Designing stress studies, part 2: how do you get your sample?

I recently posted about how to choose what bodily substance to use to test for cortisol in a stress study: blood, saliva, urine, feces, or hair. Once you have your substance of choice, though, you have to actually extract it from the dog. This can present more or fewer challenges, you know, depending.

Blood

When people first started measuring cortisol, they used blood to do it. Blood is where cortisol shows up first. All the other substances that we measure cortisol in have had their cortisol levels compared to blood cortisol levels, to make sure that they correlate strongly. Researchers had to do studies to prove that these other substances worked for this measurement, which cost a lot of effort and money. They did this because blood is pretty hard to get hold of, in most cases. Sticking a needle in a dog will usually stress it out, and it's hard to get the blood extracted before the stress of the restraint starts changing the blood cortisol levels.

But even aside from that, sometimes a blood draw is simply out of the question. For my Master’s work, I had to cold-call hospital clients and convince them to let me enroll their dog (already in the hospital for some procedure or other, in other words, already having a bad day) in my study. If I had told them that the dog would need a blood draw too, I guarantee that most of them would have said no.

In a comment on the previous post in this series, Tegan pointed out that animals can be trained to submit calmly to blood draws. For some studies, this approach would be invaluable. For my study, again, it wouldn’t have worked. Training an animal to accept a needle is an arduous process, and I had access to those dogs once, on one night. For most shelter dog studies, this would also be an impossible hurdle. But it’s a pretty cool thing to do, if you can do it.

I wish I had a video of another approach to stress-free blood draws. I have seen other vets slide a needle into the lateral saphenous vein, the vein that bulges out of the side of a dog’s hind leg just above the hock. If the dog is distracted (say by someone feeding it), a competent venipuncturist can get it done using this vein with little to no stress. I have seen this technique used in shelter dogs who would not allow restraint for a more traditional draw. But it takes a dog with short, smooth fur and a particularly lovely bulgey vein. It does not work in little dogs. And it definitely requires a competent person to do the draw. After a few years of practice in blood draws, I was just getting to the point during my internship where I could do this one. There can’t be too much poking around to find the vein, or the game is up.

(I did find a video of a technician drawing from the lateral saphenous of a dog who is lying on his side, with an assistant holding off. This is the same vein as the one I am talking about, but in the procedure I’ve seen, the dog can be standing and you actually don’t need someone else to hold off the vein. You come at the vein from above, not below, in a standing dog. Just in case any of you blood-drawers out there want to try this yourself.)

Since blood was such a pain to get, people started trying other substances, figuring anything had to be easier than a blood draw.

Saliva

Saliva is now used much more often than blood in human cortisol studies. You hand a person a cup and they drool into it. No needles, no added stress. Dogs are not so easy. You can’t ask a dog to drool into a cup; you have to get the drool out yourself.

For my study, I used Sorbettes, also known as eye sponges. The instructions say to put one Sorbette into the dog’s mouth for 30-60 seconds, and voila, it has enough saliva on it for an assay. You then put the Sorbette into a tube and spin the tube in a centrifuge to get the saliva out. You only need 25µg, which is hardly anything! What could go wrong.

Sorbettes

First of all, when you are analyzing the saliva later on, you use 25µg per well in the plate of saliva samples, and you get one cortisol value per well. But it turns out that the assay is fairly imprecise, and gets it wrong a decent percent of the time, sometimes close to 10% of the time. So it makes sense to use two wells per sample (now we are at 50 µg per dog). This way, if you get two very different answers for your two wells, you know that the assay went wrong and not to use one of the samples. Wait, which sample is good and which sample is bad? To avoid that problem, just use three wells per sample (now 75µg per dog). Then you can throw out the bad one and keep the two good ones. I had to do this maybe 4-5 times total out of my 90-odd samples. Every time, I was really glad that I had three wells. With two wells I would have had to discard that sample (and that dog) from the study. With one well I would have included bad data in my results.

So 75µg is still not all that much saliva, but it turns out that it is enough to be pretty difficult to get, especially from dogs who are stressed out in a hospital. I used three Sorbettes and rolled them around in the dogs’ mouths for up to four minutes, at which point I had to stop in case the stress of restraint was affecting the cortisol levels. Even then, I had a lot of dry sponges. It was incredibly disheartening. In the end, we saved most of my samples by a) diluting them and changing our calculations, and b) showing the dogs cans of cat food to make them salivate.

I am currently engaged in an email exchange with other researchers who are having similar problems, particularly in small breed dogs and puppies. These days, the new tech to use to get saliva out of dogs is a small rope which the dog can chew on. I like that better than the little sponge-on-a-stick, which dogs could possibly break off and swallow (I had one come perilously close to doing just that). But even so, the problem of getting enough spit remains.

Could you give the dogs food? There is a study suggesting that cheese will not interfere with the cortisol assay, and would be safe to give. [1] It makes me nervous, though.

Could you condition the dogs to salivate when you present the little rope? This is currently under discussion, but some of us are concerned that messing around with the dog’s experience of sampling would invalidate the sample. It’s worth a small study to test it out, though, for sure. I hope someone does it.

By the way: I heard a story, which may be apocryphal, but I will repeat it anyways (and maybe someone out there can corroborate): supposedly a rhino salivary cortisol study used the procedure of collecting saliva with a very long-handled spoon. If true, it is awesome.

To come: urine, feces, and hair, oh my.

References

[1] Ligout S., Wright H., van Driel K., Gladwell F., Mills D.S. & Cooper J.J. (2010). Reliability of salivary cortisol measures in dogs in training context, Journal of Veterinary Behavior: Clinical Applications and Research, 5 (1) 49. DOI: 10.1016/j.jveb.2009.09.027

Open access dog salivary cortisol data

I finally got around to sharing the data from my study of dog salivary cortisol levels on figshare. I have meant to do this for months. Particularly, I wanted to do it so that I could wear the cool “I’m a figsharer!” t-shirt that Mark Hahnel gave me at scio13. How embarrassing would it be to wear that shirt and have someone ask what you shared and have to admit that you still haven't actually shared anything? But I am a figsharer now. So if you want numbers, go check it out.

Oh, and in case you’re interested in the associated paper, that’s here (but, sadly, not open access):

Hekman, Jessica P., Alicia Z. Karas, and Nancy A. Dreschel. “Salivary cortisol concentrations and behavior in a population of healthy dogs hospitalized for elective procedures.” Applied Animal Behaviour Science (2012). http://dx.doi.org/10.1016/j.applanim.2012.08.007

Why won't they drool?

The saga of getting spit out of dogs continues. Why is it so hard? My roommate’s dog Stella will drool enormous strings of spit whilst contemplating our dinners, but I can’t get a sufficient sample out of her when I swab her mouth. I actually convinced someone who had written a peer-reviewed article about collecting saliva from dogs to have a phone conversation with me. She provided technique suggestions. These seemed to help quite a bit for a few weeks, but then I recently went through a run with about ten dry dogs in a row. Maybe it was all just chance?

My response to the run of dry dogs has been to frantically attempt to enroll more dogs. (The time when we need to be done collecting data is rapidly approaching.) Advisor feels that this is not the right approach, though, as she has to stay late after work to perform our stress-reduction intervention (or placebo) on these dogs, and she wishes not to stay late after work until the end of time. This is understandable.

On Advisor’s advice, I have constructed a mix of nasty-smelling cat food, placed in a tin with a cap that has airholes. I display this concoction to my subjects, who may or may not find it tasty-smelling. Some will become enthused and lick the tin. Others will turn their heads away. (I think this is stress and not lack of interest in nasty-smelling food per se. These are mostly labs we’re talking about, after all.)

This evening I tested the Concoction out on a few dogs in the hospital for whom I had permission to take saliva samples. Apparently it works on at least some dogs, because when I wandered back into A Ward a few minutes after testing the Concoction out on a Rottweiler, the techs said to me, “What did you just do to that dog? After you left she drooled a whole gob of spit on the floor!” Success? Or was that a reaction to being allowed to taste the Concoction as a reward?

It just floors me that after all this time and effort, I still seem to be doing only a little better than 50/50 on getting sufficient sample size from any given dog. Who knew dogs don’t like to drool?

But what do you do with the spit after you get it?

Last night, before bed, I collected some saliva from my dog Jack and my roommate’s dog Casey. Jack has, by this point, become inured to the idea that occasionally I am going to grab his muzzle and stick a sponge on a stick inside his lip and swab around for two minutes. He was ready for bed when I did it last night and decided he might as well just doze through it.

Casey is another matter. When I cornered him and grabbed his muzzle, he freaked out.

Casey: OH MY GOD I THINK I AM GOING TO DIE

Me: Casey, I have done this to you five times. Why do you always act like it is going to hurt?

Casey: IT DOES HURT IT HURTS HORRIBLY THE PAIN IS BLINDING

Me: I have done it to myself. I know it doesn’t hurt.

Casey: IT HURTS DOGS! I CAN’T HOLD MYSELF UP ANY MORE

Me: I have done it to about 30 dogs and it didn’t seem to hurt any of them much.

Casey: Oh.

Most dogs are annoyed when I try to insert the swab, but then discover it doesn’t hurt and just deal with it. Casey inevitably squeezes his eyes shut, hyperventilates, and sometimes even collapses to the ground. (You might at this point ask “how do you know there isn’t something special about Casey?”, to which I would reply that, while there are certainly many special things about Casey, he also reacts this way to having his nails clipped, so I am pretty sure he is not experiencing any unusual physical sensations when I collect his saliva.)

I imagined Casey asking me “Why are you doing this to me?” and what my answer might be: “To find out how stressed you are.” As bizarre as that might be from a dog’s perspective, in fact my plan was to analyze his saliva to find out how much cortisol was in it. Cortisol has been called the “stress hormone,” and is an indicator of how much stress an animal is under.

Because I’ve never worked in a lab before, I wanted to do a few sample assays with unimportant data (i.e., saliva from my own dog that is easy to collect) before using the irreplaceable data collected from hospital dogs. It has taken me four months to successfully enroll 24 dogs; if I waste those samples, I am SOL. (Six dogs a month? Well, I’m a lot better at it than I used to be, and I had to take some time off in the middle. I am actually averaging 3-4 dogs a week at this point.) I performed my test assay this morning.

The cortisol assay is a kit costing a couple of hundred dollars which I purchased from the ever-helpful Salimetrics. It is an ELISA, or enzyme-linked immunoabsorbent, assay. The main equipment in the kit is a little plate (I was surprised by how tiny it was) made up of 96 wells, into each of which one might pour a very small amount of liquid. Each of the wells has some cortisol antibodies in the bottom (or so I’m told, since obviously antibodies are too small to see). These will grab on to any cortisol they see and hold on to it.

When I got into the lab this morning, I took the kit out of the fridge, and some older samples out of the freezer, so that they could warm up to room temperature. Then, while last night’s samples from Casey and Jack centrifuged, I sat down with the lab tech’s computer and figured out where everything was going to go on the plate. Aside from all my samples, I had to have room for the standards, which are samples given to you with the kit containing a known amount of cortisol. If your assay doesn’t tell you that the well with the 3.000 standard has about 3.000 µg/dL of cortisol in it, you know you’ve done something wrong. There are also the controls, which contain an arbitrary “high” and “low” amount of cortisol, and are also useful for telling you when you’ve messed up.

Then there are the “zero” and “blank” wells. The zero wells contain nothing but the diluent, or the liquid used to dilute some of the chemical agents in the kit. When the calculations are done at the end, you need to have this well to be able to tell the difference between a real effect and an effect caused by the diluent. And, finally, there are the blank wells, which don’t have any cortisol binding antibodies in them, and therefore will behave differently than the zero wells when the plate is read.

For me, the hardest part of the process is making sure that everything goes where it’s supposed to. This is more complicated than it sounds, involving lots of finding the right well in a small plate that is packed with small wells; remembering to leave space for a sample that I’m going to add later, after it has been diluted; getting the standards and controls in the right places; etc. Doing something that doesn’t require a lot of brainpower but does require a lot of precision, over and over and over, is hard for me.

When all the standards, controls, and samples were in place, and diluent was in the zero wells and the blank wells, I added conjugate to everything. The “conjugate” is cortisol which is attached to an enzyme. Of course, the whole point of the exercise is that you are putting in your saliva samples which contain cortisol, and now you’re adding this new source of cortisol as well. The two sources of cortisol are going to compete for the chance to bind to the cortisol-binding antibodies in the bottom of each well. When there is more cortisol in the sample, less of the conjugated cortisol (with its attached enzyme) can bind, and vice versa.

I put the plate with its samples and conjugate on to a rotator. This is a small device with a flat top which basically waves the plate around so that its contents mix. Every time I use it I am terrified that the plate is going to go flying off and spread my samples across the lab floor. It has not happened yet.

After 55 minutes for the whole competition thing to happen, I washed the plate off (four times), which includes blotting: picking the plate up, turning it upside down, and slamming it on to an absorbent pad. This is also a scary procedure, especially as sometimes strips of wells break off and you have to figure out in which direction to reattach them. Attach them upside-down, and you won’t know which sample is which.

At this point, in theory, the wells were mostly full of bound cortisol. Some of the cortisol, which came from the conjugate I added, had enzyme bound to it. Some of the cortisol, which came from my samples, did not. The wells which contained samples high in cortisol had less of the bound enzyme, and the wells which contained samples low in cortisol had more of the bound enzyme. Next I added TMB (tetramethylbenzidine), which reacts with the bound enzyme to change its color. Obviously, the wells with more bound enzyme will change to a different color than the wells with less.

More plate rotating is next, and then a brief time (sequestered in the dark) for the plate to let the reaction run its course. When I took the plate out of its light-proof bag at the end of this waiting period, the wells were all full of a lovely blue liquid, and different wells were visibly different shades. I added stop solution to stop the reaction, which immediately changed the color to yellow. And then I put the plate in the plate reader, which is a small machine attached to a Windows computer on the lab tech’s desk. It made some thumping noises as it determined the optical density of each well — its color, expressed as a number. It relayed these numbers to the waiting computer, which did the calculations to convert the numbers into concentrations, and I exported the results into Excel.

So how did I do? This was actually my second attempt at doing this assay. Both times, the results were fairly close to what I expected, based on the standards. However, part of what I wanted to figure out today was whether I could dilute small samples and still calculate values which were close to undiluted values. (In other words, if I dilute one sample by 50%, and multiple the calculated concentration by 2, will the result be similar to what I get when I put an undiluted version of that sample in a different well?) My dilutions were way off, so that’s something I need to figure out. I’m hoping to find a lab on campus which does these assays frequently, and ask if I can work with them for a little while, to get some experience.

By the way, the above implies that I competently performed this entire operation on my own, which is completely untrue: I worked under the close guidance of the extremely talented hospital lab technician. I am very lucky that she enjoys teaching; she told me today that she once considered being a school teacher. Her patience and good nature made the day much more enjoyable than it might have been.

Spit Girl

Monday through Thursday evenings around five pm, I drive in to my school’s small animal hospital to collect spit from dogs. I’ve learned a lot about how to get spit out of a dog. It helps to have the right tools. If your subject is an adult human, you can ask them to drool into a cup. (Don’t actually spit! The manual doesn’t say why, but when I tried it, my spit contained long mucin strands which made it difficult to pipette up into a vial for storage.) If your subject doesn’t speak the same language as you do — mine don’t — or is otherwise recalcitrant, you might extract the spit with a small cotton rope. However, a new tool is now on the market: the Sorbette, a small sponge on a stick available from Salimetrics. (My endocrinology professor, upon hearing about Salimetrics, said: “An entire company dedicated to saliva? Who knew?”) Oh, wait. The Sorbette isn't actually all that new as technology goes: it turns out to just be an eye sponge. Not knowing much about eyes, I presume it is for adding things to or taking things away from them. But it's now also marketed to us spit collectors.

Salimetrics tells me to insert two Sorbettes under my subject's tongue and hold them there for a full minute. I should not swab. I did try this a few times. My subjects inevitably act like I have stuck a hot poker under their tongues, rolling their eyes and chewing madly. No Sorbettes have been damaged in this process, but it was only a matter of time. Also, the amount of saliva I actually acquired was too low for the assay I intended to perform on it.

So now I stick three Sorbettes in the side of the dog's mouth; this bothers them less, even with the greater mass. There also seems to be a fair amount of spit accumulating in the forward lower part of the dog’s mouth, between the outer edge of his teeth and inside his lip. And I do swab. The collection manual, when consulted, explained that swabbing is bad because it means you'll collect saliva from various different salivary glands, and so the levels of whatever you're testing might vary based on how much spit you get from which gland. This is a more important consideration for some other substances than for cortisol, which is what I care about.

Even so, the quantities of spit still occasionally leave something to be desired. My suspicion is that anxious dogs — which my study dogs are; that’s the point — just make less saliva than calm dogs, because their sympathetic nervous system is activated and telling them that making things which are useful for the process of digestion is not appropriate at this time. Eat later, worry about being stuck in the hospital now. In order to turn on the rest-and-digest parasympathetic nervous system, I have attempted waving treats under my subjects’ noses. Well, I’ve tried this once so far, and I did get a nice large sample that night. I left the treat behind with the dog; he was too anxious at the moment to ingest it, but I have hopes that he came around in my absence.

Once I have my sample, I balance it on a plate of ice and carry it upstairs to the clinical sciences laboratory, where I centrifuge it (3250 rpm, 15 minutes) so that it spins its way out of the sponge and into the tapered bottom of its vial. I then pipette it into a cryovial and secrete it in a -80° freezer. Done! When I have enough samples, I’ll perform an ELISA assay on them to find out how much cortisol is in each. I imagine I’ll report on that in detail here.