Okay, so we've covered a number of issues. Now lets look at a systematic way to choose the data collection parameters for a tomography project. The first step is to select a resolution target. For instance five nanometers, or four nanometers. Or do you intend to use sub tomogram averaging and get them to two nanometers, or even one nanometer or higher? Then choose a magnification that will give you a pixel size three to four times smaller than the target resolution. Don't go higher than that, because it limits your field of view, but have at least that high of a magnification. Next, record and inspect several dose series. In other words, put your sample in the microscope, and just image it over and over and over and over and watch it and see when it starts bubbling. Choose a target dose as half of the dose which first produces visible damage, or even less if you're targeting higher resolution. Now, when you do this, remember that the energy filter may be removing electrons. So be careful that you are measuring the total incident dose to your sample, not the dose that's actually emerging through the energy filter and hitting your detector. Next, choose a tilt increment that will give you a resolution just better than your target resolution. Again, you don't want to go to a tilt increment even smaller than that because, then you have to record a lot more images, it'll be slower, you'll have less dose per image. So you don't want to go to extremely low, unnecessarily low tilt increments. But you do have to choose a tilt increment low enough to support your target resolution. Then to a first approximation, the dose per image is just the actual total dose you want to use for the whole tilt series, which you figured out in this point above. That total dose divided by the number of images you're going to include. Now I say that to a first approximation that's true. In fact, you'll want to increase the dose with increasing tilt angle. So it turns out that the sample is the thinnest of course when it's un-tilted. As you tilt the sample to higher and higher angle, the path that the electrons have to go through gets thicker, longer, and so more and more will be scattered inelastically, and removed by the energy filter. In order to maintain a constant number of electrons in each image pass the energy filter. That means you'll need to use higher doses for the higher tilt angles. And there's different ways to increase that dose. One of them is called, for instance, the 1/cos Scheme. And that just means that the total dose per image scales as one over the cosine of the tilt angle. So as the tilt angle increases so does the dose being invested or used to record that image. Another one is called the Exponential Scheme where the dose being used rises as an exponential function of tilt angle. There's another one called the Saxton Scheme. And the differences between these are beyond the scope of this series. And the differences between these are beyond the scope of our discussion today. But never the less you will want to choose one of these schemes and then choose a dose per image. So that the total dose invested through the tilt series, is that target dose that you found by recording example dose series of your sample. Next, you choose an exposure time and a beam intensity that will deliver the dose you want per image, in as short as time as possible, to reduce the drift. But that still separates individual electron hits if you're using a direct detector. In other words, we can increase the beam intensity and if we do, we can deliver the dose per image we want in a shorter time and that's good because it'll reduce drift during the image. But it's bad if you're using a direct detector because you might increase the dose rate so high, that the detector can't detect individual electron hits. And so there has to be a compromise between those two factors. Finally, if you're going to low pass filter everything past the first zero of the CTF, choose a de-focus who's first zero is beyond your resolution target. If you're going to CTF correct, then choose a de-focus where the spacings of interest are in a CTF maximum, so that they'll be clearly there. The last issue we'll address is that of Handedness. Now just as your right hand is related to your left hand because they are mirror images, so too biological objects have hand. Even individual molecules for instance, there are enantiomers of each other. Anytime you have a chiral center, you can have either of two configurations around that chiral center, and this produces enantiomers. And so, all of the biological objects of interest have hand. Alpha-Helices have a particular hand for instance. And if you're looking at an entire cell, sometimes there are features who's hand matters inside. Now unfortunately this can be tricky in tomography, so you have to pay attention to a number of factors. So these are factors in tomography that influence handedness. First of all, there are conventions for the tilt angles. When you first insert a sample into the microscope column, who's to say whether this way is a positive tilt angle, or that way is a positive tilt angle. So conventions have been established for this. Another one is that the images are rotated as they pass down through the microscope. We talked about this, how electron lenses cause rotation of the images. So as the image comes down from the column as the electron wave passes through the sample, it's going to rotate any number of times before it comes down and forms an image on your detector. And so, the images are rotating as they come down a column, and one must pay attention to how much they're rotating. Or else they can flip more than 180 degrees, and confuse the hand of the reconstruction. Another issue is that, cut. Another issue is that, if you're looking at digital images presented on a computer screen, or stored in a particular image format. There have to be conventions about whether those images are being viewed as were from above or are they being viewed from below. And there just simply has to be convention about that. Finally, there are conventions in the reconstruction software. About how you take those images and build them into a three dimensional reconstruction. Finally, after you've produced the 3D reconstruction and you view it in some kind of software, that visualization software has conventions in it of how it reads the 3D reconstruction and how it presents it to the viewer. And if you're looking at slices, slice by slice in a viewer, and you think you're moving, say up in Z, you have to know up means. Is it coming through the reconstruction this way or does up mean it's moving through the reconstruction that way? And so one way to handle this challenge is to establish clear conventions for all of these different steps and make sure that they are followed consistently. This can become challenging though because you may have multiple users on a particular microscope that you may change the camera some day. You may change the magnification and you have different rotations of the images coming down the column. You may potentially change which software you're using to record the tilt series in the first place. You may change the software that you use to do the reconstruction. And so, a fail safe method is to embed in your sample some kind of molecular handedness standard. So now I'm gonna talk about DNA origami, gold nanoparticle helices which can be used for this purpose. It turns out that you can design DNA sequences to bind to each other in interesting patterns. And these can be custom-made to form really complex patterns, even letters and words. And one of our colleagues designed a DNA origami self assembly to form a tube with linkers coming out of that tube with either a right handed helical path, or a left handed helical path. And these linkers could be tethered to gold nanoparticles. So these objects are called DNA origami, gold nanoparticle helices. And they can just be purchased now, and you can just add these to your sample before you do the tomography. And here's and example where we've done that. This is a slice out of a tomogram of a cell. And in this cell. Cut. In this sample were added DNA origami gold nanoparticle helices. And so here are the gold beads from one of those that you can clearly see in the reconstruction. Now, the gold has much higher contrast than anything else in the sample. And so, with a simple ISO surface, you can visualize these gold beads in 3D space. And that's what's being shown here. And if we start at the top, the bright golds are close to you. They're in the foreground. And then, as you go down this series, the dark golds are far from you in the background, and so then here they come back into the foreground. So you can see, if you follow this from one side to the other, that this is a left-handed helix. This is another way to view it. Here, the bright beads are in the foreground. This is the same helix. And as you go around this helix you see that they get darker, showing you that this is, in fact, a left-handed helix. So if you add left handed DNA origami gold nanoparticle helices, then you can go through the whole tomography process, you can produce reconstruction and you can visualize it. And ask yourself simply the way I'm visualizing these result, are the helices, the same hand as the ones that I put into the sample originally. And if they are, you can trust the handedness of all the objects in the tomogram. If, however, they're reversed, then you know that some step along the way, the handedness got reversed and you need to flip it.
