. S5). Except for the focal injection cores, no enhancement was present anywhere in the brain, including the thalamus, at any MR imaging time point.
Overall, these data further support the conclusion that the GdDOTA-CTB conjugate is actively and selectively taken up and transported within the brain, as a MR-visible anatomical tracer.
To clarify the relative advantages or disadvantages of GdDOTA-CTB, we compared GdDOTA-CTB results with comparable data using the tract-tracing contrast agent manganese. For both tracers, the signal intensity was measured at both the S1 injection site and the thalamic transport site, at comparable time points across tracers (i.e. immediately post-injection, at and after peak transport times for the two tracers). The two contrast agents were injected at the same concentration and volume.
In the S1 injection site, these experiments confirmed that the GdDOTA-CTB was strikingly punctate, and spatially stable throughout the following weeks, in each of the major cortical layers (Fig. S6A, C). By contrast, comparable injections of manganese spread widely and quickly (Fig. S6B, D).
The increased spatial specificity of the GdDOTA-CTB injections was evident in both horizontal (i.e. along the dorsolateral surface of the cortex) and vertical (i.e. across cortical layers) directions. As shown in Fig. S6A and C, the pattern of signal enhancement at 50 hrs after injection was near-identical to that measured 170 hrs after injection, with a half-amplitude at half-maximum (HAHM) of 1–1.8 mm, measured away from the midline, in all layers. This remarkable spatial specificity was found in all animals studied (Fig. 7A left panels, and D top middle panel). As expected, the center of the injection core was not enhanced, reflecting signal dropout due to the T2 shortening effect at high concentrations of the contrast agent.
In comparison, enhancement due to manganese at the injection sites spread rapidly, in both axes. As early as 1–2 hrs after injection (i.e. the earliest possible data acquisition point), manganese enhancement at the injection site (involving the supragranular layers) was quite extensive (HAHM = 4–6 mm) in the horizontal direction within the supragranular layers, nearly triple that of the GdDOTA-CTB extent (Fig. S6, top panels of B and D, Fig. 7B, top left panel). By 10 hrs post-injection, the MR enhancement spanned all cortical layers in the vertical dimension, and also increased in the horizontal dimension (Fig. S6, lower panels of B and D). Existing evidence suggests that these rapid changes in manganese enhancement at the injection site reflect a combination of diffusion and continued uptake. The diffusion may be mediated via the CSF, at least in part (Liu et al., 2004; Chuang and Koretsky, 2009), due to the small molecular weight of the manganese. Therefore, the growing size of the injection site likely reflects manganese transport by the neurons at the site of diffusion, followed by further diffusion, uptake and transport, and so on.
Similar differences were also observed in the thalamic transport sites, in comparisons between these two tracers. After a relatively short period of time, manganese enhancement appeared in multiple subfields and nuclei, including some that are not confirmed by classical neuroanatomical tracer data. As early as 12 hrs following manganese injection into S1, regions such as the subthalamic nuclei and sustantia nigra are prominently enhanced (data not shown, but see Tucciarone and Koretsky, 2009, Fig. 2A) – even though these regions do not have direct connections with S1 (Fujimoto and Kita, 1993; see also Paxinos, 2004 for review).
We found that once GdDOTA-CTB is transported to its target zones, the enhancements remain at the same location and at a constant size (Fig. 7A, right panels, Fig. 7C, middle panel, and Fig. 7D lower middle panel). In addition, the pattern of GdDOTA-CTB enhancement in the transport zones is patchy, consistent with the pattern obtained from conventional tracers, which indicate that only specific subfields of the nucleus are connected with the injection site (Fig. 7C middle panel). By contrast, manganese enhancement at the transport zones, within a few hours after injection, rapidly spread into neighboring regions, including different subfields of the same nucleus, and even into different nuclei (Fig. 7B right panels, Fig. 7C, lower panel, and Fig. 7D lower right panel).
Thus, compared with the GdDOTA-CTB, it is more difficult to use manganese to reveal the precise zones that are directly connected with the injection site, if transport results are not timed precisely. This can be especially challenging if transport is faster for some targets (e.g. closer targets) compared to others (farther from the injection site). In such cases, perhaps no single transport time is optimal.
To test for transport in a peripheral neural pathway, we injected GdDOTA-CTB unilaterally into the nostril cavity (n=2). Strong signal enhancement was observed in the olfactory epithelium, exclusively ipsilateral to the injection, as early as 12 hrs following the injection (i.e., the second MRI time point). By day 2, robust enhancement was clearly detected throughout the olfactory epithelium and along the olfactory tract ipsilateral to the injection (Fig. 8A). Weaker enhancement was also found in the outer layer of the inferior olfactory bulb (OB, i.e. the glomeruli layer). Some individual glomeruli in the specific region of the OB could be easily identified based on the MRI enhancement patterns (Fig. 8B). Enhancement in these regions lasted up to 7 days. The injection of GdDOTA-CTB into one nostril did not enhance signal in the contralateral nostril pathway, consistent with the known anatomical evidence (for review, see Imai and Sakano, 2008; also see Kikuta et al, 2008, Fig. 1). Together, these results suggest that GdDOTA-CTB can be used to trace peripheral anatomical pathways, in addition to central ones.
Following unilateral injection of the OB, MR signal enhancement was found on day 7 in other regions of the OB, and in part of the ipsilateral anterior olfactory nucleus (AON; Fig. 8C). Weaker enhancement could also be detected in the ipsilateral projection of the central olfactory pathway to pyriform cortex (Fig. 8C). The location and pattern of GdDOTA-CTB transport is consistent with known olfactory pathways using conventional tracers (Smithson et al., 1989).
To our knowledge, this study is the first to demonstrate brain connections in vivo, using a purpose-designed compound combining a classic neuroanatomical tracer (here, cholera-toxin subunit-B, CTB) with a known MRI-visible label (gadolinium-chelate, GdDOTA). When injected into S1 of rats, this compound produced a distinct enhancement of MR signal in multiple known thalamic targets of S1, including VPL, Rt, Po, and VM, in the hemisphere ipsilateral to the injection site (see Figs. 1C, ,2).2). Moreover, the presumptive transport was somatotopically specific: injections aimed in the forepaw representation of S1 produced MR enhancement in the middle subfield of VPL, which corresponds to the forepaw representation in that thalamic nucleus (see Figs. 1C, ,2,2, ,3,3, 5A–B, ,7C7C middle panel).
This MR evidence for neural transport was confirmed by histology. Histological staining showed definitive CTB transport in the same thalamic nuclei that showed enhancement in the MR images, within the same animals, in the expected cellular compartments. For example, cell bodies and terminals were labeled in VPL, whereas only terminals were labeled in Rt (Fig. 5B–D).
Outside the thalamus, the GdDOTA-CTB also showed additional MRI properties consistent with those known from classic tracers. This evidence included stable and long-lasting enhancement of MRI at the injection site, laminar-specific intrinsic connections near the injection site, connections with ipsilateral striatum and M1, and white matter projections beneath the injection site.
Crucially, the time course of the thalamic MR enhancement is consistent with the interpretation of axonal transport of the GdDOTA-CTB compound. That MR enhancement began in the thalamic targets only after 2–3 days, and the enhancement peaked from 1–4 weeks post-injection (see Fig. 4B).
To the extent that it is known, histological evidence on CTB transport matches the time course of the presumptive transport of GdDOTA-CTB into thalamus, based on MRI. For axonal distances comparable to those in this study, CTB transport can first be detected 3–4 days following injection, and 7–14 days yield optimal results (Bruce and Grofova, 1992; Erison and Blomovist, 1988; Angulucci et al., 1996; Sakai et al., 1998). This similarity in time courses strongly supports our hypothesis that the MR signal enhancement in thalamus reflects active neuronal transport of GdDOTA-CTB to/from S1.
By comparison, MR enhancement due to passive extracellular diffusion (from GdDOTA injections into S1) peaked and then cleared within a day (see Fig. S4B–C) – i.e., 4 days before the thalamic MR enhancement due to presumptive transport from GdDOTA-CTB reached statistical threshold. Moreover, the extracellular diffusion (GdDOTA alone) spread quite widely, unlike the specific target(s) enhanced following GdDOTA-CTB injections. Thus, the GdDOTA-CTB results were quite distinct from those due to extracellular diffusion, both temporally and spatially.
Although the GdDOTA-CTB showed strong and stable MR enhancement for long periods of time (at both the injection site and the targets), injections at similar concentrations of the control contrast compound, Gd-Albumin, cleared rapidly at the injection site - despite having a similar molecular weight. This suggests that local astrocytes and neurons take up Gd-Albumin non-specifically. Moreover, the lack of transport of the Gd-Albumin here and previously (Nagaraja et al., 2006; Astary et al., 2010) indicates that the GdDOTA-CTB works successfully as a unique MRI-visible tract-tracer, based on active uptake and transport processes.
Using conventional T1-W MR sequences, GdDOTA-CTB produced a thalamic enhancement of 10–20% above the background MR level. A more targeted background-suppression T1-IR MR sequence yielded much higher signal increases (~ 80%). However the exact level of statistical sensitivity of this technique will vary widely depending on multiple technical factors. For instance, increases in the number of scans will increase the SNR ratio, in accord with the well-known inverse square law of signal averaging (I = 1/d2). Difference imaging (e.g. Fig. 2C) will also increase the statistical sensitivity. Difference imaging has been crucial in the fields of fMRI and optical recording, which are routinely based on significant signal variations as low as 0.1%. Thus, the current GdDOTA-CTB procedure produces signal changes that are well above the limits of statistical uncertainty.
Another crucial factor is the tracer molecule itself. For instance, the optimal ratio of Gd to CTB is not known. Results here were achieved with a ratio of 1.3–3 Gd/protein. However in a separate batch with up to 5 Gd per CTB (not described here), transport was not detected. Thus there may be an upper limit to the number of Gd that can be chelated and still yield effective CTB transport. Presumably, ratios that are too low sacrifice MRI sensitivity, whereas ratios that are too high may compromise uptake and/or transport.
The level of MR enhancement will also vary with the density of Gd reaching the target, which in turn reflects the divergence or convergence of those neural connections. Here our injections were concentrated in ~ 3–4 mm3 of S1 cortex. S1 projections converge onto, and arise from, much smaller (~ 1 mm3) thalamic targets in VPL and Po; other thalamic targets are even smaller. Thus the convergence of these connections may concentrate Gd levels in thalamus. Anatomical studies support this idea: it has been reported that connections to/from S1 are more abundant with the thalamus (up to 1:40), compared with cortical targets (Sherman and Koch, 1990). This factor may partially explain why cortico-cortical connections (e.g., from S1 to ipsilateral S2) were not apparent in our experiments, because cortical-cortical connections do not show such convergence. Technical limitations due to coil size and placement also reduced the detection of MR enhancement in S2 (see Supplemental Information).
Inevitably, insertion of an injection needle into the brain produces tissue damage along the needle track at the site of injection; it can also cause a small necrotic zone at the center of the needle tip (Fig. S1). The relationship between transport and such tissue damage has a long and complex history in the literature on classical tracers. For example, it has been reported that injections of CTB in large volumes, or rapidly by pressure, can produce damage at the injection site (Conte et al., 2009a, b). However, such tissue disruption is comparable following injections of the same amount of saline. Therefore, the important issue is whether damage occurs in the transport zones. Based on saline injection controls (Fig. S1) and histology, we found no evidence of damaged cells in the transport zones. Even after long survival times post-injection, we did not observe decays in MR enhancement, thus ruling out the possibility of secondary degeneration in these remote transport zones.
Histological studies have shown that CTB is transported in both directions along axons, both retrogradely to cell bodies, and anterogradely to presynaptic terminals (Luppi et al., 1986; Bruce and Grofova, 1992; Angeucci et al., 1996; Sakai et al., 1998, 2000; Wu and Kaas, 2000). Is our CTB-based compound also transported in both directions? The current results clarify half of this two-part question. As is typical in the brain, transport from S1 to VPL, and from S1 and Po, are known to be reciprocal. Thus in these areas, our MR enhancement does not distinguish between the two directions of transport. However connections between S1 and Rt and CPu are atypically unidirectional: S1 projects to both Rt and CPu, but neither Rt nor CPu projects back to S1 (Kaas and Ebner, 1998; Liu and Jones, 1999. Gerfen, 1989; Kincaid and Wilson, 1996; Hoover et al., 2003). In addition, it is known that sensory neurons in the olfactory epithelium project anterogradely to the OB. Thus we can conclude that the GdDOTA-CTB is transported anterogradely, at least.
Injections of manganese chloride, coupled with MR imaging (MEMRI), have also been widely used to map brain connections in vivo (Pautler et al., 1998; Saleem et al., 2002; Wu et al., 2006; Tucciarone et al., 2009; Chuang and Koretsky, 2009). However, it is complicated to interpret the relationship of MEMRI data to the density of anatomical connections. First, MRI enhancement due to manganese reflects functional (i.e. calcium-related) activity (Lin et al., 1997, Aoki et al., 2002, 2004; Yu et al., 2005; Eschenko et al., 2010a) as well as anatomical connections. Secondly, manganese may be released from tissue after uptake (unpublished observations); manganese at the injection sites spreads quickly and continuously (e.g. Figs. 7 and S6, see also Tucciarone and Koretsky, 2009). Thus MEMRI has not been used to measure connections across very small distances, such as those across cortical laminae. Thirdly, manganese is transported multisynaptically, not monosynaptically. In some experiments, this multisynaptic transport can be an advantage. However as described above, this can be a disadvantage if it is crucial to define each serial step of a given circuit. Moreover, diffusion of the manganese, coupled with the multisynaptic transport, could produce non-specific transport.
For these reasons, it can be challenging to distinguish which factor produces the observed MR enhancement in MEMRI studies. Additional measurements may be necessary. For instance, results from MEMRI have been compared to those from a classical tracer, to distinguish activity-dependent transport of manganese from anatomically based transport (Wu et al., 2006; Saleem et al., 2002).
The GdDOTA-CTB technique does not have these problems. Moreover, apparent disadvantages of the GdDOTA-CTB may be resolved by slight changes in procedure. For instance, multisynaptic connections could still be resolved using the monosynaptic transport of GdDOTA-CTB, using serial injections. For example, injections of GdDOTA-CTB into site A would produce transport to site B. Then a later MR-targeted injection into site B would produce transport to site C, and so on.
Previous studies (Enochs et al. 1993, van Everdingen et al. 1994) reported slow transport (~ 5 mm/day) of dextran-coated iron oxide compounds, which were visible using MRI. However, that compound was specifically not transported in the central nervous system (CNS), when injected into either the superior colliculus or the eye (Enochs et al., 1993). Prior to our use of GdDOTA-CTB, we also tested for CNS transport using an iron-labeled compound (biocytin conjugated with iron oxide). Consistent with the above findings, we also found that the biocytin-iron oxide compound did not produce transport, perhaps because iron-based compounds are too heavy to be transported easily in the CNS.
The in vivo MRI-based tracer approach reveals connections that would be difficult or impossible to study otherwise. However, current MRI tracers will not supplant classical tracers (e.g. HRP, CTB, WGAHRP, etc.) because the latter can distinguish labeled cells from labeled presynaptic terminals, and thus reveal retro- vs. anterograde transport. Accordingly, classical tracers remain the gold standard, when such tracers are compatible with the experimental goals.
Based on MRI, connections between specific brain areas have been inferred based on DTI (LeBihan et al., 2001, Beaulieu, 2002; Tuch et al., 2005) and correlated resting state activity in fMRI (Shmuel and Leopold, 2008; Margulies et al., 2009; Teipel et al., 2009). However neither of those non-invasive techniques can definitively show whether or not cells in a given brain region send or receive axons from another specific brain region.
Recently, invasive studies in animals have demonstrated functional connections more directly, by combining fMRI with electrical microstimulation of a targeted neural site (Tolias et al., 2005; Ekstrom et al., 2008, 2009; Moeller et al., 2008; Field et al., 2008). Although this technique raises exciting new possibilities, it has its own limitations. Anatomical connections can only be inferred, because the white matter pathways are not revealed. Moreover, the technique does not distinguish monosynaptic connections from multisynaptic ones. By comparison, the interpretation of anatomical connections using GdDOTA-CTB is simpler; it has all the properties of a classic neural tracer, without requiring animal sacrifice.
Presumably, future applications of the GdDOTA-CTB approach (perhaps in combination with other MRI-based techniques, such as DTI, resting state fMRI, and/or MEMRI) will furnish a much richer in vivo diagram of brain circuitry. At a practical level, this MR-visible tracer can be used to target anatomically connected regions, using electrophysiological microelectrodes and/or serial tracer injections. More generally, this in vivo approach can reveal changes occurring during plasticity induced by normal or abnormal physiology (e.g. axonal pruning or sprouting), in both central and peripheral nervous systems (Rakic et al., 1986; LaMantia and Rakic, 1990; Wu and Kaas, 2000, 2002; Raff et al., 2002; Madana and Esiri, 2003; Prince et al., 2009).
GdDOTA-CTB was made in-house using commercially available products. The detailed synthesis procedures are described in the Supplemental Information. Gd-Albumin was purchased commercially (Biopal Inc, Worcester, MA. Cat# P-00P01-100), and was lyophilized and re-diluted in double distilled water to a final concentration of 30% in protein and 65mM in Gd solution.
41 adult Sprague-Dawley rats (280–350 g) were used in these experiments (29 injected with GdDOTA-CTB, 4 GdDOTA injections, 3 Gd-Albumin injections, 1 saline injection, and 4 MnCl2 injections). In experiments #1–3, a total of 24 rats received unilateral intra-cortical injections of GdDOTA-CTB into area S1, and 1 received saline injection into area S1. In experiment #4, 4 rats received injections of GdDOTA in area S1 and 3 received injections of Gd-Albumin in S1. In experiment #5, we compared injections of manganese chloride (MnCl2; n=4) with GdDOTA-CTB to measure diffusion and transport dynamics in both the S1 injection site and the thalamic transport zones. In experiment #6, 4 additional rats received a unilateral injection of GdDOTA-CTB in the nostril cavity (n=2) or the OB (n=2). 1 additional rat was scanned for 13 hrs ex vivo, 7 days after an S1 injection of GdDOTA-CTB. All experiments were performed in compliance with guidelines set by the ACUC of the NINDS, NIH. Surgical details and procedures regarding the injections are given in Supplemental Information.
Animals were imaged prior to injection (‘baseline’), and immediately (1–2 hrs) after injection, on days 1–7 post-injection, and at longer intervals for up to 2 months post-injection. Four animals were imaged for 7–10 days post-injection, then sacrificed for histology.
GdDOTA alone was injected into the S1 forepaw, using the same coordinates as in the GdDOTA-CTB experiments. MRI data were acquired at the baseline, immediately after injection, and every 12 hours until day 2. The last image was acquired on day 5 post-injection.
Animals received unilateral injection of Gd-Albumin into the S1 forepaw. Animals were imaged immediately after injection and imaged on days 4 and 7 after injection. MRI data were acquired at the injection sites and throughout the brains.
MnCl2 was injected into S1 forepaw, using the same procedures described above. MRI data were acquired every 2 hrs until 10 hrs post-injection.
2D and 3D spin-echo multi-slice multi-echo (MSME) and rapid acquisition with relaxation enhancement (RARE) images pulse sequences were used to acquire T1-W MR images. The 3D modified driven equilibrium Fourier transform (MDEFT) pulse sequence was used to acquire T1-IR images. Additional details regarding 11.7T and 7T MRI data acquisition parameters and procedures are described in the Supplemental Information.
To measure the enhancement in the thalamic target zones due to GdDOTA-CTB transport, we used both region of interest (ROI) and 3D image volume substraction analyses. To measure the speed of signal decay at the injection site, we used ROI analyses. The details of these two analysis techniques are given in the Supplemental Information.
Details concerning animal perfusion, histology and photoimaging are described in the Supplemental Information.
We are grateful to Steve Dodd for pulse sequence optimization, David Yu for brain slicing, and Kathy Sharer for animal ordering and care. This work was supported in part by the NIMH and NINDS IRP, NeuroSpin/CEA, the Martinos Center for Biomedical Imaging, the NCRR, the MIND Institute, NIH grant R01 EY017081 to RBHT, and the French L’Agence Nationale de la Recherche grant ANR-09-BLAN-0061-CSD8 to CWHW.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.