Quantitative Morphology in Kidney Research
February 13-14, 2012 Conference Videos

Frontiers: Imaging Approaches to Cortical Volume and Glomerular Number
Kevin Bennett, Arizona State University

Video Transcript

1
00:00:00,000 --> 00:00:08,100
KEVIN LEMLEY: I’ like to introduce our last formal speaker of the day. Kevin Bennett’s from Arizona State where he’s Assistant Professor and

2
00:00:08,100 --> 00:00:18,133
Program Chair of Biomedical Engineering and MRI Director. He also works at Mayo in Scottsdale and he’s going to talk to us about some noninvasive

3
00:00:18,133 --> 00:00:27,399
methods which we heard people ask for over and over yesterday. So, I think we’re all looking forward to it.

4
00:00:27,400 --> 00:00:33,400
KEVIN BENNETT: Thanks for inviting me out. I knew this was a pretty hard-core nephrology group and I didn’t really know quite how hard-

5
00:00:33,400 --> 00:00:45,100
core until I had this particular picture up on my desktop and somebody came by and asked me if that was a glomerulus. So, I’m going to talk about

6
00:00:45,100 --> 00:00:54,200
MRI techniques to do morphology in the kidney and specifically I’m going to talk about the work we’ve done to develop MRI contrast agents that

7
00:00:54,200 --> 00:01:05,966
let you detect things and sort of single glomerular concentrations. I just wanted to give kind of an overview of MRI. I can’t go into a lot of detail but

8
00:01:05,966 --> 00:01:17,732
this is just kind of a big picture. So, the basic idea of MRI is that the sample, whatever it is, goes into a large magnet and the spins have a magnetic

9
00:01:17,733 --> 00:01:29,899
moment that is on all the water and they align in a certain direction and you can use external gradients to sort of put the amount of spin in one

10
00:01:29,900 --> 00:01:39,933
place and make it different than it is in another place. This is very rough; don’t take this as sort of a textbook version of MRI. You also use sort of

11
00:01:39,933 --> 00:01:49,499
RF antennas, little coils, to sort of talk to the spins in different places and you can use this to develop an image. So really, MRI is a combination

12
00:01:49,500 --> 00:01:58,000
of these big magnets as well as these pretty highly sensitive RF coils that both transmit and receive, and that will become very important

13
00:01:58,000 --> 00:02:14,733
when I talk to you about the artifacts that we get when we do renal imaging. On the left is a picture. Is this the pointer? Oh, here’s the pointer.

14
00:02:14,733 --> 00:02:22,433
Thank you. Right here is a picture of a typical, small animal MRI scanner, in fact that’s the one that we typically use for our animals, and this is

15
00:02:22,433 --> 00:02:30,199
very similar to the one that you would see in a clinic except the bore size is smaller and the field strength is usually quite a bit higher. The animal

16
00:02:30,200 --> 00:02:41,300
system is a 7 Tesla MRI, a typical clinical field strength would be something like 1.5T to 3 Tesla, so it’s quite a bit stronger and what that does is it

17
00:02:41,300 --> 00:02:48,166
gives you more magnetization and therefore more signal, okay? So, the higher field you get, the more signal, but right now really high field

18
00:02:48,166 --> 00:02:59,832
strengths aren’t cleared for human use, so a lot of this really high field stuff is done in animals. One of the other things that we have—and a few

19
00:02:59,833 --> 00:03:10,133
groups have this—is a super-high field system. So, this is a vertical bore. Has anybody ever done NMR in here? A few people? One or two?

20
00:03:10,133 --> 00:03:16,699
So, this is what you would see. If you go into an NMR lab it would typically look like this. It’s a vertical bore and this is exactly the same, again,

21
00:03:16,700 --> 00:03:24,900
as all the other MRI systems you see, except it’s vertical. And so, you put the sample in usually from the bottom—you could also top load—and it

22
00:03:24,900 --> 00:03:33,933
usually goes into a little tube. So, this is how we image excised things like kidneys and we’ve also done some in vivo work, and I’ll show you a

23
00:03:33,933 --> 00:03:45,599
couple of pictures at the end. MRI is very powerful and there’s been quite a bit of work in imaging the kidney both in structure and function.

24
00:03:45,600 --> 00:03:54,500
This is just a picture from a paper in 2008 where they did…I think this is in a pig kidney…just to show you what typical images would look like,

25
00:03:54,500 --> 00:04:02,733
and you can do this in three dimensions. One of the issues we’ve been talking about is morphology. You don’t really get a big contrast

26
00:04:02,733 --> 00:04:10,966
difference between, say, a glomeruli and the rest of the tissue just from baseline MRI, and so you can see that it looks fairly homogenous. You can

27
00:04:10,966 --> 00:04:17,899
tell the difference between the cortex and the medulla and that’s used to do morphometry in that way, and you can do all kinds of imaging post-

28
00:04:17,900 --> 00:04:25,633
processing tricks to pull these things out and to make volume measurements. So, this has been done quite a bit and there’s actually a literature of

29
00:04:25,633 --> 00:04:30,233
people that have been doing functional measurements in the kidney, looking at flow and that kind of thing, but that’s not what I’m going to

30
00:04:30,233 --> 00:04:39,199
talk about today. Today I’m going to be talking about molecular imaging techniques with MRI. If you think about molecular imaging, if you talk to

31
00:04:39,200 --> 00:04:46,666
somebody that works in a company, usually they don’t think about MRI and there’s a couple of reasons for that. If you take a couple of very,

32
00:04:46,666 --> 00:04:54,866
very sensitive molecular imaging techniques, optical microscopy and PET/SPECT, some of the advantages of these things are that they’re both

33
00:04:54,866 --> 00:05:04,866
super-sensitive and they’re highly flexible. For instance, light microscopy is pretty well developed, so it’s kind of the guide for molecular

34
00:05:04,866 --> 00:05:07,999
imaging technique. The problem with light microscopy, as you all know, is the light penetration isn’t very deep, so if you’re thinking

35
00:05:08,000 --> 00:05:15,366
about doing whole 3D imaging, it doesn’t work that well and similarly with PET/SPECT, it has some disadvantages in resolution. It’s super-

36
00:05:15,366 --> 00:05:26,599
sensitive—you can detect single molecules—but you can’t resolve very far. If you compare that to MRI you can get these exquisite pictures with

37
00:05:26,600 --> 00:05:34,566
very high resolution. In fact, I’ll show you some that we’ve taken at 35 micron isotropic and we’re getting even lower than that, down to 10 microns

38
00:05:34,566 --> 00:05:40,599
isotropic. So, you get these beautiful pictures and one of the other big advantages of MRI is the contrast is so flexible. So, these are things that

39
00:05:40,600 --> 00:05:51,766
make it pretty attractive. The problem is it’s not a very sensitive technique. We’ve developed a contrast agent that you can use to detect single

40
00:05:51,766 --> 00:06:00,666
glomeruli and I’ll tell you a little bit about the history of this. The agent is cationized ferritin. Ferritin, of course, is an iron storage protein, and by

41
00:06:00,666 --> 00:06:11,566
cationized it means we just stick amine groups on the surface of it. This idea for MRI, actually, we’re doing it from this work at the NIH when I was

42
00:06:11,566 --> 00:06:19,499
actually in the NINDS and I was using this as something to cross-link and we were studying the effects of particle cross-linking on the MRI

43
00:06:19,500 --> 00:06:25,200
signal, and I was buying this stuff because it was easy to functionalize—it was already functionalized, you could just cross-link things to

44
00:06:25,200 --> 00:06:34,433
it—so it was easy to use. So after I had been using it for a while I started asking, “Why would they even sell this stuff?” It turns out there’s a

45
00:06:34,433 --> 00:06:41,933
long literature of electron microscopy using cationized ferritin; it binds to the basement membrane and you guys probably all know this.

46
00:06:41,933 --> 00:06:49,766
These are some of the original images. This was published in a paper in 2008 basically showing that you could do intravenous injection of cationic

47
00:06:49,766 --> 00:06:56,832
ferritin. You can detect the labeling of the cationic ferritin in the glomerular basement membrane—you can see the ribbon—and you can see it with

48
00:06:56,833 --> 00:07:03,633
electron microscopy, too. If it’s anionic—the native ferritin—you don’t see it, okay? So, it’s very specific to just the basement membrane, at

49
00:07:03,633 --> 00:07:13,999
least in the kidney. That’s not entirely true, but I’ll talk about that in a second. So, how does this work and why would it be a potential MRI

50
00:07:14,000 --> 00:07:26,400
contrast agent? There are quite a few different types of MRI contrast agents. One very useful one is called superparamagnetic, so it’s basically

51
00:07:26,400 --> 00:07:37,200
a magnetic particle, and what it does is it creates a perturbation in the MRI field. So water, which is what you actually detect with MRI, diffuses

52
00:07:37,200 --> 00:07:46,300
around this and while it diffuses around, its spin accumulates phase and what that does is create this phase dispersion and a change in the signal

53
00:07:46,300 --> 00:07:52,700
locally where they are, and so if you have enough of them you can get a bigger change. So, the amount of change in the MRI signal at that

54
00:07:52,700 --> 00:08:02,000
place that you get is dictated or it’s kind of described by this equation where it relates the concentration of the agent and then the T2

55
00:08:02,000 --> 00:08:12,100
background—this is a background signal—to the total T2, so the more agent you have, the higher concentration, the bigger the change in T2 you

56
00:08:12,100 --> 00:08:19,666
get; it also depends on sort of the background T2 values. So, T2 is just a relaxation time that we use to measure and it creates a darkening in most

57
00:08:19,666 --> 00:08:33,599
of our pulse sequences, so it makes a dark spot in the image. Ferritin, as I said, is an iron storage protein. In its native form it actually has a

58
00:08:33,600 --> 00:08:37,100
superparamagnetic—somewhat superparamagnetic—core. It’s partially disorganized, that makes it sort of diamagnetic

59
00:08:37,100 --> 00:08:41,033
but there is part of it that becomes superparamagnetic, so it does have an MRI relaxivity; it acts like a contrast agent, so in high

60
00:08:41,033 --> 00:08:48,566
enough concentrations you can see it. This is just a picture of the particles. The ones that we use you could actually functionalize off the surface.

61
00:08:48,566 --> 00:08:59,099
Some of the advantages of this are: it’s potentially nontoxic, I’ll show you some of the toxicity studies we’ve been doing; and it’s also

62
00:08:59,100 --> 00:09:01,833
very, very small, and so that’s why it actually gets into the basement membrane in the kidney. Another big advantage is we can use

63
00:09:01,833 --> 00:09:14,199
fluorescence or immunohistochemistry to see where it is, so it’s a pretty cool agent. So, this is the first experiments that were described in the

64
00:09:14,200 --> 00:09:21,633
original paper where we basically just did an intravenous injection of native ferritin and cationic ferritin, and these are the MRI images over on the

65
00:09:21,633 --> 00:09:36,399
right and you can see that the normal kidneys that were perfused were totally clear in the cortex. When you have native ferritin there’s no

66
00:09:36,400 --> 00:09:39,833
accumulation, it looks very similar, and then when you have cationic ferritin you see the dots and these correspond to individual glomeruli. So, I

67
00:09:39,833 --> 00:09:44,599
should mention that all this work at this point, when we saw this change we started working with Rob Star’s group at NIDDK, so this is a

68
00:09:44,600 --> 00:09:54,733
collaboration between NINDS and NIDDK and Xiaodu, who’s actually here, was a great collaborator at that time. At this time I had no idea

69
00:09:54,733 --> 00:10:03,466
what a kidney was but this was interesting and so we started thinking about what we could do with it. So recently, we’ve taken this a little bit

70
00:10:03,466 --> 00:10:11,532
further and have tried to look at whether we could be more quantitative about the labeling, and so this is our paper in AJP Renal that came out

71
00:10:11,533 --> 00:10:20,766
this last year, basically showing the same effect, but these were all done at 19 Tesla with the microimaging system, and you can actually see

72
00:10:20,766 --> 00:10:27,099
this in really good resolution and we’ve started to ask: can we actually measure total number of glomeruli? Can we measure the volume in the

73
00:10:27,100 --> 00:10:32,900
glomeruli inside the kidney? And so, what I’m going to talk about today is really: what are some of the pitfalls to this and can we do it? These are

74
00:10:32,900 --> 00:10:40,233
actually 3D…these are taken from a 3D data set in MRI, so we’ve got it all the way through the kidney, and you can acquire these in

75
00:10:40,233 --> 00:10:50,633
roughly…so, at this field strength you can actually still get a good measurement of glomeruli within 2-4 hours; these were taken at 12 hours

76
00:10:50,633 --> 00:11:02,033
just to get a really good signal, but you can dial things in. Then you can do things like this where you can make a big map of all the glomeruli in the

77
00:11:02,033 --> 00:11:07,499
kidney and then you can start to ask things like: what’s the intensity of labeling in different parts of the kidney? What are the apparent sizes of the

78
00:11:07,500 --> 00:11:18,000
glomeruli that you’ve labeled? Things like that. So at this point, after the MRI is done, we work with our image processing people, in this case Dr.

79
00:11:18,000 --> 00:11:27,500
Teresa Wu, who is at Arizona State as well, to do the 3D rendering. We can compare the cortical volumes in these 3D data sets to total numbers of

80
00:11:27,500 --> 00:11:44,766
glomeruli in size, etc. And then we’ve been working with John Bertram at Monash to basically validate the MRI data, and in the paper in AJP

81
00:11:44,766 --> 00:11:53,366
Renal that came out we basically did the 3D imaging and then sent the kidneys to him for the stereology—the disector-fractionator method—so

82
00:11:53,366 --> 00:12:03,932
we were able to compare the techniques directly. We also have done acid maceration and then counted with a hemocytometer just as a second

83
00:12:03,933 --> 00:12:11,399
measurement. So, from the MRI data you get this big 3D data set and there’s quite a bit of post-processing that goes into it. I’m not going to go

84
00:12:11,400 --> 00:12:17,600
into too much detail; if somebody wants to talk about it, we can. It’s an interesting problem and I think it’s similar to the questions you guys were

85
00:12:17,600 --> 00:12:25,400
asking about, “How do you make really good measurements from stereology?” I think once this gets more developed we’ll start having the same

86
00:12:25,400 --> 00:12:32,133
arguments about: what’s the best way to process the data and what’s the best way to threshold? But the basic idea is that you want to

87
00:12:32,133 --> 00:12:38,433
put some kind of global threshold on the image and get rid of everything except for the labeled glomeruli and then somehow tell what’s a

88
00:12:38,433 --> 00:12:46,199
glomerulus versus an artifact, so it’s basically just a lot of other problems that we have to deal with; there’s susceptibility artifacts and that kind of

89
00:12:46,200 --> 00:12:55,533
thing. I’ll show you some examples in a mouse kidney, specifically. So, these are the results of that study. Again, these are the labeled glomeruli.

90
00:12:55,533 --> 00:13:06,199
This is just a 2D projection out of the 3D data set and then these colored circles are the glomeruli that we counted from this counting algorithm, and

91
00:13:06,200 --> 00:13:15,600
we were able to both pull out the total number of glomeruli, we compared that to…this was the stereological count on the left. Let’s see, this is

92
00:13:15,600 --> 00:13:26,333
stereology, this is…oh, sorry. This is stereology, this is MRI, and this is the acid maceration technique, so we were in the range of stereology

93
00:13:26,333 --> 00:13:34,599
and acid maceration slightly underestimated, but we’re in the same ballpark, and then similarly with glomerular volume we were able to get roughly

94
00:13:34,600 --> 00:13:43,566
the same numbers as the stereology, but this is in the entire kidney. One of the other interesting things that we and the group in Heidelberg were

95
00:13:43,566 --> 00:13:54,132
also able to do was to create a distribution of the apparent glomerular volumes. I’m not sure that this has been able to be done before, at least not

96
00:13:54,133 --> 00:14:01,466
for every single glomerulus in the kidney and it brings up an interesting question because, when I say apparent glomerular volume I say that

97
00:14:01,466 --> 00:14:07,699
because there’s a little bit of a susceptibility artifact inside the glomerulus and it doesn’t necessarily correspond to the size of the

98
00:14:07,700 --> 00:14:15,433
glomerulus. It could be intensity of labeling, it could actually extrude a little bit from the glomerulus, so it’s not clear that it’s exactly a

99
00:14:15,433 --> 00:14:20,899
direct measurement of size, and that’s something that we haven’t really ironed out yet. So, when we talk about this we’re going to be very careful;

100
00:14:20,900 --> 00:14:30,600
this is just an apparent measurement but it is based on the size of this little dot. We’ve also extended this in the mouse; I’ll show you how

101
00:14:30,600 --> 00:14:38,633
we’ve done the processing. So, this is just a thresholded image. This is a typical mouse kidney on our 19 Tesla MRI and you can actually see the

102
00:14:38,633 --> 00:14:48,533
labeling in the glomeruli there as well, and it’s harder. So, this question came up yesterday about what’s holding this kind of technique back

103
00:14:48,533 --> 00:14:57,233
from being clinically useful. This is one of the technical issues. So to do this, we had to use a 19 Tesla MRI system which is a super-high

104
00:14:57,233 --> 00:15:03,833
field—that gives you a lot of field and a lot of signal. The other thing is we have really powerful, what are called field gradients and

105
00:15:03,833 --> 00:15:11,533
they let you manipulate the field, the local magnetic field, and what that does is let you image over a much smaller area so it lets you get

106
00:15:11,533 --> 00:15:19,599
the resolution way up. Most clinical systems don’t have gradients like this that will let you do this on anything like this order. That’s not to say it

107
00:15:19,600 --> 00:15:26,366
couldn’t be developed, in fact, we use them all the time for animal imaging, it’s just something that is a technological issue. The other thing is it gets

108
00:15:26,366 --> 00:15:32,866
harder to make coils that good when they get bigger. So, I think the take-home message is the RF coil, the RF technology—we were talking

109
00:15:32,866 --> 00:15:39,766
about this earlier—RF technology is going to be really important as this gets developed because that’s where most of the artifacts are coming

110
00:15:39,766 --> 00:15:53,332
from and most of the technological breakthroughs have to happen. I told you that field strength is important, so this is kind of an example of what

111
00:15:53,333 --> 00:16:02,533
you get at different field strengths. These are just…these are all from published data except for this one. This is soon-to-be-published data, but

112
00:16:02,533 --> 00:16:10,933
basically showing the same thing at different field strengths. So like I said, at something like 19 Tesla, you can get very good resolution. We’ve

113
00:16:10,933 --> 00:16:18,366
gotten it on the order of 35 microns isotropic and you can get very good pictures of this effect and actually, you know, I was talking about the size of

114
00:16:18,366 --> 00:16:24,832
the glomerulus or the size of that field drop-off; it’s actually not a problem as far as we can tell, so it seems to correlate very well with glomerular

115
00:16:24,833 --> 00:16:37,966
size but we haven’t done it on a single glomerular basis yet. This is an image from the original paper. I intentionally made this look kind of like there are

116
00:16:37,966 --> 00:16:44,832
some RF artifacts. I mean, I didn’t manipulate it but I’ve accentuated the artifacts just to show you the kind of issues we’ve had. This is from the

117
00:16:44,833 --> 00:16:51,699
Heidelberg group, the recent publication, and this is also from our 7 Tesla, so we’ve got a range from 7 Tesla up to 19 Tesla where you can do

118
00:16:51,700 --> 00:16:58,900
this. So, you notice another issue at 7T. If I went to the same resolution, if I drove the gradients the same way, then I would not get very good-looking

119
00:16:58,900 --> 00:17:07,033
images; in fact, I would barely have any signal. So, we have to lower the resolution and what that does is it gives you a lot of what are called

120
00:17:07,033 --> 00:17:15,033
partial volume effects and it’s similar to this issue in slicing where if you’re….in histological slicing where if you slice across things, you can’t tell

121
00:17:15,033 --> 00:17:22,299
what they are; it’s very similar in MRI. We bump it and you sort of average everything around it and you get a reduction in the signal intensity. So, that

122
00:17:22,300 --> 00:17:34,733
becomes a problem, sort of, at a lower field. The other thing that comes up, though, if you look at these images you can see that it’s brighter up

123
00:17:34,733 --> 00:17:40,866
here than it is down here and there’s ways around this in post-processing, but it brings up this issue that we have a really good coil here,

124
00:17:40,866 --> 00:17:48,432
which is why this is so homogeneous, but most of our coils are not good enough to get a very clean picture. So this is what I was saying, is that

125
00:17:48,433 --> 00:18:00,133
the RF technology, especially if this is going to go in vivo, has to get developed further. This is just to show you how this works. This is a mouse

126
00:18:00,133 --> 00:18:08,833
kidney and this was all processed by our collaborator, Teresa Wu, but this is one potential way of processing the data, that you can

127
00:18:08,833 --> 00:18:15,199
calculate sort of a gradient map, which is how fast the signal’s changing locally and then you can threshold all the voxels that change that fast,

128
00:18:15,200 --> 00:18:23,533
so that’s what they did here, and then you can identify these little circles where it changed really fast and then you use an algorithm to try to

129
00:18:23,533 --> 00:18:29,966
identify what’s inside those circles and you call that a glomerulus. So, this is one potential algorithm. We’ve used several algorithms; it’s not

130
00:18:29,966 --> 00:18:38,432
clear which one’s the best. Again, we’re getting into sort of, you know, arguments over what happens when you have the signal drop off here.

131
00:18:38,433 --> 00:18:43,666
We’ve applied a local threshold to try to get rid of this so that you can get glomeruli back up but there’s other potential ways of doing it, so there’s

132
00:18:43,666 --> 00:18:55,499
lots of things you can do on the post-processing side to be more quantitative, but I think once we iron out the RF issues, some of these problems

133
00:18:55,500 --> 00:19:11,666
will disappear. The other issue is labeling dose. In this particular study, the first study, we had a total dose of 15 mg per kg of the CF and I’ll have

134
00:19:11,666 --> 00:19:21,799
to talk about this. I’ll tell you about this, you know, how we’re also getting to lower doses in a few slides, but we’ve given up to three doses of this

135
00:19:21,800 --> 00:19:31,333
for the ex vivo study and up to five doses in vivo and you can that between zero labeling, two doses and three doses, you get a definite

136
00:19:31,333 --> 00:19:38,866
increase in intensity of the labeling with three doses, so this is another issue that we have to figure out the optimal dose for the specific

137
00:19:38,866 --> 00:19:46,899
contrast agent. But when you think about it, it also opens up an opportunity because we’ve started to look at what’s the difference in

138
00:19:46,900 --> 00:19:54,100
intensity between two and three injections and there’s some interesting thoughts about mathematical modeling of glomerular function

139
00:19:54,100 --> 00:20:06,400
when that happens. So, this is an example of what we can pull out from this. These up here are maps of the intensity of the glomerular

140
00:20:06,400 --> 00:20:16,533
labeling through the whole kidney; these are mostly just on the surface because otherwise it would just look like a black blob. On the left here

141
00:20:16,533 --> 00:20:24,433
is after we’ve given two injections and on the right is after three injections, and you can see that after three injections you get a difference in

142
00:20:24,433 --> 00:20:34,999
both the overall intensity and also across the kidney, and if you look at histograms of intensity in what we called the cortical versus the

143
00:20:35,000 --> 00:20:42,100
juxtamedullary glomeruli you can actually start to pull out two different distributions of glomeruli in the labeling. So we’re not sure what that is yet,

144
00:20:42,100 --> 00:20:50,966
but these are potential areas where we need to focus. I’m not a nephrologist. I mean, these are places where we could definitely use some other

145
00:20:50,966 --> 00:21:03,866
expertise like you guys. The question always comes up of where we are in getting this to go in vivo. So, this was our paper in 2008 basically

146
00:21:03,866 --> 00:21:12,499
showing that you can still see the labeling in vivo. Now, there are lots of issues with in vivo imaging. One is the motion artifacts. This particular image

147
00:21:12,500 --> 00:21:23,333
we had to cardiac gate, respiratory gate, and we had a surface coil that was right here. So this is the native ferritin, and then with cationic ferritin

148
00:21:23,333 --> 00:21:30,966
you can start to see the dots. Now, these were taken at 200x200x500 microns, so these are partial volumes and I told you the partial volume

149
00:21:30,966 --> 00:21:40,832
effect before, and again, they’re gated. There’s a little bit of motion artifact but it’s not too bad, but you can start to see it, so there are a couple of

150
00:21:40,833 --> 00:21:48,933
issues with in vivo imaging. One is that you can tell that the vasculature is already dark because of the deoxyhemoglobin—this is the cause of the

151
00:21:48,933 --> 00:21:55,866
whole bold effect—so to get rid of this artifact you just crank the O(2) up and this is actually with the O(2) cranked up, so if you don’t have the

152
00:21:55,866 --> 00:22:07,766
O(2) cranked up you just see this black cortex. So, that’s one potential issue. The other one is this RF penetration issue, and so what we’re

153
00:22:07,766 --> 00:22:13,699
working on now is to see if we can build an implantable coil, at least for the short term, to get very close to the kidney, so we get this

154
00:22:13,700 --> 00:22:20,733
radiofrequency and get rid of this roll-off in RF intensity because otherwise you can’t image the whole kidney. And then again, in most of the in

155
00:22:20,733 --> 00:22:27,333
vivo imaging systems you can’t really get strong enough gradients to get a good image all the way across, so it’s not going to be at the same

156
00:22:27,333 --> 00:22:34,199
resolution as you had in vitro, at least not in a reasonable scan time, but these are technical limitations and things that could be ironed out and

157
00:22:34,200 --> 00:22:47,400
hopefully will be. The other question that always comes up is: what does this do during disease? In the 2008 paper we basically did sort of a rough

158
00:22:47,400 --> 00:23:00,133
model of FSGS with puromycin and this was actually Xiaodu’s work in developing the model. What we found was, in the normal rats as you’d

159
00:23:00,133 --> 00:23:06,733
expect—these were all age-matched, etc., litter mates, actually—in the normal rats with no puromycin injection you would actually see

160
00:23:06,733 --> 00:23:15,566
individual glomeruli, but with the puromycin—and this is two weeks later—you would start to see leakage of the CF through the glomeruli into the

161
00:23:15,566 --> 00:23:22,499
tubules, and so what we think is happening is that’s the early sort of breakdown of the basement membrane and you’re getting things

162
00:23:22,500 --> 00:23:28,766
leaking out. Now, you can see this is an artifact and you could say, “Okay, I can’t count glomeruli that way,” but I’ll show you something in a

163
00:23:28,766 --> 00:23:35,232
second that might convince you that it may not just be an artifact; it might be something that we could use to get information from. For instance, if

164
00:23:35,233 --> 00:23:43,733
it’s focal and segmental then you might be able to see just parts of the kidney that are affected, and so this is just histology, kind of confirming that we

165
00:23:43,733 --> 00:23:50,866
had the disease in the models. Okay, so this is what I was talking about. If you go back to even the 2008 paper but definitely in this more recent

166
00:23:50,866 --> 00:24:01,932
paper, we just sort of saw this. If you look here you can actually see where the tubules get labeled and, in fact, if you look at the

167
00:24:01,933 --> 00:24:07,966
fluorescence microscopy you can see it, too. Some of the CF, even in the normal rats, actually gets through and we think is labeling the proximal

168
00:24:07,966 --> 00:24:14,766
tubule and potentially farther down, so right now what we’re doing is trying to threshold these guys out. You see this very well at 19 Tesla, not

169
00:24:14,766 --> 00:24:23,266
so well at anything lower as far as we can tell, but I could be wrong about that. So, we’re trying to reconstruct the entire tubule surrounding this,

170
00:24:23,266 --> 00:24:30,432
so you have these two different thresholds and then try to map it out in 3D, but you could potentially, if this is right, you could potentially

171
00:24:30,433 --> 00:24:40,299
map this out. So, I’m just showing you the kind of things that we might be able to do; I’m not saying this is all exactly right. So like I said, there are a

172
00:24:40,300 --> 00:24:49,233
few little pitfalls that we have to avoid. One is that there’s the susceptibility artifact. We have to be careful to not say that we know exactly what

173
00:24:49,233 --> 00:24:56,266
we’re measuring; in some cases there could be some artifacts. Again, this RF in homogeneity and, in fact, the sensitivity of the RF coils is going to

174
00:24:56,266 --> 00:25:03,732
be really critical. I think there are quite a few ways that we get around it. We can develop what are called parallel coils, so you could put

175
00:25:03,733 --> 00:25:09,933
multiple coils over the surface of the kidney and that kind of thing, but when we start moving this in vivo, it’s the things that you don’t want to cut

176
00:25:09,933 --> 00:25:16,766
into, then, that might become a problem. Again, I mentioned these partial volume effects. If you don’t have really strong gradients, I see that as a

177
00:25:16,766 --> 00:25:26,599
simple technical issue, and then again this whole issue about processing that we haven’t worked out. So, this is all just stuff that’s kind of been in

178
00:25:26,600 --> 00:25:34,233
the works. The most specific one is just, to give you an example, and you’ll probably just cringe when I tell you this, we had the mice that we

179
00:25:34,233 --> 00:25:42,099
injected and we got glomerular counts and we take it to post-processing and they say we have 20,000 glomeruli per kidney. That’s obviously too

180
00:25:42,100 --> 00:25:49,900
high; it’s double what it should be. Okay, fine, we can adjust the threshold. So, this is the problem, right? And so, I mean, just to be very up-front

181
00:25:49,900 --> 00:25:54,366
about it, this is an issue. When you’re post-processing you can find anything you want to find and so then the question is: what’s real? So,

182
00:25:54,366 --> 00:26:05,199
I think that we just have to stay very clear about that. So, the question is: could we develop an algorithm that will automatically set this threshold

183
00:26:05,200 --> 00:26:11,000
and then be the same for everything? That’s the real question, right? I don’t think you’re ever going to get to the point where you’re…unless you

184
00:26:11,000 --> 00:26:19,166
have the real truth, then you’re never really going to be able to develop something where you’re absolutely sure it’s 100% right. Okay. So, toxicity.

185
00:26:19,166 --> 00:26:28,632
We always get this question. We’ve had it on every grant, every paper that we’ve submitted. We just finished a full-blown toxicity study. These

186
00:26:28,633 --> 00:26:35,966
were just with simple blood biomarkers and an immune panel and we found for, at least some basic blood biomarkers for renal and liver

187
00:26:35,966 --> 00:26:48,432
pathology, we don’t see any toxicity at a detectable dose and we don’t see any immune response, either; either chronic or acute. We’ve

188
00:26:48,433 --> 00:26:56,433
also done the biodistribution assays, at least on organs that tend to have fenestrated endothelia and you can see this is the kidney and these

189
00:26:56,433 --> 00:27:08,966
were all done at 7T. This is a kidney, a liver…you can see the liver gets labeled. I’ll come back to this in just five seconds. The lung gets labeled

190
00:27:08,966 --> 00:27:19,099
and the spleen, and exactly what this tells you is it’s still in the works. I can tell you that the

191
00:27:19,100 --> 00:27:26,366
liver…we’ve published this in an abstract in [---] and we’ve got the paper going out right now. It actually labels the Disse space in the liver. So, if

192
00:27:26,366 --> 00:27:34,299
you injected native ferritin—this is an uninjected control—if you injected native ferritin it would just be black, so the native ferritin gets taken up very

193
00:27:34,300 --> 00:27:43,466
rapidly into the macrophages, the CF actually goes into the sinusoids first and then goes into macrophages. It’s excreted both through the

194
00:27:43,466 --> 00:27:57,399
kidney and the liver. So I told you a lot of stuff. One of the things that we’re working on is to try to be able to give a lower dose. Even though it

195
00:27:57,400 --> 00:28:05,000
seems nontoxic, you want to be able to have enough to detect at a lower field strength, so what we’ve started doing is trying to make a

196
00:28:05,000 --> 00:28:12,966
better ferritin that acts as a better contrast agent. So, we’ve emptied the iron core and refilled it. You can either make a darker T2 agent, so we’ve

197
00:28:12,966 --> 00:28:19,466
done that and we’ve published that, You can make a T2 agent that makes a darker spot for less concentration; in fact, about a 10-fold lower

198
00:28:19,466 --> 00:28:26,699
concentration you can see. So, we’re still working on that to try to get it streamlined and get large amounts of it. The other thing that we’ve

199
00:28:26,700 --> 00:28:35,600
tried to do is develop what’s called a T1 agent, and there’s some advantages to a T1 agent. One of them is that you can detect it in about 10-fold

200
00:28:35,600 --> 00:28:43,500
lower concentrations because of the background, and so we’ve been working to try to turn ferritin into a T1 agent and the way we’ve

201
00:28:43,500 --> 00:28:51,000
done that is just embed manganese, which is paramagnetic, into the pores of the ferritin and it actually acts…manganese is paramagnetic,

202
00:28:51,000 --> 00:29:00,633
water interacts with it, and it turns into a bright dot instead of a dark dot, so that’s in the works. So, we’ve gotten a fairly significant increase in

203
00:29:00,633 --> 00:29:16,833
T1 relaxivity in the bright signal. One of the things we’re working very hard on is developing protocols for doing this in vivo, trying to iron out

204
00:29:16,833 --> 00:29:25,033
all the RF stuff. I’ve got a student that’s actually developing RF coils. One of the interesting things that’s been coming up is to look whether we can

205
00:29:25,033 --> 00:29:32,933
see dynamic uptake of this stuff in the glomerulus and whether that might be able to tell you something about filtration capacity. You know,

206
00:29:32,933 --> 00:29:40,466
one of the cool things about ferritin is you can produce it recombinantly, so that might make it faster for us to get this into a clinic if this pans

207
00:29:40,466 --> 00:29:52,066
out, because you can very easily load it up with iron and it looks like it’s nontoxic, so you could potentially get this through a GMP facility.

208
00:29:52,066 --> 00:29:59,199
Obviously, we’ve got to do a lot of testing and disease models and then, of course, improving RF coils that we can implant. So, these are all

209
00:29:59,200 --> 00:30:07,366
things that we’re working on. So finally, I just wanted to show you a couple of things. John forced me to put this in here just to give you an

210
00:30:07,366 --> 00:30:14,366
idea; he’s very interested in development. A potential model for renal development is the Medaka and I just wanted to show you an

211
00:30:14,366 --> 00:30:23,266
example of a non-contrast enhanced version of MRI that might be useful for developmental studies of renal disease. So, this is a Medaka—a

212
00:30:23,266 --> 00:30:35,932
fish egg—as it’s developing. This is about 800 microns across and these are all imaged at about 35 microns. This on the left is just a standard,

213
00:30:35,933 --> 00:30:42,899
sort of, gradient echo image and these are individual cells, so you can actually just image individual cells without any kind of contrast agent,

214
00:30:42,900 --> 00:30:49,000
and we can do what’s called diffusion-weighted imaging, so we can do kinds of functional imaging and stuff on developing eggs and you can do this

215
00:30:49,000 --> 00:30:57,266
all in 3D. So, these are the things you can do with it, tools that we’re developing, but again, I think one of the things that’s come out of this is that

216
00:30:57,266 --> 00:31:09,999
these sort of cross-disciplinary groups are pretty productive, so this might be a good way for people to collaborate. This is our Phoenix MRI

217
00:31:10,000 --> 00:31:16,866
group. This is the guy who’s done a lot of the work, in fact most of the work, in the last couple of years; this is Scott Beeman, my stellar Ph.D.

218
00:31:16,866 --> 00:31:27,066
student. This is my lab here and then all of our collaborators—I’d like to especially thank John for his work on the stereology—and then our funding

219
00:31:27,066 --> 00:31:50,699
sources, and hopefully the next time I talk to you I might be able to say that NIDDK funded us as well. Thank you.

220
00:31:50,700 --> 00:31:59,733
MALE: Very impressive data. Have you followed these animals, that how long this cationic proteins stay in the kidney and what is the fate of it?

221
00:31:59,733 --> 00:32:09,799
KEVIN BENNETT: Yeah, that’s a good point. We’ve measured out to three weeks. It looks like it’s all cleared by about one to two days out of

222
00:32:09,800 --> 00:32:19,266
the kidney. It’s still in the liver about three days but it does clear out of there at two. Sorry, go ahead.

223
00:32:19,266 --> 00:32:23,266
MALE: Does it end up in the urine? Can you detect it in the urine?

224
00:32:23,266 --> 00:32:26,166
KEVIN BENNETT: It does. We haven’t…yeah. JOHN BERTRAM: I should know the answer to

225
00:32:26,166 --> 00:32:30,466
this, but the in vivo imaging there you showed, is that in your 7T or your 19T?

226
00:32:30,466 --> 00:32:32,599
KEVIN BENNETT: That was actually at 11.7. JOHN BERTRAM: Oh, okay. Thanks.

227
00:32:32,600 --> 00:32:45,000
MALE: I don’t know much about MRI but I seem to recall hearing that the high field in human, at least when you get up here, it actually has physiologic

228
00:32:45,000 --> 00:32:55,300
effects. So, have you got any possibility on this being used and is it something you could localize your field to the relevant area or do you think

229
00:32:55,300 --> 00:33:04,433
maybe it would have, at high enough field strength to get good imaging, you might run into some physiologic consequences?

230
00:33:04,433 --> 00:33:12,033
KEVIN BENNETT: That’s a good point. I think, you know, if we can do it at 7T—7T is approved—we can do that in humans and the biggest problems

231
00:33:12,033 --> 00:33:22,666
at 7T right now are the RF susceptibility artifacts, RF shimming is an issue, and potentially heat because the higher field, the more heat, the more

232
00:33:22,666 --> 00:33:30,566
power you dump with RF, but those all have very strict limits, so we stay within that. So, I think at 7T we’re safe. If we get past that in humans, I

233
00:33:30,566 --> 00:33:34,699
don’t know how long it will take to get that approved.

234
00:33:34,700 --> 00:33:43,300
MALE: Nice presentation. I was just wondering, you’ve made the point that you’d like to get the RF coil as close as possible to the kidney and you’re

235
00:33:43,300 --> 00:33:54,633
implanting it in the animal model. Obviously, that’s not really feasible in humans. Is it even physiologically…not physiologically…is it

236
00:33:54,633 --> 00:34:03,166
technically possible to put a coil close enough to a human kidney, especially with the 6-12 inches of tissue in between you may have to even

237
00:34:03,166 --> 00:34:09,266
generate these kind of images? KEVIN BENNETT: At the moment, it’s really hard.

238
00:34:09,266 --> 00:34:23,466
Whether it will ever be possible, I don’t know, but some of it is just, you know, some are looking at it from sort of RF power issues. I mean, the more

239
00:34:23,466 --> 00:34:29,599
power you dump in, the more heat, and you have to have a trade-off between that and the amount of signal you get. So, that will become an issue

240
00:34:29,600 --> 00:34:41,800
with high-field human systems. So, I’m not sure if you’re ever going to be able to do 3D, whole kidney, single glomerular measurements, but you

241
00:34:41,800 --> 00:34:49,233
will definitely be able to do partial volume, sort of, estimates from parts of the kidney.

242
00:34:49,233 --> 00:34:55,966
ROGER WIGGINS: Wiggins, Ann Arbor. You probably can do it in human with a kidney transplant recipient, you know, a thin one. You

243
00:34:55,966 --> 00:35:00,199
can get the magnet really close. KEVIN BENNETT: Yeah. That’s a great point.

244
00:35:00,200 --> 00:35:04,000
Thank you. That’s something that we’ve been pretty interested in.




Date Last Updated: 10/4/2012

General Inquiries may be addressed to:
Office of Communications and Public Liaison
NIDDK, NIH
Building 31, Rm 9A06
31 Center Drive, MSC 2560
Bethesda, MD 20892-2560
USA
Phone: 301.496.3583