Few studies have focused on the impact of the rapids and waterfalls in bufeos in general and in
Inia boliviensis in particular. The first studies of bufeos in Bolivia [
1] consisted of informal surveys of various waterways, descriptions of behavior, and preliminary population-density estimates of bufeos in the Ibaré River. The authors concluded that the precipitous decline in population size was because of anthropogenic influences. They also concluded that the Bolivian bufeo was a separate species (
I. boliviensis) from the animals of the rest of the Amazon and Orinoco basins (
I. geoffrensis). They determined significant differences at the morphological and morphometric levels and identified two significant morphometric differences that are fundamental to understanding the possible origins of the bufeos.
I. boliviensis has 33 teeth on each side of its jaws, whereas
I. geoffrensis has 26–27 teeth and the average neurocranial volume is significantly less in
I. boliviensis (558 cm
3) than in
I. geoffrensis (665 cm
3). Following Williston’s law, the Bolivian bufeo should have a more primitive form than the current Amazon bufeo. These authors also claimed that an allopatric speciation process (vicariance barrier) split the
Inia forms. There is a 350 km section of the Madeira River, from Guayaramerín (Mamoré River, Bolivia and Brazil) to Porto Velho (Madeira River, Brazil), which contains 18 rapids and a collective drop of 60 m due to waterfalls such as Teotônio and Jirau [
141]. Two of these rapids were formally designated as waterfalls, due to accentuated differences in mean water-surface elevation above and below the rapids [
141]. The 900 m-wide Teotônio waterfall, which is the largest of the falls, fell 30 m over a span of 600 m. The Jirau waterfall, the second largest, is upstream of the Teotônio waterfall. It is 730 m wide, spans 1100 m [
141], and allegedly split the
Inia species. Most authors follow [
1] in supposing that the Teotônio waterfall delimits the geographic distribution of both species. Nonetheless, it was suggested that during the rainy season bufeos can cross the upper Madeira River rapids (including the Teotônio waterfall) [
7]. More recently, two studies [
10,
11] analyzed the question whether these rapids and waterfalls in Brazil actually separate the forms of
Inia. The first study analyzed mtDNA of 125 specimens of bufeos from the Madeira River basin sampled upstream and downstream of the 18 rapids on the upper Madeira River (in Brazil) [
10]. The authors observed that all individuals of this river up to almost the mouth of the Madeira River belonged to the species
I. boliviensis, save those near the population of Borba—870 km downstream of the Santo Antonio rapids and 890 km downstream of the Teotonio waterfall. The authors observed that gene flow was always upstream-to-downstream and the mtDNA of
I. boliviensis invaded at least half of the Madeira River. To study the impact of the Brazilian rapids on bufeos, the second study analyzed nuclear microsatellite markers of bufeos both upstream and downstream of the Brazilian rapids [
11]. They determined that there were likely two biological groups throughout the Madeira River, corresponding to the two species of bufeo,
I. boliviensis (from the Mamoré River and Madeira River to the locality of Borba) and
I. geoffrensis (from Borba to the mouth of the Madeira River with the Amazon River). Nevertheless, within
I. boliviensis, they detected two other groups (one between the Mamoré River and upstream of the Teotonio waterfall, and other between this last waterfall and downstream of the Madeira River). These authors determined that the pattern and direction of gene flow observed for microsatellites reflected that of the mtDNA. Between the Mamoré River and the Brazilian rapids region, gene flow was 9.83 in the downstream direction and 0.00 in the upstream direction. Between the rapids and the downstream localities, gene flow was 0.67 in the downstream direction and 0.19 in the upstream direction. Therefore, these rapids did not represent a complete physical barrier to movement and gene flow of bufeos.
4.1. Genetic Diversity and HWE in the Bolivian bufeos
The multiple genetic-diversity statistics we calculated for the diverse molecular markers we used offered a range of diversities. For the CR, the diversity was in the medium range, with some haplotypes dispersed across all the rivers studied (H1, H2) whereas other haplotypes were only specific to particular rivers. Ruiz-García et al. (2018) [
9] estimated the genetic diversity of the bufeos in different basins of South America, including Bolivia. That study had a smaller sample size for the dolphins from the Bolivian rivers than in the current study. Some of the genetic diversity estimates of that study [
9] were lower than those shown in the present study (six haplotypes vs. fifteen haplotypes; H
d = 0.52 vs. 0.63; π = 0.0014 vs. 0.0051; θ
per sequence = 1.08 vs. 7.83, respectively). However, in both cases, the Bolivian estimates were considerably lower, especially for the nucleotide diversity, than those obtained for the bufeos of other river basins (Orinoco basin; 9 haplotypes; H
d = 0.79; π = 0.025; θ
per sequence = 7.7; Amazon basin: 23 haplotypes; H
d = 0.71; π = 0.012; θ
per sequence = 12.42). This is the first evidence that the Bolivian dolphins were affected by some evolutionary event that diminished its genetic diversity in reference to other
Inia populations.
Nuclear microsatellites also yielded lower levels of genetic diversity for the
Bolivian bufeos (MANA = 4; H
e = 0.345) than for bufeos from the Peruvian Amazon (MANA = 6.2; H
e = 0.591) or the Orinoco basin (MANA = 4.8; H
e = 0.524) [
4]. Other Amazon bufeo populations had higher genetic diversities than did the Bolivian population. In Brazil, H
e = 0.59 was reported for two bufeo populations in the Negro River (Ariaú and Novo Airão, Brazil) [
142], and, also in Brazil genetic diversities of H
e = 0.54 and 0.57 for the Mamirauá Reserve and Tefé Lake, were found, respectively [
143]. Thus, for mtDNA and nuDNA (microsatellites), the
Bolivian bufeos showed significantly less genetic diversity than the
Inia populations from other South American river basins. We must recall that genetic diversity has been treated by the IUCN as a priority parameter for measuring the risk of extinction of a species, because it represents the substratum on which natural selection acts to promote the adaptation of a species to changing environments. The loss of genetic diversity can lead to loss of reproductive fitness and to a decrease in the evolutionary potential of a given species [
144]. Therefore, this parameter is crucial in helping us to determine if conservation measures are warranted for the
Bolivian bufeos. In contrast, the diversity for the
DQB-1 locus was medium–high, with the three most-frequent alleles well dispersed in all the Bolivian rivers studied (11 alleles; H
d = 0.857; π = 0.031). The diversity for the
Bolivian bufeos for this marker, in contrast to the other markers studied, was like other
Inia populations in the Peruvian Amazon (11 alleles; H
d = 0.873; π = 0.038) and Orinoco (9 alleles; H
d = 0.908; π = 0.054) (Ruiz-García, unpublished). Therefore, the
DQB-1 locus has a different evolutionary dynamic in
Inia than the other three kinds of markers we used.
For microsatellites, a significant excess of homozygotes was detected when all the Bolivian rivers we studied were analyzed collectively. This result could be due to genetic differentiation among certain populations of some rivers (the Wahlund effect). Nevertheless, when each geographical population was individually analyzed, only two out of eight deviated from HWE. This finding could be due to endogamy within these two populations or that there was a Wahlund effect within them. But, in most of the geographical populations considered, there were no deviations from the HWE. It is very likely that, for the DQB-1 locus, deviations from the HWE were not revealed. This agrees quite well with the fact that all the rivers we analyzed contained a unique major genetic pool for the Bolivian bufeos.
4.2. Genetic Heterogeneity
The phylogenetic trees and the genetic heterogeneity statistics for CR indicated that the rapids in the Mamoré River are not geographical obstacles for the dispersion of the Bolivian bufeos, even though these rapids are characterized by relatively fast water, deep channels, and steep banks, with little suitable habitat for Inia. However, there was evidence of isolation by distance where the rapids did not play any significant role. With CR, the genetic heterogeneity was created by the two most distant populations (El Azul in the Iténez-Guaporé River and the San Martín River) from the nucleus of the study area. Although the genetic heterogeneity detected was significant, the values of gene-flow estimates were high. The Kimura 2P genetic distances were small or very small among geographical population pairs, with the highest genetic distance for El Azul (the Iténez-Guaporé River) for CR. The same occurred for AMOVAs. The highest differentiation was between El Azul and the remaining populations. Using BAPS, three groups were detected with CR: two small populations (constituting a very-restrictive number of individuals and thus highly affected by gene drift). All the remaining populations were considered as one.
The nuclear microsatellites showed similarities and differences with that found with mtDNA. The most relevant similarities were as follows: (1) the existence of significant limited heterogeneity, mainly caused by the most differentiated population El Azul (Iténez-Guaporé River) being the second most differentiated population with respect to that from San Martín River (Iténez-Guaporé river basin). This is identical to what was determined with CR. (2) Gene flow among geographical populations is also high (2-mod analysis detected that gene flow is 99.9% more probable than gene drift), producing low levels of correct population assignment (up to 12–14 individuals were identified as migrants of first generation). (3) If we consider non-admixed models, the BAPS analyses showed six different groups for individuals and three different groups for populations. But even in this last case, the three populations were separated by the Matacaré, Warnes, Mayo-Mayo, and Envidia rapids, yet they were considered as a single unique population because the geographical barriers did not seem to affect the genetic structure of the bufeos of this area. Thus, as with the mtDNA, the rapids did not play any role in the heterogeneity of these river dolphins in the Mamoré River. If we considered mixed models, only one unique gene pool was determined, indicating the non-significance of the rapids in shaping the genetic structure of this species in the Bolivian rivers. The main difference regarding the mtDNA is that the nuclear microsatellites did not detect the same strong and significant isolation by distance, as detected by the mtDNA. This difference can be explained by the differential gene flow of females and males. It is interesting to discuss the different results obtained with mitochondrial and microsatellite markers. The mt markers seem to be especially influenced by their neutrality and of the dynamic movement of the females and the microsatellites seem to be of neutral nature and clearly influenced by male movements. In reference to the DQB-1 locus, six alleles were shared with Inia specimens from the Colombian–Venezuelan Orinoco and the Peruvian Amazon, while five alleles were exclusive to the Bolivian rivers. However, these five alleles did not show any kind of genetic structure. The fact that some DQB-1 alleles were shared with other river basins from other countries, while the mtDNA haplotypes were unique (private) to the Bolivian rivers, is indicative of the fact that the first alleles were probably submitted to similar natural-selection pressures by microorganisms, pathogens and parasites in different rivers of South America. The mtDNA haplotypes showed a much more neutral dynamic. BAPS analysis determined four groups at the DQB-1 locus, but in no case did the rapids show any influence in the groups detected. Henceforth, for the mtDNA data set and for the two kinds of nuclear markers we used, the rapids of the Mamoré River did not play an important role in the genetic structure found in the bufeos of the Bolivian rivers.
Our results can be compared with those obtained by other authors [
10]. They estimated relatively low F
ST values for mtDNA, but they indicated that the haplotypes had non-random distribution (F
ST = 0.169;
p < 0.001) between the upstream and downstream groups of
I. boliviensis separated by these Brazilian rapids. In our case, we detected values of F
ST slightly higher (F
ST = 0.25 for CR), but the estimates of gene flow were elevated sufficiently to connect the populations of dolphins on both sides of the rapids. Henceforth, they concluded that the rapids do not delimit the distribution of
I. boliviensis upstream and
I. geoffrensis downstream of the rapids, as previously hypothesized. Our results agree well with the fact that rapids do not influence the genetic structure of the species studied. It should be noted that some of the rapids in Brazil are considerably larger than those tested in Bolivia. However, these authors [
10] determined the existence of gene flow using IMa2 and concluded that gene flow was uni-directional, moving from an upstream to downstream direction. In contrast, we determined the existence of gene flow in both directions for all the rapids analyzed in the Mamoré River in Bolivia. This finding is probably because these rapids were smaller than those in Brazil. In this last country, it was found that 15 haplotypes were restricted to the upstream group while 24 haplotypes were restricted to the downstream group, with only two haplotypes shared between the two groups [
10]. Excluding rare haplotypes, singletons recently appeared via a mutational process and did not yet have time to expand geographically from their area of origin. We observed the same in Bolivia, with some haplotypes extensively distributed in the different Bolivian rivers we analyzed. Other haplotypes—those that were generated more recently—were restricted to specific points of the area studied. The Brazilian researchers [
10] observed bufeos moving across the greatest of the Brazilian rapids in the Madeira River during high-water season, a time when several fish species have their annual spawning migrations [
145]. Most of the rapids are not submerged during the dry season and some are not even submerged during the rainy season. Additionally, these Brazilian authors [
10] commented that during periods of extremely high water, such as what occurred in 1996, 2006, 2008, 2009 and 2012, all rapids become submerged, permitting unhindered movement of animals. Although extremely high water potentially facilitates the bi-directional movement across the rapids for the bufeos, it also increases the velocity of the river. The deep channels with steep banks and high-velocity current characteristic of the rapids’ region become even more of a barrier for upstream movement of dolphins. Yet, the same phenomenon probably facilitates downstream movement and gene flow of this species. Periods of extremely high waters, therefore, most likely facilitate uni-directional connectivity between the upstream and downstream groups of river dolphins. In the Bolivian rivers, we observed bi-directional movement across the rapids even in the dry period, because the rapids are not as big as some from the Madeira River. For this reason, the gene flow in the Bolivian rivers is bi-directional across the rapids. These bufeos are capable of sustained swimming in fast waters for short periods of time, after which they need to recuperate [
146]. The animals use short bursts of energy to cross the rapid, followed by a period of rest in an area with little or no current [
10].
Another work [
11] also examined microsatellite markers, when these Brazilian authors studied the impact of the Brazilian rapids on bufeos upstream and downstream. Using STRUCTURE, they determined that the most likely number of biological groups throughout the Madeira River was two, corresponding to the two species of bufeo,
I. boliviensis (from the Mamoré River and Madeira River to the locality of Borba) and
I. geoffrensis (from Borba to the mouth of the Madeira River with the Amazon River). A principal component analysis separated out specimens from upstream of the rapids (above Guajará-Mirim, Brazil, or Guayaramerin, Bolivia), between the rapids’ region upstream of the Teotônio waterfall (between Fortaleza do Abunã and the Teotônio waterfall), and between the Teotônio waterfall and the Santo Antônio rapids. All these specimens had mtDNA typical of
I. boliviensis; however, the Teotônio waterfall separates the nuclear genomes of
I. boliviensis, upstream, and
I. geoffrensis, downstream. Thus, this barrier is partially porous, allowing low levels of unidirectional connectivity, upstream to downstream. A principal component analysis also separated out specimens (considered
I. boliviensis ×
I. geoffrensis hybrids in the Madeira River) downstream of all the rapids and specimens from the Mamirauá Reserve on the Japurá River (used as control specimens of
I. geoffrensis). Two specimens sampled below the Teotônio waterfall were genetically assigned to the between-rapids group. Most likely, these were first-generation migrants, which means that they could cross this geographical obstacle. Nevertheless, using the BayeAss 3 software, it was found they had the highest migration rates within each group, but also relatively high migration rates between the region upstream of all the rapids, and the rapid region upstream of the Teotônio waterfall, and between the downstream region and rapid region downstream of the Teotônio waterfall. Our results showed that the different rapids present within the Bolivian Mamoré River have less impact on the genetic structure of
I. boliviensis than rapids and waterfalls in the Brazilian territory, because microsatellites showed that the Bolivian rapids neither decreased nor restricted the gene flow to the contiguous populations. For instance, El Corte and El Azul (the most differentiated of all populations studied with microsatellites) are separated by the San José rapid. However, El Corte was more closely related to El Azul than any other population, although they are separated by an important rapid area. Furthermore, microsatellites showed that gene flow was bi-directional (as occurred with mtDNA) between populations separated by rapids in Bolivia. Therefore, the power of the Mamoré River does not affect the ability of the bufeos to cross the rapids in either direction. This finding is different from what was detected for the Madeira River. These differences are reflected in the BAPS analysis, assuming admixture (the real situation), which only detected one unique population of bufeos in the Bolivian rivers and multiple different groups in the Madeira River (STRUCTURE, principal component analysis). This Bolivian “gene pool” is the same as that the Brazilian authors [
11] detected upstream from the Brazilian rapids. Another interesting point is that the bufeos belonging to different river basins in Bolivia could be closely related. This is the case of the close genetic relationship between the bufeos of the confluence of the Ibaré and Mamoré Rivers area and those at the Ipurupuru River. The Ipurupuru River, although it belongs to the Iténez-Guaporé river basin, is geographically near the confluence of the Ibaré and Mamoré Rivers. Therefore, in high-water times, when the forest floods, bufeos can move from rivers of different basins, allowing genetic homogenization of bufeo populations across these rivers. Another difference between the Brazilian works [
10,
11] of Gravena et al. and our study is that they observed a much greater genetic differentiation among the groups determined above, between, and down-stream of rapids for microsatellites, than for mtDNA. In contrast, we found a greater and clearer genetic heterogeneity in the Bolivian rivers for mtDNA (CR: G
ST = 0.264; F
ST = 0.251) than for microsatellites (G
ST = 0.122; F
ST = 0.094; R
ST = 0.063). Our results are consistent with females staying near reproductive areas (female philopatry), whereas males disperse more widely, in search of breeding opportunities in faraway areas. It may also be possible that females inhabiting areas with powerful rapids [
10,
11] are moving more actively than males looking for good places to breed or feed. An alternative hypothesis is that the mtDNA showed more ancestral patterns of colonization when rapids and waterfalls of the current Bolivian rivers were higher or more prominent during Pleistocene dry periods. Furthermore, microsatellites showed patterns of more-recent colonization processes. Clearly, the current rapids from the Bolivian rivers are minor and less powerful (less of a geographical barrier to gene flow) than the Brazilian rapids and waterfalls (a bigger barrier to gene flow). In fact, our study’s higher genetic heterogeneity for mtDNA than for microsatellites agrees quite well with that reported in other studies with cetaceans and disagrees with the results of the Brazilian authors [
10,
11]. In the North American beluga whales, for example, the γ
ST values (0.047–0.072) for microsatellites were significantly lower than the γ
ST value obtained for the CR (0.409) [
147]. Similarly, it was found that female porpoises were more philopatric than males [
148], and it was also shown that there was a highly male-biased dispersal among
Phocaena dalli populations [
149], as we found for
I. boliviensis.
The second Brazilian work [
11] considered that
I. boliviensis can be divided into two distinct management units (MUs) [
150] based on mt and microsatellite markers. In the Bolivian area we studied, however, only one MU should be proposed as being one of the MUs detected by these authors [
11]. However, other potential MUs could exist in other areas of the Bolivian rivers which were not included in our study. Although some studies have strongly emphasized that
I. boliviensis was a separate species from
I. geoffrensis [
3,
4,
8], the results of [
10,
11], as well as those from [
9], raise doubts as to this idea, at least based on the Biological Species Concept (BSC) [
151,
152]. The mtDNA genome of
I. boliviensis is in
I. geoffrensis occurring in the Madeira River below the Teotônio waterfalls, and the nuclear DNA genome of
I. geoffrensis have introgressed into
I. boliviensis in the Teotônio waterfalls. This shows that pre-zygote reproductive-isolation mechanisms were not developed between
I. geoffrensis and
I. boliviensis as we would expect if the BSC were applied. Some authors [
11] affirmed this by stating that “While the two species do exist (
I. geoffrensis and
I. boliviensis), there is an extensive hybrid zone in the Madeira River. The hybrid zone appears to be ancient and is characteristic of a region of introgressive hybridization. There appears to be no physical barrier between the hybrids and
I. geoffrensis, bringing into question how and why the hybrids persist”. The question is easily answered using the BSC: they are a unique bufeo species.
4.3. Spatial Genetic Structure
The different types of molecular markers we used in
I. boliviensis allowed us to detect different spatial genetic structures. For the mtDNA, Mantel’s test determined that 13% of the genetic distance found was explainable by isolation by distance. Similarly, the spatial autocorrelation analysis of the sequences of the individuals showed significant monotonic clines related to isolation by distance (however, when only haplotypes were used no spatial structure was detected). AIDA of the CR detected genetic patches with a diameter of around 30 km, and later isolation-by-distance. GLIA detected the sample from San Martin River (Iténez-Guaporé river basin) as the most differentiated of all the samples. The results are consistent with female philopatry. In other species of marine dolphins such as the bottlenose dolphins, patterns of population structure across very small distances are commonly described [
153,
154]. Other authors determined fine-scale population structure within the Adriatic Sea for bottlenose dolphins consistent with local reports of strong site fidelity [
155,
156]. The female site fidelity is observed elsewhere in European waters, such as the Moray Firth (Scotland), the Shannon Estuary (Ireland), and the Sado Estuary (Portugal). In all locations, small populations of females show strong site-fidelity to semi-enclosed bays, with limited interaction with populations outside the bays [
157,
158]. Many studies have demonstrated that sexual differences in dispersal trends exist [
159,
160,
161] and that this affects the genetic structure of the species. For example, some studies [
159,
160,
161] demonstrated that migrant sex is the vector of gene flow and a factor of global genetic homogeneity. Philopatric sex introduces internal homogeneity within the lineages (at least promoting gene correlation between individuals in the same lineage or close geographic lineages) and genetic heterogeneity among the most distant lineages. One study [
5] used Chesser’s social models to determine asymptotic values of the F-statistics. It was showed that bufeos in the Peruvian Napo and Curaray rivers have a social reproductive system. Based on their findings, the basic genetic lineages could be composed of seven reproductive females per lineage in each breeding period (with high site fidelity) and the number of reproductive males per lineage is not important. Furthermore, on average, within each lineage there was a reproductive male with four females, and thus polygyny exists within this dolphin species (with some migrating males, although others presenting site fidelity). There was also a probability of 0.30 that females of the same lineage choose the same male for breeding. Although conditions could be different in the Bolivian rivers, these facts [
5] agree with the genetic results in the current study, specifically regarding the philopatry of the bufeo females that was detected in spatial genetic analyses with mt markers.
In contrast, nuclear microsatellites did not detect isolation by distance and even the two first DCs, in the autocorrelation analyses, were significantly negative. This finding could agree with the fact that, while the females were philopatric, some males migrated, on average, more than 30 km to successfully breed. Therefore, the population dynamics of females and males were different. This was due to habitat preference and reflected in the spatial structure of the molecular markers. Male bufeos are primarily found in the main channels of large rivers, whereas females prefer more protected areas, such as flooded forests and its channels and lakes [
142]. Several studies have determined the minimum home ranges of bufeos. These estimates varied from 12.1−27.9 km
2 (Juruena River, Brazil) to 60.5−233.9 km
2 (San Martín River, Bolivia). Some of the distances travelled by individual male bufeos are worth mentioning, such as (1) 7.5 km in the Juruena River, (2) 36.2 km in the Amazon River (Colombia and Perú), and (3) 51.8 km in the Orinoco River (Colombia and Venezuela). Some authors [
162,
163] reported 225 km for one bufeo in the Japurá River (Brazil). The distance travelled of four individual bufeos was also determined (mean of 61 km with a maximum of 220 km) [
164]. Likewise, it was detected that there were movement distances ranging from 50 to 200 km for bufeos in the Cuyabeno and Lagartococha rivers (Ecuador) [
165]. The longest total distance traveled by a specimen was recorded by a male in one of the Bolivian rivers where we sampled bufeos. It was 297.9 km [
166]. In all cases, the maximum distances were carried out by males. Long fish migrations (bufeos predated more than 43 fish species, many of them with migratory patterns) from 500 km up to 3000 km [
167] can potentially explain the relatively long-distance values recorded in tagged bufeos. It was shown, through satellite tracking, that the Amazon bufeo traveled between 20 and 100 km per day, whereas other dolphins, generally females, stayed for weeks in a small (1 km
2) area [
163]. Using photo identification, it was reported that the average range of movement detected for bufeos in Bolivia is at least 60 km, or 3 to 10 km daily [
164]. River dolphins are highly mobile top predators that can cover long distances (hundreds of kilometers) in relatively short periods (days), with a possible differential use of habitat by males and females [
162,
163,
166,
168,
169]. In fact, typical daily movements of up to 20 km, and individuals swimming at sustained speeds of 3–6 km/h were reported [
170].
On the other hand, the sequences at the DQB-1 locus did not show any spatial trend for any of the procedures used. This could mean that selective pathogen and parasite pressures are very similar throughout the Bolivian rivers, as well as for other Neotropical rivers (Perú and Colombia–Venezuela) where DQB-1 alleles were shared. Henceforth, spatial structure analyses could help to determine the different evolutionary events which occur for different kinds of molecular markers.
4.4. Demographic Changes
The MJN procedure revealed that the diversification of the mtDNA haplotypes in the Bolivian rivers studied began around 180,000–130,000 YA (for CR). We know that selection can affect effective population size, reducing the effective number for a time and increasing the coalescence rate later, which could increase “artificially” the temporal splits among populations. The same occurs with small changes in the mutation rate (μ), which can greatly affect the effective number and, in turn, the estimated divergence time. For instance, if the estimated mutation rate is smaller than the real one, the temporal-divergence split will also be smaller than the real one for a determined effective size. Contrarily, if the estimated mutation rate is higher than the real one, this increases the temporal-divergence estimates in an unreal way. Despite this, our temporal mtDNA diversification process agrees quite well with that detected by another work [
10]. They estimated isolation between the upstream and downstream groups of bufeos across the rapids of the Madeira River to have occurred around 122,000 YA (95% HPD 32,000–283,000 YA). They performed an IMA2 analysis using a harmonic mean of the average mammalian substitution rates. When they used the upper- and lower-bound substitution-rate values from another study [
171], their estimated values ranged from a mean of 97,000 (95% HPD 25,000–226,000 YA) to 163,000 YA (95% HPD 43,000–379,000 YA). Henceforth, both studies concluded that at the beginning of the fourth great Pleistocene glaciation, the rapids of the Mamoré–Iténez (Guaporé) River basin and the Madeira River basin influenced the diversification of mitochondrial haplotypes in
I. boliviensis. This divergence was not due to the formation of the upper Madeira River rapids, but rather, a different event. The formation of the rapids most likely resulted from the rise of the Fitzcarald Arch in the mid-to-late Pliocene, around 4 MYA [
172,
173]. The period of mtDNA haplotype differentiation, which occurred around 170,000–100.000 YA, coincided with beginning of the last large Pleistocene glacial period (transition from the Middle to the Late Pleistocene) [
174]. This could be connected to the haplotype diversification detected in the bufeos of the Mamoré and Madeira rivers. For instance, although 27 possible major climatic changes with a periodicity of 100,000 years during the Pleistocene in northern South America were defined [
175], there only is clear evidence for the last glaciation [
176,
177]. This last cold period began an estimated 116,000 YA, which is the same time of estimated divergence for many of the pairs of haplotypes of
I. boliviensis. Probably, these cold periods substantially decreased the level of the waters in the rivers of the Mamoré–Iténez (Guaporé) and Madeira basins. Under conditions of very low waters, the rapids become a partial barrier to the dispersion of the dolphins within this geographical area and affected the hydrological dynamics of these rivers. Additionally, the reticulations observed in the MJN agrees well with this scenario. Some lineages of bufeos should be isolated in several riverine refuges when the levels of water decreased during the last phase of the Pleistocene and accumulated unique mutations, but when the waters rose again some lineages expanded and widely dispersed their genetic characteristics across the major fraction of the Mamoré River basin. For this reason, we found exclusive and private haplotypes, together with well-expanded haplotypes, in the same population.
All the statistics used to detect demographic changes with mtDNA detected significant population expansions, except for the mismatch distribution. Bolivian River dolphins showed considerably lower mtDNA diversity than did
Inia populations from other basins (Amazon and Orinoco). Based on statistical analysis of the mtDNA marker, it seems clear that the Bolivian population crossed an initial bottleneck with posterior population expansion. BSP for CR detected a female population decrease which began around 60,000 YA, with a strong dip around 20,000 YA followed by an increase in the last 500 YA. Other marine animals showed a strong decline 20,000–12,000 YA (Last Glacial Maximum, LGM) with a subsequent population expansion. The killer whale population (
Orcinus orca) declined during LGM and then expanded [
178,
179]. Similarly, the low diversity observed in loggerhead sea turtles in Eastern Mediterranean [
41] has been linked with excessively low temperature for successful hatching during the LGM, which would imply that present loggerhead turtles are descendent from post-glacial colonizers. Additionally, it was suggested that differentiation of certain estuarine/bay populations of bottlenose dolphins in the North Atlantic has been achieved post LGM [
180]. This suggests that the LGM might have had profound effects not only on the Mediterranean, North Atlantic, and Pacific marine fauna but also on the bufeos of the Bolivian rivers.
The nuclear microsatellite results were complementary to those obtained with the mtDNA. The Kimmel test [
121] detected an initial bottleneck with posterior population increase. An interesting result from this analysis is that if the value is β > 1 (ln β > 0), then it will be present during several thousand generations (from 5000 to 10,000 generations) before these values show the signature of a population expansion (β < 1 (ln β < 0)). If, in the case of bufeos, a generation is about seven years, then the population expansion process after the initial bottleneck is not older than 35,000 Y (5000 generations; more probable) or 70,000 Y (10,000 generations; less probable). This could mean that more than one bottleneck has occurred throughout the natural history of
I. boliviensis. Also, other tests [
123,
127] detected an intense bottleneck during the foundation of the population. The Zhivotovsky et al. test [
123] is more powerful than the Kimmel test [
121] for detecting strong bottlenecks in the initial formation of a population, or during its history. It has a greater potential for detecting when populations reach a demographic stability after a rapid population growth. The values of S
k and V for this test over time weakly depend on the rate of population increase and the final population size. Only extreme differences in the growth rate could produce substantial differences between these statistics. The Bolivian bufeos presented extremely different S
k and V estimates. The negative S
k value and the small V value demonstrated the existence of a strong bottleneck in this population. The question is whether this bottleneck occurred just before or during the population expansion. This last event may greatly influence S
k but only slightly affect V (
Figure 6 from [
123]). In the Bolivian bufeo population, both S
k and V were extremely affected. Therefore, the striking bottleneck was in the original Bolivian population. This result disagrees with previous hypotheses [
181], which claimed that the original Bolivian population gave origin to all other bufeo populations in the Amazon and the Orinoco basins. An extreme bottlenecked population cannot generate other populations with more genetic diversity than itself. On the other hand, S
k is not affected by different mutation rates in the microsatellites (at least if this rate does not change over time), whereas V is greatly affected. Therefore, because all the microsatellites studied were dinucleotides, different S
k values could not be attributed to different mutation rates. Also, we have no evidence that the mutation rate changed, at least when the population began to grow. If the mutation rate had increased, then both S
k and V would have been higher than the values found. The k and g tests [
129,
130] yielded no evidence of significant population expansions for the Bolivian river dolphins. This showed that these tests have less power to detect population changes than do the previously quoted tests.
These demographic changes, together with the low levels of genetic diversity herein reported for the Bolivian bufeos, indicate that the current
I. boliviensis originated as a small founder population split off from the central Amazon into the upper reaches of the proto-Madeira River. The population became isolated via a geomorphic event that prevented gene flow with other
Inia populations for a considerable amount of time. This could have begun around 4–3 MYA [
172,
173]. This explains the low levels of gene diversity found in the three types of different molecular markers we used in this study. This event is likely associated with the migration of rivers draining the Andean front in response to eastward propagation of the Andean fold and thrust belt [
182,
183]. The modern upper Madeira River has migrated to the westernmost edge of the resistant Brazilian shield, which now constrains its course, as can be clearly seen by its unique NW–SE flow direction. The flow of the Madeira over the rocky western edge of the shield creates the barrier isolating
Bolivian bufeos from the mainstem of the Amazon basin. This first event agrees with the estimated temporal splits between the ancestors of
I. geoffrensis and
I. boliviensis (2.6–6.2 MYA, [
3]; 1.5–3.8 MYA, [
9]; 3.1 MYA, [
143]; and 2.87 MYA, [
184]). The history of population changes after this founder event is recorded in the mitochondrial sequences and in the nuclear markers studied. Around 180,000–100,000 YA, at the beginning of the fourth great Pleistocene glaciation (and maybe with the influence of the Eemian interglacial period, around 150,000 YA) climatological changes (more than geological changes) decreased the depth of the rivers within the Madeira basin. With very low waters, the rapids became population barriers, and possibly helped to create different mt haplotypes, which later expanded throughout the Bolivian rivers. Once again, during the LGM, another population reduction occurred, followed by another small increase over the last few centuries. For this reason, different tests for mt and nu markers detected different population bottlenecks and expansions for
I. boliviensis. These results invalidate the hypotheses proposed by Grabert [
181] for the origin of
Inia in South America. This hypothesis proposed a Middle Miocene Pacific entrance prior to the uplift of the northern Andes. A Pacific coastal Iniidae entered a large Bolivian lake (sub-Andean freshwater molasses) during the Miocene (15 MYA).
I. boliviensis then later appeared in this molasse lake during the Pliocene, around 5 MYA. It became highly adapted to the turbidity of these waters. Sometime afterwards, approximately 1.8 MYA,
I. boliviensis entered the Amazon basin through the Purus or the Iquitos gates, eventually causing the appearance of
I. g. geoffrensis. Later, about 10,000 YA,
I. g. humboldtiana formed from
I. g. geoffrensis within the Cassiquiare channel. This hypothesis is totally rejected. Therefore,
Inia boliviensis could not be the original form of
Inia. Its smaller cerebral capacity and greater number of teeth are not considered ancestral characteristics in
Inia. In contrast, they are derived characteristics, which could represent a violation of Willingston’ law. A strong gene drift and the prolonged bottleneck effects could change specific morphological characteristics following a genetic revolution [
152] or a “flash-crash” process [
185,
186].
The bufeo is considered an ideal biological model in helping us to understand the dynamics of aquatic ecosystems because it has a small population size and high habitat requirements, and differentially uses its habitats. Moreover, bufeos are top predators, have long gestation periods, and demonstrate long-term parental care. Therefore, bufeos are good bioindicators of aquatic environments because they are sensitive to the environmental changes [
187]. It was determined that there was a decrease of 35.7% in the number of bufeos in a population in the Orinoco (Colombia) over six years [
168]. For the Colombian Amazon, it was reported that there was a reduction of 10% per year in a populations of
I. geoffrensis between 1993 and 2007 [
188]. These reductions can be explained by the increased number of bufeos captured and used as bait for the mota (
Calophysus macropterus) fishery. Bufeos need environmental protection. Molecular-population genetic studies, like the current one, provide critical information in helping us to determine the genetic history and diversity of the bufeo. The use of different kinds of molecular markers is relevant, as we show in the current work, because some of them are considered to have a neutral behavior (mt CR, which is illustrative of the female colonization process, and nu microsatellites, which detected the important vector of gene flow which the males have), but others (
DQB-1 locus) clearly respond to pathogen selective pressure, which is similar in different Neotropical rivers. Such information is crucial to developing a meaningful conservation plan.