Study sites were selected based on the availability of historical data from the U.S. Virgin Islands Territorial Coral Reef Monitoring Program (TCRMP) (Ennis et al., 2020) and data on SCTLD distribution that was tracked weekly by the Virgin Islands Coral Disease Advisory Committee (Costa et al., 2021; VI-CDAC, n.d.). TCRMP has monitored sites in the USVI annually since 2002. TCRMP methodology captures long-term baseline data on the benthic cover, coral and fish species abundance, and diversity. Additionally, coral health metrics are recorded following modified Atlantic and Gulf Rapid Reef Assessment (AGRRA) methodology (Ennis et al., 2020). TCRMP sites have six permanent transects at each site where a diver on SCUBA records video of the benthos, maintaining a distance of 0.4 m above the substrate through use of a guide. Twenty-five images per transect are then captured from these videos, representing an area of approximately 0.31² m. On each image 20 random points are generated using either Coral Point Count with Excel Extension (CPCe, used prior to 2019) (Kohler and Gill, 2006) or R statistical analysis “imager” (Barthelme et al., 2020) and “spatstat” (Baddeley et al., 2015). The substrate under each point are identified and coral cover calculated for each transect (Ennis et al., 2020; Brandt et al., 2021). Total living coral cover and coral cover by coral species calculated for each transect were used in this study.

Three distinct phases of SCTLD on an affected reefs were recognized by VI-CDAC and defined in Costa et al. (2021). These included emergent, epidemic, and endemic phases characterized by the timing of SCTLD affecting a site, and sites were categorized into zones based on these phases. The “Emergent Zone”, or disease front, included sites with no disease present but were near the disease front or where only highly susceptible species showed signs of disease, indicating a recent emergence of SCTLD estimated to be within one month of disease arriving at the site. The “Epidemic Zone” included sites experiencing peak prevalence. In this zone, highly susceptible species were still prevalent and were highly affected, but the moderately and low susceptible species including Orbicella spp. and Porites astreoides were also showing signs of disease. In the epidemic zone, it was estimated that the disease had been present between one and nine months. Lastly, at sites in the “Endemic Zone”, Orbicella spp. were well beyond initial onset and most highly susceptible species had experienced significant mortality from disease, or may have been lost from the site entirely. This phase is estimated to represent the time period of greater than nine months from the first signs of disease (Costa et al., 2021). Survey sites were chosen from each of these disease zones (emergent, epidemic, and endemic) at the time of the surveys or long-term study site establishment. The timeline of SCTLD affecting sites and characterization of sites into the zones was based on weekly, territorial-wide monitoring and reporting of the disease (Brandt et al., 2021; Costa et al., 2021; VI-CDAC, n.d.). All sites were represented by diverse fringing reefs in 5-20 m depth ( Figure 1 ; Tables 1 and 2 ).

Rapid goby population assessments were conducted by two divers who swam three meters apart in a haphazard pattern across the reef for 50 - 60 min and searched for the presence of Elacatinus spp. on corals, sponges, and other substrates. When a goby was sighted, the coral head or other substrate was searched and the number and size ( ± 1 mm) of gobies, depth, and substrate type were recorded, and a photo was taken of the cleaning station. If the substrate type was live coral, then coral species, presence/absence of SCTLD, percent (%) recent mortality associated with SCTLD lesions, and coral width, length, and height ( ± 1 cm) were recorded.

At each long-term monitoring site, divers established one or two 25-m reference transects and marked the beginning and ending points with sub-surface buoys tied to dead coral. Live coral cleaning stations occupied by Elacatinus spp. cleaner gobies were located by divers who swam in a systematic search pattern along reference transects at each site. All corals, sponges, and other substrate types were carefully searched, and when a cleaning station was located, the distance along the transect and distance to left or right of the transect were recorded to relocate goby cleaning stations in future surveys. Each coral head with a cleaning station was tagged with a unique number plate hammered into adjacent dead substrate with a masonry nail and flagging tape, data recorded as described above. For endemic and epidemic sites, 25 cleaning stations on healthy corals were tagged. At emergent sites, 50 healthy cleaning stations were tagged to increase the probability of capturing disease stages from initial signs through colony death during the study. After corals were tagged, health data were taken following the National Oceanographic and Atmospheric Administration (NOAA) National Coral Reef Monitoring Program (NCRMP) coral demographic methods (Coral Demographics Survey Protocol for the Atlantic Region: U.S. Caribbean, Florida and Gulf of Mexico: 2018, 2020). The presence, number, and size of gobies as well as condition of corals were reassessed at least five times over the following seven months ( Table 1 ). During each subsequent survey, transect tapes were reattached to buoy lines and datasheets with goby cleaning station location were used to relocate marked coral heads. Coral health was reassessed, corals were photographed, and the number of gobies and their sizes were recorded. If no gobies were seen, a two-meter radius around the original coral head was searched for gobies that may have relocated to another coral or sponge. If a goby of similar size was found, the alternate location was recorded.

To evaluate the persistence of cleaning gobies around the marked cleaning stations, area surveys were conducted during the last survey of tagged corals at each site. These surveys consisted of delineating the study area with transect tapes so that the most distant tagged corals to the left and right side of the original transect defined the rectangular population-survey area. All live and dead corals, sponges and other substrate types were systematically searched for the presence of cleaner gobies. When a goby was located, data were recorded as described above (Coral Demographics Survey Protocol for the Atlantic Region: U.S. Caribbean, Florida and Gulf of Mexico: 2018, 2020).

Due to SCUBA diving limitations, timed surveys at deeper sites (i.e., Buck Island STT, Seahorse) were a few minutes short of the full 60 minutes ( Table 1 ). Thus, disease zones were compared using gobies encountered per minute. These data were analyzed with a 1-way Analysis of Variance (ANOVA) to test for an effect of disease zone. Area surveys, unrestricted by time, at long-term monitoring sites were used to calculate the density of gobies among sites and disease zones during the initial and final surveys, a 9-to-12-month period ( Table 1 ). Since timed surveys and area surveys used a similar methodology and collected comparable data while searching for goby-occupied corals, we compared the average size of Elacatinus spp. and the number of gobies per cleaning station using a 2-way ANOVA with disease zone and survey type (timed vs. area) as independent variables. These results were not significant for survey type. Therefore, data were combined and the number of gobies per cleaning station and goby lengths were each tested among disease zones with 1-way ANOVAs. For long-term monitoring surveys of tagged goby cleaning stations, recruitment of juvenile gobies after the initial population survey may have affected persistence estimates over time. Therefore, to standardize persistence data we excluded likely recruits (TL < 13 mm) arriving after the first survey from analysis ( Supplementary Table 1 ).

Use of habitat by cleaner gobies was calculated using percent cover values from TCRMP data and goby occupation from this study. We applied Ivlev’s Electivity Index because certain habitat options utilized by gobies were so scarce that they were not detectible by TCRMP methods, resulting in a recorded zero value for those habitat types. Ivlev’s Electivity Index assigns the maximum selection value (1) to those habitats that are occupied but have a cover value of zero (Manly et al., 1993). Electivity indices were only calculated for Fish Bay, Magens Bay, Buck Island STT, and Black Point where both TCRMP and long-term monitoring data were available.

Goby encounter rates were significantly different among SCTLD disease zones (F_2.6 = 12.97, p<0.007). Endemic reef sites had 50% fewer Elacatinus spp. (0.31 ± 0.015 SE gobies min^-1 search time) compared to epidemic (0.61 ± 0.076 SE gobies min^-1) and emergent (0.67 ± 0.055 SE gobies min^-1) sites ( Figure 2 ).

Areas surveyed at monitoring sites for total goby population varied from 300 m² to 1072 m² ( Table 1 ). Goby density ranged from 0.023 m^-2 at Flat Cay and 0.09 m^-2 at Hull Bay during initial surveys to 0.037 m^-2 at Buck Island and 0.138 m^-2 at Magens Bay during final surveys ( Table 1 ). Changes in density from first to last survey varied across sites but were relatively similar between sites on the same side of the islands (North or South). All observed density losses occurred at sites on the south side of St. Thomas or St. John and the greatest increases in density occurred at the northern sites (Magens Bay and Hull Bay, Figures 1 and 3 ). The most notable decrease in goby density occurred at Buck Island, St. Thomas where decreases were observed across all life stages and persistence of gobies >13 mm was lowest ( Figure 3 ). Low settlement, immigration and persistence of juveniles and adults resulted in negative changes in density at this epidemic site which experienced high coral mortality from SCTLD ( Table 1 ; Figure 3 ). This site is the furthest away from the shoreline of St. Thomas or St. John ( Figure 1 ). Changes in density were influenced by goby life history stages where larval settlement, persistence of all life stages, and immigration of gobies from degrading habitat resulted in positive or negative changes in density from first to last surveys at most sites ( Figure 3 ).

Similar to other studies, this study confirmed goby occupation of live coral more often than expected by chance when accounting for available habitat area ( Supplementary Figure 6 ). Combined data for rapid assessment surveys and area surveys at monitoring sites showed that the number of Elacatinus spp. per live coral cleaning station was not significantly different among disease zones (F = 0.15, P = 0.86) and averaged 1.4 gobies/coral ± 0.06. The maximum number of gobies per coral head ranged from 1 – 5 at emergent sites, 1 – 7 at epidemic sites and 1-8 at endemic sites ( Supplementary Figure 1 ). The percent frequency of gobies per coral head was nearly identical among disease zones ( Supplementary Figure 1 ).

Long term monitoring data revealed no relationship between the number of gobies per coral head and coral head maximum diameter (r² = 0.05) or surface area (r² = 0.04). Coral species that hosted greater than three Elacatinus spp. (>13mm) per cleaning station included M. cavernosa (n = 6, max = 6 gobies), Siderastrea siderea (n = 4, max = 5 gobies), O. faveolata (n = 1, max = 4), O. franksi (n = 1, max = 4) and C. natans (n = 5 max = 7). Coral species that hosted three or less Elacatinus spp. per cleaning station included D. labyrinthiformis (n = 12, max = 3), O. annularis (n =10, max = 3), P. astreoides (n = 4, max = 1), and P. strigosa (n = 12, max = 3).

Results of both long-term monitoring sites and rapid assessments showed similar Elacatinus spp. distribution patterns among live coral cleaning stations ( Figures 4 , 5 ). Live stony coral cleaning stations mostly consisted of M. cavernosa, S. siderea and O. faveolata across disease zones ( Figures 4 , 5 ). M. cavernosa was the most frequently used coral in endemic sites, more than double the percent frequency of any other live coral species ( Figures 4 , 5 ). Use of sponges and gorgonians was greater in epidemic and endemic zones when compared to emergent zones while use of some major reef building corals, such as Orbicella spp. and C. natans, was less in the epidemic and endemic zones ( Figures 4 , 5 ). In this study, adult cleaner gobies were also observed to maintain cleaning stations on previously unreported or underreported coral species, D. labyrinthiformes and C. natans, which are highly susceptible to SCTLD, and two additional moderately susceptible species, O. faveolata and O. franksi (Whiteman and Côté, 2002, Whiteman and Côté, 2004; Florida Coral Disease Response Research & Epidemiology Team, 2018; Landsberg et al., 2020; Muller et al., 2020; Costa et al., 2021).

At the beginning of the study in 2019, O. franski contributed most to coral cover at most sites, followed by S. siderea,O. faveolata and P. astreoides. By the end of the study in 2021, O. franksi and P. astreoides contributed most to coral cover at most sites, followed by S. siderea, and O. faveolata. Note the reduction in contribution made to coral cover by M. cavernosa and C. natans at most sites and relative extirpation at all sites for C. natans and at sites where M. cavernosa made up a small proportion of the coral cover in 2019 (Seahorse Cottage Shoal and Magens Bay, Figure 5 ) (Ennis et al., 2020). Even though M. cavernosa contributed little to coral cover at most sites, it was one of the most chosen live coral cleaning stations by Caribbean cleaner gobies, followed by S. siderea. When taking available live coral habitat into account, S. siderea was less preferred than C. natans and M. cavernosa ( Figure 5 ).

M. cavernosa was the only coral species habitat that gobies preferred across all sites and both years. Goby occupation of different coral species varied across sites and changed through time ( Figure 5 ). The most prominent changes occurred in the emergent zone where several species that were occupied less than randomly in 2019 were found to be occupied more than random in 2021( Figure 5 ). Most of these changes represented an increase in preference with the exception of S. siderea at Fish Bay (emergent) and O. faveolata at Buck Island (epidemic), which decreased in preference ( Figure 5 ).

M. cavernosa was also the only species to experience ≧ 75% tissue loss on individual colonies at all sites and near total tissue loss (≧ 90%) at all but one site ( Figure 6 , Supplementary Figure 3 ). During the study, near total tissue loss (≧ 90%) of tagged coral cleaning stations following SCTLD incidence was lowest at endemic sites [Black Point (5%) and Flat Cay (8%)], but otherwise did not follow any discernable pattern ( Supplementary Figure 3 ). Of the nine M. cavernosa colonies to experience 90% tissue loss, only two were observed to be occupied by gobies following this loss and each of those had only one cleaner goby occupant after reaching 90% tissue loss. Tissue loss rates ≧ 75% on tagged coral heads following SCTLD incidence was greatest at emergent sites [Magens Bay (21%) and Fish Bay (23%)], and one epidemic site [Hull Bay (26%)] and lowest at endemic sites [Black Point (15%) and Flat Cay (8%)] and one epidemic site [Buck Island (16%)] ( Figures 6 , 7 ).

Over the course of the study SCTLD-related mortality resulted in a 3.6% decline in coral cover across study sites ranging from a loss of 10.8% at Seahorse Cottage Shoal to a gain of 2.6% at Fish Bay ( Table 1 ). Higher mortality rates of more susceptible species (i.e., C. natans, D. labyrinthiformis) changed the relative abundance of species across most sites and reduced coral species richness most notably at Black Point and Magens Bay (6 species each, 29% and 40% decline respectively) ( Figure 5 ).

Elacatinus spp. size ranged from 6 mm to 45 mm across all sites with highest abundance in the 11 mm - 15 mm TL size class ( Supplementary Figure 4 ). Combined data for rapid assessments and area surveys at monitoring sites found Elacatinus spp. sizes were not significantly different (F = 0.5, p = 0.62) among emergent (20.1 mm ± 0.85 SE), epidemic (18.4 mm ± 2.32 SE) and endemic (19.44 mm ± 1.17 SE) sites.

Goby size frequencies at study sites indicated recruitment of juvenile gobies occurred throughout the study ( Figure 3 ; Supplementary Figure 2 ; Supplementary Table 1 ) but peaked between the months of September and December at endemic and epidemic sites and in June at emergent sites and decreased again throughout the spring. ( Supplementary Figures 2 , 4 , 5 ; Supplementary Table 1 ).

All study sites experienced loss of tagged cleaning stations (≧ 75% tissue loss on an individual colony). M. cavernosa cleaning stations experienced the greatest loss in the epidemic zone and was the only species to experience ≧ 75% tissue loss across all zones. Similar patterns can be seen in colonies that experienced ≧90% tissue loss. The initial die-off, captured in the monitoring of this zone, in this study and others, of a variety of coral species caused the loss of cleaning station species diversity in the early stages of the disease on a reef (Brandt et al., 2021; Costa et al., 2021). Interestingly, the disease zone with the least live coral cleaning station richness was the epidemic zone. Corals in this area were actively experiencing disease but had already experienced whole colony losses. This occupation to a few coral species may be due to a concentration of gobies onto the most preferred, readily available, least compromised live coral habitat. These losses potentially restrict cleaning station options to less optimal coral species that may not be as easily recognized by clients.

The coral most occupied relative to its availability, and thus preferred by Elacatinus spp. cleaner gobies overall was M. cavernosa. In the epidemic zone, M. cavernosa made up more than half of the tagged cleaning stations and made up the greatest proportion of live coral cleaning stations in the endemic zone compared to all other species and was consistently preferred when taking available coral cover into account regardless of disease zone. S. siderea and C. natans were the most common cleaning stations tagged in the emergent zone. Both M. cavernosa and S. siderea are categorized as moderately susceptible to SCTLD while C. natans is categorized as highly susceptible (Florida Coral Disease Response Research & Epidemiology Team, 2018; Muller et al., 2020; Costa et al., 2021).

In zones where the disease was established, C. natans was occupied less than half the frequency (n = 1 coral at Black Point) seen in the emergent zone (n = 10 corals) This tagged colony was observed with a maximum of three gobies and did not experience tissue loss throughout this study. However, it did experience partial bleaching in October 2020 during which none of its resident gobies were found. A single adult goby (28 mm) was located on this coral in April 2021 during the last survey of this site,166 days after the coral had regained color. This species was consistently preferred with the maximum electivity index value at all site-year combinations except Fish Bay in 2019 (emergent) where it was occupied slightly less than expected by chance and at Buck Island STT where it was occupied as expected by chance in both 2019 and 2021 ( Figure 5C ). The abundance of this species at Black Point, Fish Bay, and Seahorse Cottage Shoal is so low as to be undetectable by TCRMP surveys in 2019 and at all study sites with available data in 2021. In 2019 C. natans made up 3.16% of the coral cover at Buck Island, St. Thomas, 7.04% at Magens Bay and 0.21% of the coral cover at Flat Cay (Ennis et al., 2020). This represents a substantial loss of some of the most chosen cleaning station habitat by cleaner gobies.

M. cavernosa may represent a more reliable, preferred option for cleaner gobies in comparison to highly susceptible species. This is supported by its presence and high levels of occupation at all sites as well as both a high preference value and an increase in electivity index value across sites. As other corals die, S. siderea may represent the next best option for Elacatinus spp. gobies to maintain a cleaning station. It also increased in preference value at two of four sites. S. siderea is also moderately impacted but rarely experiences acute tissue loss and affected colonies do not experience the same rapid mortality as other species (Florida Coral Disease Response Research & Epidemiology Team, 2018; Muller et al., 2020; Brandt et al., 2021; Costa et al., 2021). The presence of these two species on a reef may increase the stability of an actively cleaning goby population by providing alternative, more resilient alternatives to large brain coral species such as C. natans.

Isopods of the family Gnathiidae are the most common marine fish ectoparasites consumed by cleaner fishes and represent the primary food source for cleaner gobies in the eastern Caribbean region (Losey, 1974; Cheney and Côté, 2005; White et al., 2007; Côté and Soares, 2011; Artim et al., 2017; Nicholson et al., 2024). Although other microcarnivorous fish species will occasionally consume gnathiids, and even herbivores may ingest them accidentally, in our study area, cleaner gobies consume far more gnathiids per unit biomass or abundance than any other fish functional group (Artim et al., 2017; Nicholson et al., 2024) and thus appear to fill a unique functional role. The reasons cleaner gobies appear to associate strongly with certain types of live corals is unclear. One hypothesis is that they may be more conspicuous to potential clients compared to those on other substrates (Stummer et al., 2004; Cheney et al., 2009). When corals die or otherwise become unavailable, cleaner gobies can occupy other habitat and feed on other small invertebrates. However, changes in diet associated with such habitat-switches impacts their functional role. Indeed, several studies have shown that cleaner gobies associated with live coral spend more time cleaning and comsume more fish parasites than cleaner gobies occupying other substrate (Whiteman and Côté, 2002; White et al., 2007; Artim et al., 2017).

While cleaner gobies are well-known predators of gnathiids, whose cleaning behavior appears linked to the presence of live coral, recent studies have also shown that corals themselves can be major predators of gnathiids (Artim and Sikkel, 2013; Santos and Sikkel, 2019; Paula et al., 2021). A recent study conducted in the same region (and including some of the same sites) as the present study found that among 16 reef sites, live coral cover was the single best predictor of gnathiid density, with gnathiid densities reaching highest levels below 15% coral cover (Artim et al., 2020). Thus, diseases such as SCTLD could impact gnathiid isopod abundance through impacts of two of their major predators: corals directly and cleaner gobies indirectly.

While cleaner gobies have received increased attention in recent years, they are still much less understood than other, more habitat-specialized, goby species or the Indo-Pacific cleaner wrasse which fill a similar functional role. One key difference from the latter is the strong association to the benthos.

This study detected Elacatinus spp. cleaning stations on four previously unreported coral species, highlighting the need for a comprehensive study of habitat association and live coral preference, and the multitude of factors that influence cleaner goby recruitment and retention throughout their range. Moreover, to the extent that cleaner goby diets in association with substrate have been examined at only a few sites and in relation to a small range of habitat types, more studies on this association are needed.

Among the potential ecological impacts of cleaner gobies, the impact on microbial communities and dynamics is just beginning to be explored (Xavier et al., 2019; Narvaez et al., 2022; Sazima, 2023). Because of their association with live corals and their contact with many fish and fish species per day, cleaner gobies stand to serve as distributors of microbes between fishes and corals and thus could contribute to the spread of or prevention of and recovery from disease for both corals and fishes. Thus, more studies on the role of cleaner gobies in microbial dynamics are needed.

This study offers a glimpse into the potential indirect impacts of SCTLD on the wider ecosystem through an important species that impacts many members of the nearshore fish community. The loss or reduction of ecological services provided by these cleaners may lead to consequences for the health of other reef fish that rely on them which includes many commercially important species and may have consequences for the local fishing economy. The impact of this disease on the wider ecosystem and local fisheries are not well understood. More comprehensive assessments of such diseases on the wider ecosystem are needed to begin understanding the full impacts on the ecosystem beyond coral colonies and reef structure. Our findings have important implications for the science and practice of coral reef restoration that has thus far focused on a narrow range of fast-growing coral species that do not appear to be suitable habitat for cleaner gobies.