Gastrulation occurs in multiple phases at two distinct sites in Latrodectus and Cheiracanthium spiders

In all three species studied, we show that the cells of the cumulus internalize first
and in concert. There is then a pause in the internalization process, after which
non-cumulus mesendoderm cells begin to internalize. There is variation between the
two Latrodectus species versus C. mildei in the location of this second round of internalization. In Latrodectus spp., mesendoderm internalizes along a linear blastoporal region that includes the
former site of cumulus ingression. Much later, after cumulus migration is complete,
cells also internalize in situ at the rim of the germ disc. In contrast, secondary
internalization in C. mildei occurs in situ at a ring or region approximately two–five cell lengths away from
the original blastopore. Thus, in both genera, cells internalize first at a central
blastopore (to form the cumulus) and subsequently at other positions on the blastoderm
(to form general mesendoderm), see Fig. 2. The early timing and unitary site of internalization for the prospective cumulus
strongly suggest that the cells of the cumulus are already committed to their fate
as the dorsal organizer when they internalize. The canonical model’s assumption that
the cumulus arises from a subpopulation of generalized primitive plate cells within
the deep layer is not consistent with what we observe in the species studied here,
as virtually all the early ingressing cells form the cumulus.

Fig. 2. Proposed revisions to the canonical model of spider gastrulation. Diagrammatic midsagittal
sections. Ectoderm is blue. Top row: Canonical model. A mixed population of cumulus cells and deep layer mesendoderm cells (orange) enters through the central blastopore. The cumulus (red) then differentiates and migrates towards the prospective dorsal side as the primitive
plate (yellow) enlarges. Gastrulation ends when a deep layer has formed under the entire germ disc
and signaling from the cumulus has broken radial symmetry. Bottom row: Our working model. The cumulus mesenchyme cells (red) internalize first, as a group, through the central blastopore. As the cumulus migrates,
additional mesendoderm (yellow) cells internalize at two distinct locations: through the central blastopore and
at the rim of the germ disc or, in species lacking a germ disc, a region several cells
away from the central blastopore

Latrodectus mactans and L. geometricus

Timing of cumulus internalization and its fate

In Latrodectus spp., the presumptive cumulus cells are the first cells to internalize. We show this
in two ways: by virtual tracing of cells in timelapse movies and by direct labeling
of presumptive cumulus cells with CM-DiI. Tracings of two representative L. mactans embryos are shown in Fig. 3, in which the first cells to enter the deep layer at a central point blastopore are
colored red and pink. Immediately after most of these cells internalize, even while
a few are still visible at the blastopore, the cumulus begins to migrate. It can be
seen as a prominent bulge beneath the germ disc originating below the blastopore (red
outline, Fig. 3). During this early phase of cumulus migration, there is a distinct pause of about
1 h while no further cells are internalized. Then internalization resumes at the same
blastoporal region as non-cumulus cells begin to ingress. At this stage, the blastopore
elongates to form a groove (most easily seen in Lm embryo 1—note slight depression
at times 550–1175 min, Fig. 3; Additional file 4: Movie 1); and non-cumulus mesendoderm internalizes (yellow cells). Internalization
appears to be somewhat asymmetric; more cells internalize on the side of the groove
away from the path of cumulus migration (this can be seen qualitatively in timelapse
movies). Much later, after the cumulus has migrated, additional mesendoderm ingresses
at the rim of the germ disc (discussed in more detail below). We were unable to produce
tracings in L. geometricus to confirm the same result in that species, but our histological data (see below)
are consistent with the view that the cumulus cells arise by a conserved process in
both Latrodectus species.

Fig. 3. Gastrulation proceeds in two phases in L. mactans embryos. Tracings of external views of living embryos show that prospective cumulus
cells (red and pink) internalize first at the blastopore (bp). Internalized cumulus cells are circled in red and migrate along red arrow. Other presumptive lower layer cells (shades of yellow) internalize later to form generalized mesendoderm. Note yellow cells at periphery of germ disc; these were traced and will internalize later. Blue cells are those known to persist in the superficial layer of the germ disc as long
as their daughter cells can be visualized. Times in min. Scale bars 200 µm. Timelapse movies of Lm embryos 1 and 2 are shown in Additional file 4: Movie 1 and Additional file 5: Movie 2)

To confirm that the first cells to internalize are prospective cumulus cells, we microinjected
the cell tracer CM-DiI into the deep layer cells immediately below the blastopore.
We then traced the path of migration of the first ingressors and determined their
fate later in development. When injected early in gastrulation, before cumulus migration
has started, the label was restricted to the cumulus during its migration. No labeled
cells appeared in any other region of the germ disc 5–16 h post-injection, showing
that the first cells to internalize contribute exclusively to the cumulus (N = 24, 12 followed to germ band stage; Fig. 4a, b). Subsequent position of the label is consistent with what is known about the
fate of the cumulus once its cells disperse (11], this paper); labeled cells were found in the posterior germ band 9–14 days after
injection (N = 12; Fig. 4c, d). In contrast, injection below the blastopore after the cumulus was in migration
labeled only non-cumulus cells (i.e., no later-ingressing cells joined with the cumulus).
After 14 days, labeled cells in these embryos were found in the deep layer of the
lateral germ bands (except the appendages) and extraembryonic areas (N = 6; Fig. 4e, f). These locations are consistent with a generalized ‘mesendodermal’ fate and
differ from the solely posterior position of the dispersed cumulus cells. Thus two
approaches to the problem of the timing and spatial origin of the cumulus cells yield
the same answer: all and only prospective cumulus cells internalize first and in concert,
before generalized mesendoderm enters the deep layer. This suggests that prospective
cumulus cells become committed to their fate without the requirement for cell–cell
interaction within the deep layer, as cumulus migration begins before other cells
have internalized.

Fig. 4. Direct labeling of the deep layer confirms that cumulus cells internalize first in
L. mactans. a External view just after CM-DiI injection through blastopore into early deep layer,
arrowhead indicates label (red) in cumulus. b Confocal section of similar embryo fixed 1 day after injection; red dye is limited
to cumulus mesenchyme (arrowhead). A few cells appear to have been pushed deeper into the embryo during injection.
Tubulin is green, nuclei are blue, arrowheads mark labeled cells in b–f. c, d Injection into early deep layer. c Confocal stack of embryo fixed 1 week after injection; label is in posterior germ
band consistent with normal migration path of cumulus. d Confocal section of a similar embryo fixed 1 week after injection; labeled cells
are in deep layer at posterior. Asterisk marks fluorescent debris attached to exterior of embryo. e, f. Injection into later deep layer below blastopore, after cumulus has migrated away
from blastopore (~1 day later than early injection). e Confocal stack of embryo fixed 2 weeks after injection; labeled cells are in right
germ band and extraembryonic area. f Confocal section of similar embryo fixed 2 weeks after injection; labeled cells are
in deep layer of right germ band. Position of label is consistent with non-cumulus
fate. Scale bars 200 µm

Histology confirms that the cumulus cells internalize before other cell types

Histology can reveal details of cell shape diagnostic of cell type and from which
cell behaviors can be inferred 30]. In Latrodectus spp., the cumulus cells are typically larger and more spherical than adjacent cells.
Sections of early L. geometricus embryos show that the presumptive cumulus cells ingress first (Fig. 5a): no other deep cells are found in serial sections. When internalization of other
primitive plate cells begins, the cumulus is seen as a mass of spherical cells distinct
from the other cells of the deep layer, which are spindle-shaped at this early stage
(Fig. 5b, c). In both species, internalizing cumulus cells and other primitive plate cells
show a bottle cell shape, where a cell is apically constricted with a wide basal region
(e.g., outlined cells in Fig. 5a, e). This is indicative of invagination or ingression in a wide variety of taxa
31]. After internalization, most non-cumulus deep cells appear mesenchymal whereas columnar
and cuboidal epithelial cells make up the germ disc surface (Fig. 5d–g). If non-cumulus cells appear round, they are distinctly smaller than those composing
the cumulus (e.g., Fig. 5d). Rempel 28] reported some vertical mitoses contributing to the formation of a deep layer, but
our sections confirm Rempel’s view that most mitoses in the blastoderm are horizontal
rather than vertical (Fig. 5f). Such horizontal orientation of mitosis could contribute primarily to the epibolic
expansion of the superficial layer rather than adding cells to the deep layer.

Fig. 5. Cell morphology at two loci of internalization in Latrodectus species.a Beginning of gastrulation: presumptive cumulus cells internalize through central
blastopore (arrowhead). Bottle cell outlined. L. geometricus.b Slightly later stage: most deep cells are cumulus mesenchyme (circled). Cumulus has begun to migrate. A few primitive plate cells have ingressed (green arrowheads). L. geometricus.c Closeup of boxed area in b. d Mid-gastrula: cumulus (circled) has migrated to edge of germ disc. Blastopore is not in plane of section. L. mactans.e. Late gastrula: primitive plate comprises two to three layers of deep cells. Internalization
continues at central blastopore (arrowhead). Two bottle cells outlined. L. mactans.f Mitotic spindles are oriented such that daughter cells typically remain in the same
cell layer (black arrowheads), although some cell divisions are vertical (white arrowhead). L. mactans.g Section through rim of germ disc late in gastrulation showing a chained array of
large, round cells (yellow arrowheads). Green arrowheads indicate some mesenchymal cells in the deep layer. L. mactans. Scale bars 50 µm

Evidence for internalization at the germ disc rim

Our tracings of cells in timelapse videos during late gastrulation show definitively
that cells internalize at the rim of the germ disc in both L. mactans and L. geometricus (Fig. 6; Additional file 6: Movie 3, Additional file 7: Movie 4). Internalization at the rim begins about 14 h after gastrulation at the
central blastopore has ceased. At this stage, the cumulus has completed migration
and the dorsal field has begun to form. In L. mactans internalization at the rim appears to contribute only a few cells to the deep layer.

Fig. 6. Cells internalize at the rim of the germ disc in Latrodectus species. External views of late gastrulae show internalization of prospective mesendoderm
(yellow) and, in blue, examples of cells known to persist in the upper layer of the germ disc as long as
their daughter cells can be visualized. Arrowheads point to cells that will internalize. df, dorsal field, region to which cumulus has
migrated. Lm, L. mactans. Lg, L. geometricus. Times in min. Scale bars 200 µm. Timelapse movies of Lm embryo 1 and Lg embryo 1 are shown in Additional file
6: Movie 3 and Additional file 7: Movie 4

Round cells of intermediate size are visible in the deep layer of histological sections
late in gastrulation (Fig. 5g, green arrowheads). In sections of P. tepidariorum, Montgomery 22] notes apparent cell internalization around the rim of the germ disc, and in L. mactans Rempel 28] notes a chain of deep-layer cells linked in series and connected to the superficial
layer at the germ disc rim. Our sections show virtually the same static image as presented
by Rempel (Fig. 5c cf. his Fig. 26), and is compatible with internalization at the rim of the germ disc,
perhaps by involution. This is consistent with the expression of At–twist and fkh at the rim of P. tepidariorum18], 32]. However, the problem shared by histology and static gene expression patterns is
that such studies cannot reveal the origin of the cells; for example, in the case
of At–twist and At–fkh, cells expressing these markers are found at the rim but may have originated at the
central blastopore and simply migrated. In P. tepidariorum, this issue was ultimately addressed by following a labeled clone in timelapse video,
to obtain direct evidence of internalization at the rim 29]. In Latrodectus, we followed the movements of individual cells at the rim in video recordings.

Although the chained array of cells seen in sections was suggestive of involution,
our videos show no mass movements or inrolling—internalizing cells left the surface
only by ingression. This pattern of individual ingression is also seen in embryos
of L. geometricus, with more cells internalizing at the rim than in L. mactans (Fig. 6). As with L. mactans, L. geometricus cells internalize at multiple positions around the rim of the germ disc. We did not
see internalization of cells in the central germ disc at this stage in timelapse movies
of either species, and neither did we see bottle cells in the central region at these
later stages in our extensive histological series.

Cheiracanthium mildei

Central and ring blastospores are distinct sites of internalization

Similar to many spiders (Ischnothele karschi12], C. salei21]), C. mildei embryos lack a sharply defined boundary at the edge of the germ disc. Nevertheless,
at germ disc stage the majority of cells occupy one hemisphere of the embryo, and
the blastopore is located near its pole. Frame-by-frame cell tracings of timelapse
movies of C. mildei show that early gastrulation is similar to that of other species: the cumulus cells
internalize as a group through the central blastopore (Fig. 7; Additional file 8: Movie 5), as in L. mactans (this paper), Z. x–notata26], and P. tepidariorum (our unpublished observations). However, internalization of the cumulus is followed
approximately 3 h later by a novel gastrulation process: cells ingress in a ring approximately
2–5 cell lengths from the original blastopore. The new region of ingression is spatially
distinct from the point through which cumulus cells internalized, although some cells
are adjacent (e.g. Embryo 2 in Fig. 7). Cells internalizing via the ring probably contribute mesendoderm to the primitive
plate. It seems likely that additional mesendoderm also internalizes at the original
blastopore at this stage, but the central cells are so tightly packed that they cannot
be resolved individually. Because the number of cells seen to internalize at the annular
blastopore does not account for the entire lower layer in C. mildei, it is likely that both areas continue to internalize mesendoderm during the second
round of gastrulation.

Fig. 7. Gastrulation at two distinct loci in Cheiracanthium mildei. A1–E1, A2–E2 External view of two representative embryos show internalization of different cell
populations at two distinct loci. First, cumulus cells (red and pink) internalize through a central blastopore; then generalized mesendoderm (yellow) internalizes in situ at a separate ring outside the blastopore. Some mesendoderm
likely internalizes at the original blastopore as well, see text. Times in min. Scale bars 200 µm. Timelapse movie of Cm embryo 1 is shown in Additional file 8: Movie 5. A3–E3 Representation of gastrulation in C. mildei from our timelapse and histological data (not shown). Diagrammatic midsagittal sections.
Ectoderm is blue, cumulus is red, mesendoderm is yellow

The cumulus does not begin its migration until mesendoderm ingression through the
central and annular blastospores is underway and a distinct lower layer has formed.
This could mean that the cumulus in C. mildei forms by cell–cell interaction within the primitive plate, as postulated by the canonical
model. To examine this possibility, we labeled early-ingressing cells with CM-DiI:
embryos were injected directly into the nascent deep layer below the early blastopore.
During cumulus migration, in both external view and in histological section, label
was found in only the cumulus mesenchyme cells (N = 15, Fig. 8a, b). If the cumulus cells were to differentiate from a mixed population of cells
in the early deep layer, CM-DiI labeling of these cells should yield label in both
cumulus and generalized mesendoderm. Our results show that only cumulus cells are
labeled by early injection, so specification of cumulus cells likely occurs before
general mesendoderm internalization.

Fig. 8. The cumulus internalizes first and its cells have an opisthosomal fate in C. mildei. a Early gastrula soon after injection of CM-DiI (red) into the early deep layer below the blastopore. b Representative vertical section (10 µm, paraffin) through cumulus 6 h after CM-DiI
injection. Nuclei stained with DAPI. c First-instar juvenile 10 days after injection, parasagittal section near midline.
White arrowheads indicate label. d Heat map showing positions of label in serial sections spanning ~160 µm of the sagittal
midline. Counts were summed for four juveniles with a maximum value of 32 points of
label counted per grid position. Darker shades of color represent more points per grid position, summed for the four individuals. Additional
labeled cells were found in more lateral sections (not shown). Scale barsa, c 200 µm; b 50 µm

Fate of the cumulus

Once the cumulus has migrated to the edge of the germ disc or the equivalent position
in C. mildei, its cells disperse as the dorsal field forms. Until the present study, only Holm
11] was successful in following the fate of cells beyond dorsal field formation by labeling
cumulus cells in A. labyrinthica with carmine powder. Holm was able to find carmine particles in the opisthosoma,
but presented only one case. We raised four injected C. mildei embryos to hatching, and all labeled cells were found in the opisthosoma, confirming
Holm’s basic finding (Fig. 8c). Most labeled cells were in the dorsolateral region of the abdomen. A limitation
of our study is that development in oil (necessary for the injections) interferes
with chitinization and molting, and development arrests after hatching. Organogenesis
is not complete at this stage, particularly in the opisthosoma; so there is no absolute
certainty regarding the ultimate adult fate of the labeled cells. However, histological
sections of uninjected controls, including second instar and adult, suggest that the
dorso-lateral abdomen is composed largely of gut and gut diverticula (Additional file
9: Figure S4). Therefore, the labeled cells probably form some portion of the digestive
mass. This is consistent with molecular data that cumulus cells express endodermal,
but not mesodermal markers 17], 18]. It is also possible that the labeled cells are fated to form visceral mesoderm.