Factors mediating coexistence of closely related species have long been of interest in the origin and maintenance of biodiversity.  Niche differentiation, for example, facilitates stable coexistence by reducing interspecific competition. Here we study adaptive divergence in congeneric seed bugs that often co-occur at small spatial scales: the small milkweed bug (Lygaeus kalmii) and the false milkweed bug (L. turcicus). L. kalmii primarily feeds on milkweeds (Apocynaceae) and sequesters toxic cardenolides as an anti-predator defense; little was previously known about L. turcicus. We report the primary host plant of L. turcicus is false sunflower Heliopsis helianthoides (Asteraceae) and that its seeds produce cardenolides, the second report of this toxin among Asteraceae. We find divergent trends in performance and defense of the bugs on the two hosts: L. kalmii had >30% higher growth rate on milkweeds than on false sunflower, while L. turcicus performed equivalently on each species. Conversely, L. turcicus sequestered more effectively from false sunflower, while L. kalmii sequestered equivalently on the two seeds. The thick shell of false sunflower seeds contributes to the disparity in performance, and differences in seed chemistry are hypothesized to drive differences in sequestration. In particular, false sunflower cardenolides are converted to a diversity of breakdown products by its specialist (L. turcicus), while L. kalmii sequesters these compounds intact. Thus, adaptation to host-plants is asymmetrical for growth and defense between these two seed bugs. Shared plant chemistry apparently bridged host-shifts in this group of insects, with adaptive divergence to their food niches contributing to diversification.

H. helianthoides seeds were purchased from commercial vendors. Freeze-dried seeds (50 mg for seeds) were ground to a powder (Retsch Mixer Mill 300). 1.6 mL meOH with 20 μg digitoxin (Sigma-Aldrich, MO, USA) was added to each sample, along with ~30 FastPrep beads. Samples were agitated twice for 45s at 6.5 m/s on a FastPrep-24 homogenizer and centrifugated for 15 min at 12,000 rpm. Solids were discarded, and samples were dried in a centrifugal concentrator at 35 ^oC. Samples were resuspended in 250 μL of meOH, and each resuspended sample was vortexed 3 times with 1.5 mL of hexanes and centrifugated for 15 min at 12,000 rpm. The hexane layer was discarded to de-fat each sample. Samples were once again dried down, resuspended in 250 μL of MeOH, and filtered with a MultiScreen Solvinert 0.45 μm low-binding hydrophobic PTFE 96-well plate (Millipore, Burlington, Mass). We analyzed samples using HPLC following the method used by Züst et al. (2019). Fifteen μL of the samples were injected into an Agilent 1100 series HPLC (Agilent, Waldbronn, Germany) equipped with a diode-array detector. Cardenolides were eluted at a flow rate of 0.7 mL/min across an acetonitrile gradient (0–2 min 16% acetonitrile, 24 min 70% acetonitrile, 30-40 min 95% acetonitrile, and a 10 min post-run with 16% acetonitrile). Peaks with symmetrical absorption maxima between 216 and 222 nm were considered to be cardenolides. Cardenolide concentrations were calculated based on peak areas at 218 nm and were standardized by the concentration of digitoxin in the sample and the sample’s dry mass. Peak areas were calculated from HPLC chromatograms using the chromatographR package.

Nuclear magnetic resonance (NMR) spectra were recorded on a Bruker 600 Avance III HD console and BBFO⁺ cryoprobe.  All 1D data (¹H, ¹³C) and 2D data (DQCOSY, gCOSY, HSQCAD, gHMBCAD, HSQCTOCSY, ROESYAD) were obtained with Bruker standard sequence. Sample was measured in CD₃OD (99.5%) and the residual solvent signals were δ_H 3.31/δ_C 49.15. NMR data were processed with MestReNova version 15.0.0-34764 Mestrelab Research S.L.