The dorsal renal epithelium is the largest portion of the kidney and shows conspicuous hemocyte islets in the hemocoelic spaces between the epithelial crypts (41). A field view of the dorsal renal epithelium of P. canaliculata is shown in Figure 5A . These hemocyte aggregates are a constitutive part of the kidney of healthy individuals and are composed of agranular cells (hyalinocytes, agranulocytes) and just a few granulocytes ( Figures 5B–D ). The islets are interposed in the blood drainage of this major part of the kidney (29) where they are loosely arranged and anchored by basal extensions of the renal epithelial cells. These epithelial extensions show no cellular junctions, which may facilitate the in-and-out hemocyte migrations, while the entire islet remains in place because of the network of basal epithelial extensions ( Figures 5C, D ) (29). Also, some of the circulating progenitor cells (32) may be caught in that network. Many hyalinocytes in the islets, but not granulocytes, contain non-membrane-bound cytoplasmic areas that are likely glycogen deposits and may have a role in storing this energetically valuable substance ( Figures 6A–C ).

Similar hemocyte aggregates have been observed in other gastropods since the initial observations of Perrier in L. littorea (58), and though under different organ names, there were always associated with reno-pericardial derivatives (59–61).

The area of the constitutive hemocyte aggregations in the kidney significantly increases after an immune challenge (inoculation of yeasts, Saccharomyces cerevisiae) to P. canaliculata (29), that may indicate either attraction of hemocytes to these aggregations or the generation of new cells within the islets, i.e., hematopoiesis, as it will be reviewed in section 4.2 . Also, hemocyte nodules are formed as outgrowths of these islets ( Figure 7 ), sometimes dislodging the renal epithelium, so that the nodule becomes in direct contact with the cryptal lumina and thus with the renal cavity. In P. canaliculata, we have substantiated the existence of internal “immune barriers”, referring to those organs that are able to prevent the dissemination of intruders, either because of its position in the circulation, or because their intricate and stagnant microcirculation, which would increase the probability of hemocyte contact with the intruders (29). These organ barriers are here called “internal” to distinguish them from the large “external” organ barrier formed by the shell and the integument (section 4.1.1 ).

In P. canaliculata, the kidney is an immune barrier with constitutive hemocyte aggregations, which makes it an analog of lymph nodes in mammals. Indeed, these nodes are interposed in the lymph flow, which stagnates in their sinuses as the renal hemocyte islets are interposed in the blood flow of the ampullariids. Another analogy is that the lymph of mammals and the blood of ampullariids both find a locally high concentration of immunocompetent cells when confronting the barrier, which makes them more effective.

It has been reported (62) that repeated bleeding (2 mL each) within two consecutive days did not significantly affect the circulating hemocyte concentration in P. canaliculata, which would imply that capacious hemocyte reservoirs should exist in this snail, that would be able to release hemocytes to the circulation, thus contributing to homeostasis.

Constitutive hemocyte islets found in the kidney ( Figure 5 ) (41) are candidates for this function in P. canaliculata because they are major hemocyte aggregations that could contain sufficient cells to significantly affect the circulating concentration after hemocyte release. These islets are contained within the hemocoelic draining spaces of the kidney, where they are retained by a meshwork of basal extensions of the renal crypt epithelium, but with no intervening intercellular junctions ( Figure 5 ) (29). Also, some of the circulating progenitor cells (32) may be caught in that network and would be the basis of the role of these islets as hematopoietic sites (see section 4.2.3.2 ). Likewise, in vitro observations of hemocyte aggregates showed a dynamic exchange of cells by diapedesis, between the aggregate and the immediate environment (see section 4.1.3.3 ). By diapedesis, the hemocytes may also leave the islet traversing the mesh of epithelial projections that hold the hemocyte islet in place. Afterward, the hemocytes can be carried away with blood drained from the kidney, which is collected by the efferent renal vein and goes directly to the heart auricle and hence to the general circulation. It would be crucial, however, to determine the release of marked cells from the islets in response to peripheral depletion of hemocytes.

Alternatively, it has been proposed (62) that the aortic ampulla would serve both as a hemocyte reservoir and as a site of hemocyte maturation (the ampulla is an extensible outgrowth of the pericardial portion of the anterior aorta, and this small organ is a synapomorphy of the Ampullariidae). We think that the proposition of the ampulla as a reservoir was based on a misinterpretation of the storage cells’ nodules occurring in the ampullary walls (63). In fact, the latter study combined light and electron microscopies, subcellular fractionation, and biochemical determinations, and could indeed show that the cytoplasm of these large cells is loaded with urate crystalloids. On the basis of this study one may reinterpret the light microscopy observations of (62) as follows: (a) the “white spots” shown in their Figure 3A would correspond to the nodular urate tissue described in (63); (b) the “compartmentalized lobules” would correspond to the large storage cells whose cytoplasm is extracted in light microscopy preparations (63); and (c) the stocked “hemocytes” would correspond indeed to the nuclei of these storage cells, which may still retain a cytoplasmic rim. These reinterpretations were further supported by a recent TEM study of the same group that confirmed the occurrence of storage cells and crystalloids, but no hemocytes, in the walls of the aortic ampulla (64).

Regarding the proposed role of the aortic ampulla as a site of hemocyte maturation (62), we think it was proposed not as a formal hypothesis but as a mere possibility, and it has no empirical support. Curiously (and wrongly), the same idea has been attributed to some of us in a rather recent paper (65), but it has not been mentioned in another recent study by the same group (64) and so it was probably abandoned.

In principle, however, contingent hemocyte aggregations can occur anywhere in the body, but are certainly facilitated in the internal immune barriers, such as the lung in P. canaliculata, an organ that lacks constitutive hemocyte aggregations, but where large hemocyte nodules are also formed in response to yeast injection (29). Indeed, nodule formation in the lung relies on the stagnant circulation of the respiratory lamina, an extensive network of blood sinuses serving gas exchange (29). Therefore, nodulation in the lung follows a somewhat different pathway, because small hemocyte aggregations form within the large vessels supplying the organ, and these aggregations are later dislodged and clogged in the narrower sinuses of the respiratory lamina, where they may grow by the local generation of hemocytes or by the attraction of circulating or wandering ones (29).

Cueto et al. (15) developed a method for the primary culture of circulating hemocytes of P. canaliculata, which may be interesting to study nodulation. Hemocytes in those conditions showed the evolutionarily conserved behaviors of attachment, pseudopodia emission, and phagocytosis. Real-time, time-lapse videomicroscopy showed the dynamics of the early hemocyte aggregations, with cells both entering and leaving the nodules ( Figure 8A ). Besides, these cells showed a marked tendency to form large nodular aggregates that may later merge in masses that could be visible with the naked eye ( Figure 8B ).

Both the large single nodules and the merging ones detach from the bottom surface and float freely in the medium, while migrating cells emitting lamellipodia and filipodia occur near their surface ( Figures 9A, B ). The nodules may show a central compact zone, an intermediate zone with large extracellular lacunae and an outer zone of flattened cells (15). Dual DAPI/propidium iodide staining revealed the coexistence of nodules with varying proportions of viable and non-viable cells ( Figures 9C–H ). Their formation in axenic culture parallels the in vivo generation of hemocyte nodules in response to immune challenges ( Figure 7 ) (29), although an obvious difference is that nodule formation in axenic cultures could only be elicited by some kind of contact factor.

Nodular hemocyte formations are known to occur in the major clades of Protostomia (e.g., 66–69) and Deuterostomia (e.g., 70). A particular case among the latter are the nodules formed in the zebrafish larvae (Danio rerio, Teleostei, Cyprinidae) in response to Mycobacterium marinum inoculation, because these nodules (usually called granulomas or tubercles in the context of tuberculosis studies) have received a particular attention as a model of mycobacterial pathogenesis (71–75). Since the zebrafish larvae lack adaptive immune system early in ontogenesis, it is thought that the initiation of nodulation can occur without any contribution of the adaptive immune system (72). The same can be said of nodules induced by different stimuli in P. canaliculata and other invertebrates, where an adaptive immune system would never develop. Also, Ramakrishnan (76) elaborated on the idea that though the tuberculous granuloma was traditionally thought as a protective structure committed to wall-off and destroy the tuberculous bacilli, they may also provide a suitable environment for the multiplication of this intracellular parasite and also, that the in-and-out migration of macrophages from the nodule would result in the dissemination of the infection, rather than on its contention. Similar in-and-out hemocyte migration has been observed during the early phases of nodule formation in P. canaliculata (15) (see section 4.1.3.3 ). Therefore, Ramakrishnan’s ideas should be considered while exploring the nodular reactions to intruders in gastropods and other invertebrates.

Besides nodule formation, diffuse hemocyte infiltration of the connective tissue has been reported in M. cornuarietis in response to A. cantonensis infection (13), but this was not frequent in P. canaliculata, at least in response to yeasts (29). The single report of the cellular reactions of M. cornuarietis to an infection indicated a first phase of loose hemocyte aggregations around the larvae of A. cantonensis (first 24 h after exposure) that was followed 10-40 days later by massive granulocyte aggregations within large hemocoelic spaces and the formation of nodules by non-granular hemocytes in the connective tissue. A follow-up of these observations and a closer comparison with the cellular reactions in P. canaliculata is therefore needed.

There is an old issue of anatomical nomenclature that has caused much confusion, and that it is still causing troubles in recent literature. In a way, the question may be posed as simply as such: “How many kidneys do the ampullariids have?” And the right answer to the question may also be simple: “Just one, as all caenogastropods do”. Therefore, one can ask, “Why there are so many references in recent literature to an “anterior” and a “posterior kidney” in Ampullariidae? This needs to be clarified because the kidney is important as a potential hematopoietic organ in this family and the confusion is growing (see section 4.2.3.3 ).

The confusion originated in Bouvier’s and Perrier’s writings (58, 109), at the end of the 19th century. Perrier was involved in a discussion on the fate of the right renal primordium in adult “monotocardian”, while Bouvier, who was himself a pioneer of ampullariids anatomy, had wrongly interpreted that what is now recognized as an ectodermal ureter (25, 26) was indeed a derivative of the right renal primordium, and therefore it was a true kidney (109). Bouvier called it “anterior kidney”, as opposed to the kidney proper, which he called the “posterior kidney”. And because both “kidneys” were connected by a “nephrostome”, Perrier became convinced (p. 171) that the ampullariids represented an intermediate case between the two-kidney condition of Patella and the single kidney of other “monotocardians”, and that it was un passage intermédiaire (a “missing link”)! supporting his interpretation that a complete fusion of both primordia originated the single kidney of “monotocardian” gastropods.

Even though Perrier’s (58) interpretation did not gain general acceptance and was finally forgotten, his paper contributed to make better known the views of Bouvier, and the names of anterior and posterior kidneys did survive for more than a century and continue to cause difficulties today. Formally, the confusion was brought to an end by Andrews in 1981 (24), when she introduced the term “ureter” for the Ampullariidae, and referred to it as “the so-called anterior kidney of the earlier authors”. Those “earlier authors” included even herself (110), and we have also used such terminology once (63). Also, the malacological community assimilated slowly her formal corrigendum and, until recently, many continued to fall into the two-kidney confusion (e.g., 62, 65, 111). We should emphasize that these organs are morphologically (29), developmentally (25), and functionally (24) different, and that confounding them has had significant negative consequences (see section 4.2.3.3 ).

Yousif et al. (13) first suggested the occurrence of an “amoebocyte-producing organ” in an the ampullariid M. cornuarietis. The proposed site was a cellular band extending along the lung roof that became markedly thickened in snails infected with A. cantonensis larvae. This report seemed to disprove our evo-devo hypothesis that restricted the constitutive hematopoietic sites to the reno-pericardial complex, and therefore we carefully looked for an equivalent tissue in the lung of the only ampullariid we had at hand, P. canaliculata, but we did not find any such tissue in serially sectioned lungs, neither in the roof nor in any other part of the lung (112).

Another study that was made in P. canaliculata (62) had proposed the existence of a hematopoietic tissue within the pericardial cavity, more precisely in the external aspect of the “veins” entering the “heart”, as well as in the pericardial fluid. The same idea was put forward in a collaborative chapter (113) and the study was referred to in an influential recent review of hematopoiesis (21). We think it is erroneous, however. For untangling this matter, it should be first said that the “veins” shown in micrographs of both (62) and (113) were actually the auricle, that shows a thin wall and characteristic muscular bundles traversing its lumen, as it was described by Andrews (110; p. 83), while the referred “heart” was in fact the ventricle, whose connection with the single auricle is shown in Figure 2B of (62) and microscopically in Figure 1.1 of (113). It should also be said that the mitotic marker used, phospho-Ser10 histone H3, may not be specific, because it also marks non-dividing cells, such as neurons (108) in P. canaliculata. Notably, a large amount of free hemocytes in the pericardial fluid was shown (62; their Figure 6), all of which were marked with anti-phospho-Ser10 histone H3, both in control and challenged animals. Furthermore, even though no quantitation was attempted, the authors stated their number was not affected by blood withdrawal, i.e., the hematopoietic challenge they used (62; legend to their Figure 6). Also, some large nuclei were marked in the pericardial fluid, which were larger than those of hemocytes (shown in their Figures 6D–F but not mentioned in the legend). An important hint for the identification of the latter cells came indeed from (62; their Figure 5), and from (113; their Figure 1.1) where these cells appear orderly arranged in the epicardial epithelium of the auricle, indicating these marked epithelial cells were not hemocytes but the larger “epicardial cells” described by Andrews (114) in M. cornuarietis. Besides that, it should be said that there is no “pericardial side” of the pulmo-branchial and renal veins, because they are embedded in connective tissue before debouching into the auricle and do not have an intrapericardial path (110, 115). Epicardial cells are functionally involved in ultrafiltration and hence in the generation of the primary urine (22, 114), and they seem to constitutively express phospho-Ser10 histone H3, as some neurons of P. canaliculata may also express (108). Therefore, there is no evidence of a hematopoietic response to the experimental challenge used, and we should conclude that there is no empirical indication of a hematopoietic response in the pericardium of P. canaliculata, neither in the pericardial fluid nor in its epithelial lining.

Besides that, as it has been pointed out elsewhere (29), a potential hematopoietic site should have a direct connection with the hemocoel, where the newly generated hemocytes should pass, while the pericardial cavity connects not with the hemocoel but with the renal cavity and, if hemocytes were generated there, they would get finally expelled with urine. All previous propositions of hematopoietic sites in the pericardium of other species were in the hemocoelic sinuses that are contained within the pericardial wall (i.e., not in the pericardial cavity), because only those hemocoelic sinuses warrant a direct access to the systemic circulation.

Therefore, we had to look in a different direction and, according to our evo-devo hypothesis, we put to test the possible hematopoietic role of the only constitutive hemocyte aggregation we had noticed in P. canaliculata, i.e., the renal hemocyte islets. To test this possibility, the animals were injected with yeasts in the visceral mass (or with buffer) and cell proliferation in the islets was compared by (1) histological examination, (2) a bromodeoxyuridine incorporation assay, and (3) by measuring the changes in the renal area occupied by the hemocyte aggregates. All these different approaches were concurrent in indicating a hematopoietic response to the immune challenge: (1) nodules were formed in the hemocyte islets ( Figure 7 ), and (2) both bromodeoxyuridine incorporation and the area occupied by the hemocyte aggregates increased significantly (29). This was the first quantitative demonstration of a hematopoietic response to an immune challenge in an ampullariid and, to our knowledge, in the wide diversity of the Caenogastropoda.

However, we may still wonder whether the ampullariids have a hematopoietic organ in the pericardial blood sinuses, which certainly have direct passages to the hemocoel. This would be homologous to the APOs shown in several lymnaeoid and helicoid snails. However, this remains to be explored.

There is much to be learned about the molecules and mechanisms that guide the hemocyte proliferation and differentiation in gastropods (21), and especially in ampullariids. Therefore, we were much interested when we knew of the proposition of a cytokine-candidate that would increase hemocyte proliferation in P. canaliculata (65). It was molecularly related to the prokineticin-domain containing proteins, astakines-1 and -2, which were discovered in the crayfish Pacifastacus leniusculus (116). Astakines promote the release of new hemocytes from the hematopoietic tissue of this crayfish and induces cell division and the partial cell differentiation of hematopoietic cells in primary culture (117). Similar results were obtained in the shrimp Penaeus monodon (118). Also, a polypeptide containing a homologous sequence (named CgAstakine) was identified and found active in the Pacific oyster, Crassostrea gigas (119). In P. canaliculata, the cytokine-candidate was provisionally named Pc-PlP (for P. canaliculata-prokineticin like protein) (65), and changes in the expression of the Pc-PlP gene were compared with that of the housekeeping gene rpL5 in several tissues/organs, both in control (constitutive) conditions and after two different challenges, repeated blood withdrawals and LPS injection. The results obtained were not easy to interpret and the authors attempted to correlate them with propositions made in previous publications and with which we should dissent, as it has been previously discussed in this review. Those propositions were that the aortic ampulla was a hemocyte reservoir and a site for hemocyte maturation (reasons discussed in section 4.1.3.2 ) and that the pericardial cavity and its fluid were sites of hematopoiesis (reasons discussed in section 4.2.3.2 ).

Also, the paper considers the ureter (i.e., the “anterior kidney” of the older terminology, see section 4.2.3.1 for the source of this confusion) as the most responsive organ to LPS injection but, while attempting to relate their findings with hematopoiesis, the ureter is referred to as “the rostral component of an organ containing islets of hemocytes and proposed to be an immune barrier (Cueto et al., 2015)”, i.e., that the ureter and the kidney were mixed-up. Indeed, the immune barriers we proposed were the actual kidney and the lung (29), and the hemocyte islets were found only in the kidney (41), not in the ureter (see section 4.2.3.2 ). In fact, the ureter is not the rostral component of the kidney (24), it is a quite different organ (see section 4.2.3.1 ). In the same context, the occurrence of commensal microorganisms in the kidney was surprisingly attributed to a paper by us (i.e., 41), in which we said nothing regarding commensal microorganisms. The paper (65) also has some other inaccuracies regarding sampling and organ identification. One of them is that some intrapericardial “vessels” were sampled, which cannot be vessels but the auricle (as already explained in section 4.2.3.2 together with the consequences of this misidentification).

In our view, the reported pattern of control and challenged PcPlP expression does not suggest any clear role for this prokineticin-containing protein in the regulation of hematopoiesis in P. canaliculata. However, all the objections above should not obscure the interesting finding of a protein that bears structural similarities with the astakines that promote hematopoiesis in two crustaceans (116–118) and a bivalve (119) and that is encoded in the P. canaliculata genome (20). Though this is interesting in itself, we still have to learn about the actual function of this protein in P. canaliculata, which may or may not be related to hematopoiesis or to any other function of the immune system.

Hemocyanin is the most widespread respiratory pigment in molluscs, though it is lacking in the Patellogastropoda (120) and the Planorbidae, where it is replaced by hemoglobin (121). Sminia (52) was the first to observe polymerized hemocyanin in rhogocytes (also called “pore cells”, see section 4.2.4.2 ) of L. stagnalis and to propose that this pigment was synthesized and/or stored in these cells. More recently, Martin et al. (122) and Kokkinopoulou et al. (123) have also shown evidence for the localization of hemocyanin synthesis in rhogocytes, but no ultrastructural evidence was obtained for P. canaliculata (64, 112). The size and structure of hemocyanins vary among different taxa and are among the largest proteins in nature, with molecular masses ranging from 3.5 to above 13.5 MDa (124) and are thus visible under the transmission electron microscope (in P. canaliculata: 29, 125, 126). Hemocyanins have biomedical importance, and the first to be studied was that of the keyhole limpet (Megathura crenulata) which has been used for the specific treatment of bladder carcinoma (its efficacy probably due to a cross-reacting carbohydrate epitope) and is also widely used as a hapten carrier and generalized vaccine component (127). However, the hemocyanin of at least four other gastropod species have been studied: Concholepas concholepas, Haliotis tuberculata, Rapana thomasiana and Fissurella latimarginata (128).

The hemocyanin of P. canaliculata has received particular attention in recent proteomic and genomic studies. It is a multimeric protein encoded by four similar but sequentially distinct genes (126). Mining in the known genomes and transcriptomes of other three ampullariid species (Marisa cornuarietis, Pomacea maculata, and Lanistes nyassanus) (20, 129) indicates that the four different subunit isoforms seem typical of the family. Those subunits are polymerized forming di-, tri- and tetradecamers (being di- and tri-decamers more abundant than tetradecamers) which could be seen under TEM.

These cells are thought to be exclusive of molluscs and may be the only distinct character exhibited by all extant species of the phylum (130, 131). A distinctive morphological character of these cell is a “slit apparatus” occurring in the plasma membrane as the outer limit of “extracellular lacunae” (=“subsurface cisternae”; 132) and that may serve as a molecular sieve (see below) (121, 123). They are usually embedded in the connective tissue, as resident hemocytes are, but free-floating rhogocytes have also been mentioned in the hemocoel (e.g., 133, 134). Haszprunar (130) proposed the homology of rhogocytes with podocytes, nephrocytes, and cyrtocytes, which links these cells with the excretory system in molluscs and some non-molluscan phyla, and thereby with the gastropod reno-pericardial system and with our evo-devo hypothesis.

Rhogocytes in P. canaliculata ( Figure 10 ) show large, pleomorphic, and frequently notched nuclei, containing heavy heterochromatic clumps and conspicuous nucleoli. The tubuli and cisternae of both the rough (RER) and smooth (SER) endoplasmic reticulum are frequently dilated and filled of a uniform material with an electron density similar to that of the surrounding cytoplasm, but there are no distinctive signs of phagocytic activity. Another feature of rhogocytes are the cytoplasmic granules of varying electron density, that may become very large, likely by the merging of smaller ones.Taken together, these nuclear and cytoplasmic features would suggest polyploidy and a high level of transcription and post-transcriptional processing.

Regarding hemocyanin synthesis, it is intriguing that P. canaliculata rhogocytes do not show the typical hemocyanin polymers seen in some gastropod species (e.g., 52, 122, 123) while loads of hemocyanin polymers may occur in the blood spaces around the renal hemocyte islets of this species ( Figure 11 ) (29). Could it be possible that hemocyanin subunits were synthesized in rhogocytes, and transported to the kidney for polymerization?

We first found rhogocytes in the lung of P. canaliculata at the interface between the fibromuscular layer that underlies the respiratory lamina and the storage tissue that forms most of the pulmonary wall in this species (112) ( Figure 12 ). These rhogocytes were in close association with the storage cells that accumulate urates (63, 135) and possibly glycogen (136) in this species. The distribution of these storing cells in ampullariid tissues is well known (63), but not that of rhogocytes, that have only been searched for in the lung and the aortic ampulla (64, 112).

Cells with storing capacity for quite different substances have received different names in gastropods (e.g., vesicular cells, globular cells, chondroid cells, plasmatic cells, Leydig cells, etc.) and were frequently associated with glycogen storage (22, 28). Because the phylum Mollusca lacks a specialized tissue as the adipose tissue of vertebrates (137), glycogen storing cells and tissues may be serving an analogous role in energy metabolism. However, other substances are stored in these tissues in P. canaliculata and other gastropod taxa [such as calcium carbonate in the congeneric P. maculata (138), or copper in a likely homologous crystalloid-containing tissue in L. littorea (139)].

A functional interpretation of the close association of rhogocytes with cells of the storage tissue of P. canaliculata (63, 64, 112) is wanting. Uric acid is accumulated intracellularly in this tissue in the form of spheroidal crystalloids (135). This purine, which has been generally considered an excretory compound, circulates in plasma at a concentration of ~5 mg/L (140) and is totally resorbed from the primary urine, so that it is not excreted (135). The storage cells are integrated in a stress-related tissue system that is anatomically discontinuous because it is present in some specific organs only (mainly the lung, the digestive gland, the wall of the style sac, the mesenterium of the coiled gut, the testis and the heart ampulla) (63). Physiological studies have shown uric acid released from these sites participates in the adaptive responses to oxidative stress when the animal enters or leaves the hypometabolic states of estivation and hibernation (141, 142), and proteomic studies have confirmed they also partake in other adaptations to stress (hypoxia or pesticide exposure) (64, 143).