Devastating disease can cause increased breeding effort and success that improves population resilience

Novel and invasive diseases are a key threat to wildlife and can cause massive, unprecedented declines and extinctions. The amphibian fungal disease chytridiomycosis has had devastating global impacts, but after severe declines some amphibian species can persist and even rebound. Understanding how these species survive is critical to discovering management techniques for supporting declining species. Here we explored the impacts of disease on reproduction in frogs, investigating its effect on primary and secondary sexual characteristics and breeding effort and success.  Male frogs are increasing various facets of their breeding efforts resulting in increased offspring. Infected male frogs have 1) increased vocal sac coloration, 2) increased sperm quality, and 3) higher mating success and father more egg masses than uninfected males. Ongoing high mortality due to chytridiomycosis in these species appears to be balanced by high reproduction. Management efforts should target the natural mechanisms (e.g., breeding) that species use to overcome key threats because they are more likely to succeed and be sustainable.

Methods
Study species

Litoria verreauxii alpina, the alpine tree frog, is endemic to the Australian Alps and is considered critically endangered in both New South Wales and Victoria, Australia. Since the introduction of the fungal pathogen Batrachochytrium dendrobatidis in the 1980s, this once widespread species has experienced dramatic declines and is now present in <20% of its former range (Osborne et al. 1999). The species now inhabits 8-10 sites across the Alps. While the species is inhabiting a much smaller area than they were historically, the populations at these remaining sites appear to be relatively stable. The species is highly susceptible to B. dendrobatidis infection and experiences near complete population turnover every year (Brannelly et al. 2015, Scheele et al. 2015). It is well understood that the species is persisting in these habitats due to breeding (Brannelly et al. 2020a, 2021a), but it was unclear if the species was actively increasing their breeding output to persist or if their baseline reproductive effort was high enough to maintain the populations. However, since the introduction of B. dendrobatidis to the populations, there has been changes in baseline spermatogenesis, indicating that the populations with endemic disease are evolving increased reproduction (Brannelly et al. 2021b).

Secondary Sexual Characteristics
Clinical infection trial

Adult male L. v. alpina were bred in captivity at the Melbourne Veterinary School (Brannelly et al. 2023). Males were 1 year old and sexually mature at the time of the experiment, and randomly selected from three clutches of origin, and evenly allocated to each treatment group (n = 30, 15 infected and 15 control; block randomisation across clutch and treatment). Frogs were housed individually in terraria (34cm x 20cm x 15cm) on a gravel and moss substrate at 16-19ºC, 12:12hr light:dark cycle during the Austral spring, which is their natural breeding season. Terraria were flushed daily with carbon-filtered water and were fed crickets dusted with vitamin powder (Repashy) twice per week for the duration of the experiment.

Males were inoculated with a known virulent strain of B. dendrobatidis from New South Wales Victoria, AUS#46 (WastePoint-L.verreauxii #5-2013-LB.RW). Bd was grown on agar and tryptone, gelatin hydrolysate, lactose plates and incubated for three days at 20°C. The plates were flooded with 3ml of sterile Milli-Q water and agitated for 10 minutes to allow the Bd zoosporangia to release zoospores. The inoculum was poured off the plates and the Bd zoospore concentration was determined were counted with a haemocytometer. Litoria v. alpina males in the Bd treatment group were inoculated with 5×105 zoospores in 5ml of inoculum. The inoculum was dripped onto their dorsum with a syringe and allowed to pool in the 20ml inoculation container. The control frogs were given a mock inoculation by flooding sterile agar plates with Milli-Q water, and then dripping 5ml of the control inoculant onto the animals. All frogs were left in inoculation containers for 12 hours, and then returned to their individual terraria with gravel and moss substrate. 

Frogs were swabbed for B. dendrobatidis (see methods below), measured, and analysed for breeding colouration weekly. The snout-to-vent length (SVL) to the nearest 0.1mm was measured using dial callipers, and each frog was weighed to the nearest 0.01g using digital laboratory scales. Colour readings were taken using a portable spectrophotometer (see methods below) for three measurements on the throat, dorsum, and venter. The experiment ended after 4 weeks when some animals began to show clinical signs of chytridiomycosis (e.g., anorexia, irregular skin slough, splayed leg posture).

Batrachochytrium dendrobatidis testing
To collect a sample for B. dendrobatidis from each frog, a dry sterile rayon swab (MW-113, Medical Wire and Equipment Wiltshire, UK) was rolled five times along each of: the ventrolateral surface of the abdomen, the ventral surface of all limbs and digits, and the medial surface of the thighs. The tip of the swab was then broken off into a 1.5 mL Eppendorf tube and stored at -20ºC until processing (approximately 3 months). Skin swab samples were extracted using PrepMan Ultra (Thermo Fisher Scientific). The extraction method followed the manufacturer’s directions which included: adding 50 μL of PrepMan Ultra and 30-40 mg of 0.5 mm silica beads (BioSpec) to each, homogenizing samples (using a cell homogenizer) for 2 min at 1400 oscillations per sec, incubating samples at 95 °C to lyse the cells for 10 min, and collecting and diluting the supernatant 6:100 in ultra pure water before directly analysing for pathogen presence and quantity using qPCR (Brannelly et al. 2020b). The remaining extracted DNA was stored at 4 °C. With every extraction performed, one B. dendrobatidis positive control sample (zoospores from culture) and one negative control (swab only) were extracted.

We used qPCR (Rotogene, Qiagen) to amplify and quantify the B. dendrobatidis DNA in each sample following Boyle et al. (2004) with modifications (Jadwani-Bungar et al. n.d., Brannelly et al. 2020b). We used a 15 μL reaction volume, with 5μL of template DNA, 7.5 μL of lo-ROX 2x master mix (SensiFast, Bioline), and a final reaction concentration of 900 nM ITS1-3 (Purification: RP-Cartridge Gold, Sequence: CCTTGATATAATACAGTGTGCCATATGTC, Eurogentec, Integrated Science), 900 nM 5.8S Chytr (Purification: RP-Cartridge Gold, Sequence: AGCCAAGAGATCCGTTGTCAAA, Eurogenetec, Integrated Science), 250 nM Chytr MGB2 5’ 6-FAM-labelled probe (5' 6-FAM - CGA-GTC-GAA-CAA-AAT - MGB-Eclipse® 3'; Integrated Science) and 400 ng/μL bovine serum albumin (Fisher Biotechologies). The amplification conditions were 2 min at 50°C, 10 min at 95°C, followed by 15 s at 95°C and 1 min at 60°C for 40 cycles. On each qPCR reaction plate we included a series of seven plasmid-based Bd standards (purchased from Pisces Molecular, containing 4.2, 42, 420, 4,200, 42,000, 420,000, and 4,200,000 DNA copies per reaction). On each qPCR reaction plate we included a no template control, where ultrapure water replaced the template DNA. The extraction positive and negative control samples were analysed via qPCR like all sample extracted DNA. 

DNA copies of the B. dendrobatidis ITS gene extracted from each swab were estimated by extrapolating the gene copies detected by qPCR (Roto-Gene Q 2.3.5 software), considering the elution volume and dilution of template DNA. Any sample was determined to be Bd positive if the reaction well was amplified, representing at least 2 ITS DNA copies detected in the sample. If the sample was considered negative for Bd, the Bd DNA copies were coded as 0 for that sample (Brannelly et al. 2020b). Samples were analysed via qPCR in singlicate, which is a common and accepted practice for chytrid infection studies (Brannelly et al. 2015, 2017). All positive extraction controls were positive, all negative extraction controls were negative, and all no-template qPCR controls were negative for B. dendrobatidis DNA copies. 

Skin colour methodology
Colour readings were taken using a portable spectrophotometer (OceanOptics USB2000+), probe (OceanOptics R400-UV-VIS), and light source (OceanOptics PX-2 Pulsed Xenon light source). We set the light source to single pulses, with an integration time of 50msec, a data update range of 5msec, and the scans to average set to 20. The spectrophotometer readings were obtained to create a spectral curve (OceanView software) from 300-700nm via reflection of the surface of the skin. From that curve we were able to calculate brightness (area under the curve from 300-700nm), UV chroma (area under the curve from 300-400nm) and yellow-orange chroma (area under the curve from 570-620nm). Three repeat colour readings were taken from the dorsum, throat, and venter of each frog. The probe was disinfected with ethanol after every individual. The spectrophotometer was recalibrated using a white diffuse reflection standard (OceanOptics WS-1SL) every 30 minutes at a maximum. The spectral curve was exported from the OceanView software and imported into R within the RStudio interface (RStudio Team 2020, R Core Team 2022). The auc function in the MESS package (Ekstrøm 2020) was used to calculate the area under the curve for the colour wavelength readings.  We used an interpolated average to calculate the mean colour readings at 5nm subdivisions (Cassey et al. 2012). This approach ensures that wavelength curves are smooth between spectrophotometer readings with slightly different sampling slit-widths (Cassey et al. 2012).

Secondary Sexual Characteristics
Field trial for colour and calling characteristics

Study site 
Fieldwork was conducted at two sites in the Victorian Alpine National Park, Blue Rag Dam (-37.0841881 °S, 147.1217652 °E, 1600 m elevation) and Dead Timber Creek (-37.019346 °S, 147.239182 °E, 1590 m elevation). The region is classified as Alpine Sphagnum Bog and Associated Fen, which is listed under the Environment Protection and Biodiversity Conservation Act (EPBC) 1999 as a nationally threatened ecological community. Both sites have a permanent water body and are fed by alpine streams and seeps. The vegetation is comprised of snow gum (Eucalyptus pauciflora) overstorey with sedges, rushes, wetland heaths and sphagnum moss understory and ground cover. Batrachochytrium dendrobatidis is endemic to both sites.  Litoria v. alpina breed in the spring following snowmelt and winter torpor (Brannelly et al. 2015, 2016). 

Field sampling
Our surveys were conducted during the austral spring over five sampling times between October 19 and November 18, 2018. We sampled a total of 45 male L. v. alpina for call recordings, colour readings, and infection. 

We located frogs by triangulating their calls and visual spotlighting. Once a calling male was located, we approached the animal quietly and carefully to ensure that natural calling behaviour was not disturbed. We used a Marantz Professional hand-held solid state recorder (PMD6661 MK11, InMusic Japan) and directional microphone (Rode Microphones, Sydney, Australia) to record calls of male frogs. The recorder was set up one metre away from the calling animal and the field team retreated to ten metres away from the frog once the recording began to minimise disturbance to the animal. We recorded calls using mono recording settings in .wav format at a sampling rate of 48kHz with a 24bit resolution. We recorded the calling male for a maximum of 20 minutes, with the first and last five minutes of the recording disregarded to allow the frog to acclimate to our presence. Once the recording was complete, we measured the background noise level using a decibel reader (specs).  

Prior to capture, we measured the temperature of the frog and the substrate temperature using a dual laser infrared thermometer (RS PRO 8861 ±1°C). We captured frogs using clean, gloved hands and by dipnet. We placed each frog in individual zip lock bags to prevent potential cross contamination of Bd infection. We weighed frogs to the nearest 0.1g with spring scales and measured snout-to-vent length (SVL) to the nearest 0.1mm using dial callipers. 

We collected colour readings for male L. v. alpina in the field using the same method as outlined in the lab methods above, except in the field we only took throat colouration measurements. 

Call analysis 
From the 10-minute call recordings, we chose five complete calls per male for analysis. We disregarded any calls that were overlapping or where the frog was not caught after the recording to confirm their infection status. We obtained call and infection data for 39 male L. v. alpina. We analysed the calls using the sound analysis software Raven Pro (Cornell Lab of Ornithology, Bioacoustics Research Program). We included the following call characteristics in our analysis: call duration (length of the call in seconds from the start of the first pulse to the end of the last pulse), intercall interval (length of time in seconds between two calls), pulse rate (pulses/sec), total number of notes per call, note duration (length of note in seconds), internote interval (length of time in seconds between two notes), peak note frequency (frequency, Hz, occurring at the highest amplitude) and the number of pulses per note. 

Primary sexual characteristics
Clinical lab experiment on sperm quality and quantity
Animals were housed individually following the same husbandry procedures as clarified above, where animals were housed in individual containers with a gravel and moss substrate. Enclosures were flushed with carbon filtered water daily and fed twice weekly. In this experiment, we selected individual males (n = 58) from our L. v. alpina captive colony when they were 1 year old and sexually mature in the austral spring. The animals were randomly selected from 6 different clutch groups and evenly distributed across the exposed and unexposed groups, and randomly allocated a number within the two treatment groups (block randomization across treatment and clutch of origin). Animals were exposed to the fungal pathogen (n = 29) or unexposed control animals (n = 29). We followed the same exposure protocol as listed above, however, animals in the infected group were exposed to less zoospores (3x105 zoospores in 5mL Milli-Q® water). We exposed these animals to a lower infectious dose to try to extend the experimental duration to 6 weeks before clinical signs were present. 

Each week animals were swabbed for B. dendrobatidis (following the procedures above), weighed and measured. Every two weeks a subset of the animals was induced for spermic urine collection. Animals were only induced once during the duration of the experiment. We successfully collected samples from all 29 control animals (9 on week 2, 10 on week 4 and 6). Two exposed animals had either cleared infection or were never infected at the time of sample collection, therefore the total samples collected for the infected animals was 27 individual samples (10 on week 2, 8 on week 4 and 9 on week 6). Some samples produced few sperm therefore not all analyses were conducted on all samples, for example, viability analysis was able to be conducted on a total of 48 individuals (see Table S3). 

We used exogenous hormones to stimulate and increase the release of sperm into the spermatic urine. We selected the drug Chorulon (a purified hCG) at a dosage of 120 IU as it has been shown to induce high levels of spermiation in L. v. alpina 1 h after exposure (Pham & Brannelly, 2022). Prior to administration of hCG we moved the frogs into 9cm x 9cm x 3cm enclosures containing filtered tap water to allow adequate hydration and aid in spermatic urine collection. After 2 h 120 IU of hCG was injected into the peritoneal cavity from the ventral side using a 27 g needle. We then placed animals back into individual enclosures for 1 h before collection of spermatic urine. Prior to sample collection we used a Kim wipe to dry cloacal area of the animal. A fire-polished glass Drummond microcapillary tube (20μL) was inserted into the cloaca with spermatic urine excretion promoted through light massage and rotation of microcapillary tube.

To determine sperm cell concentration, 2µL spermatic urine was diluted 1:10 in Simplified Amphibian Ringer solution. Sperm cells were counted at 200x using a hemocytometer with 10uL of diluted sample added to each of the two chambers. For each sample a minimum of 50 sperm cells were counted (repeated for both chambers and averaged) or where 50 cells were not present, both chambers were fully counted and every sperm sample was included in analysis for analyses of concertation, total sperm cells within the sperm produced and volume of sperm. 

Once sperm cell concentration was determined, 30 sec videos were recorded at 200x magnification totalling at least 35 sperm cells per sample. If 35 sperm cells could not be recorded within the diluted sample, the sample was not included in the analyses. Within the videos the sperm cells were scored as non-motile (no observable activity), or motile (if any activity was observed over the 30 sec observational period). Within the motile category, individual sperm cells were categorised as forward moving (the cell moved in a forward motion for all or part of the 30 second observation), not-forward moving (the sperm head moved but not in a forward motion for all or part of the 30 second observation) and stationary moving (the tail moved but the head did not for all of part of the 20 second observation). 

Sperm viability was determined using Eosin-Nigrosin staining procedure (Agarwal et al. 2016), with eosin being a supravital stain and Nigrosin acting as counterstain increasing contrast to better visualise sperm. We added 1µL undiluted spermatic urine sample, 2µL Eosin Y (1% aqueous) and 2µL Nigrosin (10% aqueous) on a slide and mixed with pipette. A thin smear of the stain was created using a clean slide. Stained slides were air-dried and then mounted with Permamount Mounting Medium (Fisher Scientific). Viability scoring was completed using bright-field microscopy at 1000x with oil-immersion objective lens. Live/viable sperm cells had white/clear or light pink heads, whereas dead sperm cells displayed red or dark-pink heads that were larger in size Fig S3. For each sperm sample, a minimum of 90 total sperm cells were analyses. If the slide did not contain at least 90 high quality sperm cells to analyse, then they were excluded from the analysis.

Sperm morphology of the live sperm cells was conducted via 1000X oil immersion photos taken of individual sperm cells stained with Eosin-Nigrosin. Photos were taken and analysed using ImageJ software (Wayne Rasband, NIMH) to determine the length of head and the length of the tail using the segmented line tool and measurement analysis (Pham and Brannelly 2022). We aimed to analyse 20 individual sperm cells for morphology where possible. If we could not analyse at least 10 individual sperm cells per sperm sample, then the sample was not included in the analysis. 

Mating success and offspring production
Study overview
The fieldwork for this study was a capture-recapture survey where animals were surveyed weekly for six weeks (12 October – 21 November 2018) at one site in the Victorian Alpine National Park, Blue Rag Dam (coordinates withheld because the species is endangered (Lindenmayer and Scheele 2017), 1600 m elevation). The study overviewed here also included the animals captured within the field study for “Secondary Sexual Characteristics” described above. 

Data collection 
Frogs were captured via hand with a fresh pair of gloves and individuals were placed in a clean ziplock plastic bag. Upon initial capture individuals were sexed and had morphometric data including snout-vent length (mm) and mass (g), skin swabs for Bd infection load, toe clips, and pictures for identification for recaptures taken. Any subsequent captures of an individual included all the former listed above minus the acquisition of toe clips. At the time of capture, if a male was actively in amplexus with a female, we separated the two animals, but noted which female the male was actively engaging with. We identified pairs to be in amplexus if both the male had a tight grip and the female was not actively attempting to escape the male’s grip. 

Once the animals were collected, photographs were taken using a Panasonic model DMC-G5 with a Lumix G VARIO 14-42/F3.5-5.6 lens. Pictures of the left side of the groin and dorsal were taken and used for identification of individuals. Individuals were swabbed for B. dendrobatidis (described above). Toe-tip clips were taken upon initial capture of an individual using bleach sterilized scissors, severing one toe-tip at the first phalange, and storing the sample in 100% ethanol.  

Upon each survey timepoint, new L. v. alpina egg masses were identified, and photos were taken to estimate egg mass size (by estimating the proportion of the egg mass in the photograph and counting the visible eggs. For each newly laid egg mass, 4 eggs were collected with a disposable 2mL pipette from different areas on the outside edge of the egg mass (to minimise egg mass disturbance). The individual eggs were stored 100% ethanol in individual 2mL tubes. 

In total, 197 unique individuals (179 males and 18 females) were captured during the field season with 272 recapture events of males. A total of 451 capture events of male frogs was conducted. Female frogs were never recaptured during the sampling period. 55 egg masses were sampled, and 41 unique egg masses were identified, ranging in size from approximately 5 – 220 eggs. 

Parentage analysis
To assign parentage to egg masses, all adult toe clips as well as three eggs per egg mass (165 total eggs) were genotyped. Tissue samples and eggs were sent to Diversity Array’s Technology which utilizes the genotype technique DArTseq™. DArTseq™ uses a combination of DArT complexity reduction methods and next-generation sequencing platforms (Jaccoud et al. 2001, Sansaloni et al. 2011, Kilian et al. 2012, Cruz et al. 2013).  

A dataset of approximately 85,374 SNPs was returned and these were stringently filtered in R using the RStudio interface (RStudio Team 2020, R Core Team 2022) using the dartR package (Gruber et al. 2018) to 3570 SNP loci that had a call rate of higher than 90% and reproducibility higher than 98%. From these SNPs and the successfully genotyped individuals we used the program Colony (Wang 2004, 2012, 2013a, 2013b, Jones and Wang 2010a, 2010b) to determine parentage of egg masses. 

Determination of male breeding status
Within our genetic parental analysis, we found that no egg masses were produced between males and females that we captured in amplexus with each other. While amplexus does not always lead to spawning (Jennions et al. 1992, Orton et al. 2023, Brannelly et al. 2023), in many cases the females in amplexus pairs began to oviposit directly after capture. We believe that we disturbed the breeding pairs for sample collection, and while no breeding event took place, it would have if we had not interrupted. Therefore, when we determined the breeding status of the males at each sampling point throughout the breeding season, we included both egg masses produced and observed pairs in amplexus. Additionally, by including observations of both amplexus and the production of egg masses we bolstered our results: we associated 21 egg masses to 20 male capture events throughout the breeding season, and 18 instances of amplexus.