Data I’ll never publish: Antirrhinum herbivory

Inspired by this post, I’m going to try to put the results of small (but interesting) experiments up here every once and awhile. In the summer of 2014, I spent a lot of time washing plants. I was – and still am – curious of the function(s) of plant exudates. I primarily did this with Trichostema laxum and Atriplex rosea (in 2013), but I also did it with Mimulus layneae and Antirrhinum cornutum (California snapdragon). The snapdragon gave me interesting results.

(this post should also be regarded as potential project for someone else: I started it in May – there is plenty of time to get up to McLaughlin and do it again this year).

One of the experimental A. cornutum, showing leaf damage.

This snapdragon, while not as heavily glandular as Trichostema or that Mimulus, is fairly glandular-sticky, even entrapping a small number of minute insects (see the table/supplementary material). Under the microscope, you can see the fairly dense short glandular trichomes (the longer trichomes are mostly nonglandular) on the stalk and flower bud.

Stem of A cornutum with an entrapped insect.
Flower bud showing short glandular and long nonglandular trichomes.

Wondering whether the glandular exudate is defensive, I did an experiment where I removed it with water. Most glandular exudates in CA summer annuals seem water soluble, so a spray bottle rainfall takes off much of the exudate (observationally verified in situ with a 20x loupe – plus whatever was in this exudate made suds on the plant!). This manipulation was my first treatment group. Of course, adding water to a plant has an effect of its own, so I also had a water control group, where I added the same amount of water below the plant’s leaves, as to not wash off any exudates. Finally, I had a true control group, which received no water whatsoever. I instituted these treatments on the 30th of May and reapplied them on the 17th of June. Each time, I recorded the number of leaves, flowers, fruit, and plant height, as well as any damage. I also checked the plants, but did not reapply treatments on the 2nd and 19th of July (the last check all were senescent).

During the experiment, plants suffered two main forms of herbivory. The first type, which was most common and most destructive, was that the stems were entirely clipped off. I’m nearly positive this was by jackrabbits (indicated by a single flat cut diagonally across the stem) and it usually killed the plant. The photos below shows what remained.

A killed experimental A. cornutum plant. See it?!? Its the little stem to the bottom left of the flag. Also notice a nice healthy Lessingia in the background. They, too, are extremely glandular and sticky.
A survivor of mammalian herbivory. If the meristem was not completely destroyed, they often came back and branched like this. Like the classic overcompensation “herbivore-plant mutualisms”, the resulting plants were often bigger than the others, with more reproductive structures, but unlike this “mutualism”, it was too late in the season and they had low fitness, as they could not mature these structures.

The mammalian herbivory was not random. Of the 25 plants per treatment, 11 in the control group, 13 in the rainfall simulation (exudate removal) and a whopping 20 in the water control group were eaten by mammals (this is nonlethal, lethally was 10, 12, 18). With a simple chi-squared test, we can demonstrate that this was likely nonrandom (X2 = 7.3688, df =2, p = 0.025) (for lethal, X2 = 5.5714, df=2, p = 0.062). Why were the mammals targetting the water control plants so heavily?

Were they bigger and thus easier to find or just more profitable to eat? They were not significantly different in height, fruit or flower numbers from the other two groups during any check. I don’t have data on plant quality (perhaps the less water-limited plants were more nutritious or something?).

The other type of damage was equally-interesting. Heliothis phloxiphaga is a generalist caterpillar on glandular plants. It was the primary herbivore on my columbines, as well as a common herbivore on Trichostema laxum and other sticky plants. Like most heliothiine noctuids, it feeds primarily (but not exclusively) on reproductive structures. I only observed it once on Antirrhinum (eating a fruit), but all the fruit damage I found was consistent with it (and that’s one more time than I saw a jackrabbit eat it!).

The other type of damage: caterpillar fruit predation.

I had hypothesized, that if the exudate were defensive, the washed plants would be most heavily eaten. This hypothesis was supported with the fruit damage. Rainfall plants received far more damage than the other groups. (note: I didn’t actually analyze this with zero-inflated binomial, as it should be. There is a problem, in that only 7/25 of the water control plants had any fruit at all because of the rabbits.)

A crumby excel graph of proportion fruits damaged.

What does this all mean? Obviously, it means that mammalian and insect herbivores are responding to different plant traits. What they are exactly, I’m not sure (especially for mammals). If anyone (nudge, nudge, wink, wink) were to repeat this experiment, with a larger sample size, and maybe some other mechanistic experiments (perhaps cage controls and lots more trait data to see what is different in the water control and rainfall manip groups), I think its a pretty good system that someone could get a paper – if not a few – out of.

Trichostema laxum research update: the first interesting data?

Awhile back, I wrote about the beginnings of some research on Trichostema laxum. I’ve been slogging through the disappointing amount of data I gathered this field season and doing a little bit of writing. While I was, and still am, really excited about the project on T. laxum, it took a backseat to columbine and tarweed work this field season (most of which burned up). I did get some new data and perhaps gained some insight into the system.

A normal array of plants in the site: normal purple Trichostema laxum, an individual with the common white and purple lower lip phenotype and some Zeltnera trichantha intermixed (a really cool plant)
My main question in the system is: how is this polymorphism in flower color maintained? If it were a fitness benefit, we might expect a high proportion of it. If it were deleterious, it should be lost (especially as it is at reasonably low frequency). If it is neutral, it might be drifted out. I actually suspect the answer is quite a bit more complicated.
Once you start looking for variation, you find it! I don’t know what this doubled lower lip is about (it showed up in a plant grown in the greenhouse – on most flowers). The plant was male-sterile, I believe. I’ll be looking for it in the field though!
The first question is, of course, how common is the color polymorph? I censused the focal patch/population (separated by ~300 meters from others) in 2014 and 2015. In 2014, the polymorph was 2.0% (46/2278 individuals) in 2015, 3.7% (102/2757). Neither of these censuses was a complete census of the population – necessarily, I cannot assess the phenotype of any pre- or post-flowering individuals. Both were done roughly in the peak flowering time (over several days), so I do think it is close to accurate. I think its safe to conclude that the proportion stayed the same or even went slightly up in 2015.
A rather large w/p morph individual.
The next logical question is: do the two morphs have similar field fitness? Any “fitness” measure (e.g. reproductive success, height, etc.) of these plants is dictated mostly by microhabitat location. In this rocky, heavily serpentine site, most plants stay under 20 cm tall and never put out more than 50 flowers (mints have 4 ovaries per flower, so maximum seed set is four times flower number). In a wetter, less serpentine and rocky meadow, I once found a plant on a gopher mound (which brings up nutrients) that was nearly a meter wide and better than a half meter tall. It probably had >5000 flowers throughout the season.
A veritable field of Trichostema! Not my field site – this site has huge plants (~500 flowers/plant) and very little flower color variation. It is a nice place to look at the insect communities on T. laxum, as it has really high densities of herbivores and predators (T. laxum gets most of the sticky plant predators)
Because of this microhabitat variation, the best comparison to make is nearest neighbors which differ in flower color. In both 2014 and 2015, I took data on 41 pairs (coincidentally!) of white/purple and purple/purple neighbors. I found no significant differences, either year, between any fitness variables (number of buds, flowers, fruit, height and, in 2014, number of leaves and herbivory [too low in 2015]).
A more normal-sized (for this population) individual.
This, ostensibly, seems like the trait is fitness neutral (and lab growouts seem to bear this out – more data soonish). Given that this site burned in August this year (after most had flowered, but some [probably few] were still maturing seeds), I was curious about whether the morphs differed in phenology. Hindsight is 20/20 (I should have censused biweekly!), but the neighbor pairs data can be used to examine this in a roundabout way; I have data on buds, flowers and fruit, so later phenology plants should have a higher proportion of buds to flowers and fruit than earlier phenology plants.
In both years, the white/purple plants had a lower proportion of buds than the purple plants (it is marginally significant). This suggests that they have a earlier phenology – which could be what is under selection – not the flower color itself. I am super, super, super, excited about this (the only positive result so far from anything in T. laxum) – there was possibly a big selective event (a fire) on 12-August (I think – could have burned on the 13th). From this, I’d predict that the w/p morph may have dehisced a higher proportion of their seed set by the fire. I’ll be paying far more attention to the phenology, and recensusing more often this upcoming season.
I also analyzed the pollinator data from 2014 and got no particularly useful insights (I wondered if there was some degree of isolation between the morphs). The pollinator communities using each morph were pretty similar and the only real differences were:  a bee on a purple flower was more likely to visit a w/p next than a bee starting on a w/p* and, only bees that started on w/p flowers next visited a snapdragon, Antirrhinum cornutum (but only 3% of the time). This last result is interesting as the snapdragon also has whitish purple flowers AND the T. laxum population with the w/p flowers is the only one  (out of ~15) I’ve found interspersed with large numbers of A. cornutum. I’ll have to get MUCH better data for any hypotheses about its effect.
Antirrhinum cornutum, grown in lab, showing the pale purple/white flowers.
I’m working now on the “genetics” (well, inheritance, but that’s as close as I ever get to ATGC) of the polymorphisms (this one and selfing). Could w/p be recessive and heterozygous in more individuals (~25% under HW assumptions)? I don’t think it is (entirely) developmentally induced, as in the first grow out, I only got this polymorphism from this population (I grew individuals out from 4 populations). More soon! Do let me know if you have other ideas about the system!
Heliothis phloxiphaga was a very common herbivore on T. laxum in 2014 (this plant had two – I didn’t stage this), but nearly absent in 2015 – though it was still common on columbines and tarweeds.
*I think this is confounded, as I watched two plants during each observation – nearby plants that were similar in size. Therefore, I suspect that it was more likely that a bee on the p/p plant would encounter a w/p than one from the w/p.