Go with the flow – bulk transport by molecular motors

Cytoplasmic streaming, an active bulk intracellular flow of cytoplasmic contents, was first discovered in algae cells more than 200 years ago (Corti, 1774). This self-organized movement of cytoplasmic contents can reach speeds of up to 100 µm/s in some algal cells (Shimmen and Yokota, 2004). For comparison, synaptic vesicles move in the axon of a mammalian neuron at a velocity of less than 2 µm/s (Nakata et al., 1998). Over the past two decades, an increasing number of examples of bulk cytoplasmic movement have been reported in large animal cells. Despite these cytoplasmic movements occurring in numerous species at various developmental stages, with different patterns and velocities and for distinct purposes, they share a common feature of being mostly driven by cytoskeletal filaments and their associated molecular motors – protein machines that move on the cytoskeletal filaments in a step-by-step fashion powered by ATP hydrolysis. Here, we first provide an overview of how molecular motors organize cytoskeletal components and power large-scale intracellular cytoplasmic movements in different systems, then discuss the biological functions of these movements, and finally describe intercellular cytoplasmic transport into oocytes.

Since its initial discovery in internodal cells of the algae Nitella and Chara (Corti, 1774), cytoplasmic streaming has been documented widely in higher plants, including Elodea and Arabidopsis (Allen and Allen, 1978; Goldstein and van de Meent, 2015) (Table 1). During cytoplasmic streaming, chloroplasts – among other cargoes – rotate in a synchronized manner, which is prominently observed within the large leaf cells of Elodea (Fig. 1A).

Cytoplasmic streaming in plants is dependent on the action of myosin-XI, a plant-specific myosin class, which walks on actin filaments (Goldstein et al., 2008; Haraguchi et al., 2022; Morimatsu et al., 2000; Shimmen, 2007; Tominaga and Ito, 2015; Tominaga et al., 2013; Ueda et al., 2010; Williamson, 1972, 1975; Woodhouse and Goldstein, 2013). Myosin-XI moves large, interconnected organelles such as the endoplasmic reticulum (ER) along actin filaments, generating a bulk flow of cytoplasm (Kachar and Reese, 1988; Stefano et al., 2014; Tominaga and Ito, 2015; Ueda et al., 2010) (Fig. 1A, Table 1). Since the ER is the largest membrane network in eukaryotic cells, with close contacts or connections with other organelles (Guo et al., 2018), movement of the ER presumably exerts enough mechanical forces on the surrounding cytoplasm to generate cytoplasmic movement via viscous drag.

Myosin-XI proteins contain an N-terminal motor domain, a neck region built of six calmodulin-binding IQ motifs, a dimerizing coiled-coil rod region and a C-terminal globular cargo-binding tail (Fig. 1A); this resembles the structure of the evolutionarily related animal myosin-V (Ryan and Nebenfuhr, 2018; Tominaga and Nakano, 2012). However, myosin-XI is a remarkably fast motor, with a velocity of 7 µm/s or higher (Haraguchi et al., 2022; Ito et al., 2007; Tominaga et al., 2003; Tominaga and Nakano, 2012), which is faster than all myosins in animal cells, including myosin-V (which has a velocity of less than 1 µm/s; O'Connell et al., 2007; Tominaga and Nakano, 2012). In the giant internodal cells of the algae Chara, which can reach 20 cm in length, the speed of cytoplasmic streaming can reach 70 µm/s (Yamamoto et al., 2006) (Table 1). Matching this high velocity of cytoplasmic streaming, two myosin-XI motors from Chara braunii (CbXI-1 and CbXI-2, encoded by CBR_g50407 and CBR_g48390, respectively) display striking velocities of ∼60–70 µm/s (Haraguchi et al., 2022), which is ten times faster than the fastest animal myosin, skeletal muscle myosin-II (∼6.9 µm/s) (Uyeda et al., 1990).

What is the underlying molecular basis of such striking differences in velocities? Velocity of a motor can be simply calculated by multiplying its step size by its stepping rate. As both myosin-XI and myosin-V have similar step sizes on actin filaments (35 nm and 36 nm, respectively; Cappello et al., 2007; Tominaga et al., 2003), the higher velocity of myosin-XI could be a result of a higher stepping rate. The stepping rate is largely influenced by the ATP turnover rate and by the actin–myosin binding and dissociation rate. For instance, the ATP turnover rate for myosin-Va (MYO5A) is ∼12 inorganic phosphates released per second per myosin head group (Pi/s/head), whereas the Chara myosin-XI motors can reach up to 500 Pi/s/head (Ito et al., 2003; Tominaga and Nakano, 2012). This high ATPase activity of Chara myosin-XI is at least partially attributed to its dramatic acceleration of ADP release by actin and extremely fast ATP binding rate (Ito et al., 2007). Moreover, a recent study on the ultrafast Chara myosin-XI motors CbXI-1 and CbXI-2 has identified a unique actin-binding region that is not found in other slower myosin-XI motors and myosins of different classes (Haraguchi et al., 2022). Remarkably, introduction of the CbXI-1 actin-binding region significantly increases the motor velocity of a slower myosin-XI motor (Haraguchi et al., 2022). The unique actin-binding region might enhance the actin-activated ATPase activity, as well as the actin–myosin binding and dissociation rate, therefore increasing the stepping rate and optimizing the myosin-XI motor for extremely high velocities (Haraguchi et al., 2022).

Typically, animal cells use myosin-V for short-distance transport or local dynamic anchorage on a randomly organized meshwork of actin filaments (Lu et al., 2020; Rodionov et al., 1998; Rogers and Gelfand, 1998). In contrast, plants mainly rely on myosin and actin tracks for long-range cargo delivery (Sparkes, 2011). Thus, it is not too surprising that plant myosin-XI motors are more efficient for long-range transport (for example, myosin-XI from tobacco has a velocity of 7 µm/s and a run length of ∼1.2 µm; Tominaga et al., 2003) than animal myosin-V motors [for example, Drosophila myosin-V (also known as Didum) has a velocity of ∼0.5 µm/s, human myosin-Vb (MYO5B) has a velocity of 0.2 µm/s, and the average run length of myosin-V is ∼1.3 µm; Hathcock et al., 2020; Toth et al., 2005; Watanabe et al., 2006].

As cytoplasmic streaming was first observed in algae and later found to occur widely in higher plant cells, it was initially considered an exclusive mechanism for active circulation of cytoplasm in immobile plants. However, cytoplasmic streaming has also been observed outside the plant kingdom, including in the plasmodium of slime molds (Stewart and Stewart, 1959). The plasmodium is a tubular network generated from a single large multinucleate syncytium after synchronized division of nuclei without cytokinesis. In the species Physarum polycephalum, the plasmodium can grow up to a meter in size (Kuroda et al., 2015) and displays a characteristic periodic cytoplasmic streaming with a speed of up to 1 mm/s in large tubes (Stewart and Stewart, 1959) and ∼50 µm/s in small tubes (Alim et al., 2013) (Table 1). Cytoplasmic streaming in this system is self-organized and is driven by the cross-sectional myosin-II-dependent contraction of the actomyosin cortex at the outer layer of the tube (Fleig et al., 2022; Kawamichi et al., 2007) (Fig. 2A). Streaming in the plasmodium reverses flow direction every ∼100 s (Stewart and Stewart, 1959), displaying a classic back-and-forth shuttle streaming pattern (Goldstein and van de Meent, 2015). The change of direction is caused by rhythmic cycling between contraction and relaxation, which is associated with periodic Ca²⁺ waves that inhibit myosin-II activity and thus lead to the relaxation of the plasmodium tubes (Farkas et al., 2003; Smith and Saldana, 1992; Yoshiyama et al., 2009) (Fig. 2A).

In addition to plants and slime molds, several examples of cytoplasmic streaming have been described in animal cells over the past couple of decades, mostly in the largest animal cells, such as oocytes and one-cell zygotes (Table 1), suggesting that bulk cytoplasmic movement is a universal phenomenon occurring in large eukaryotic cells. Unlike plant and slime mold cells, which use actin-based myosin motors for streaming, animal cells employ either microtubule-based or actin-based motors to drive cytoplasmic movement (Table 1).

Directional rotation of the entire cytoplasm has been characterized in both Caenorhabditis elegans meiotic embryos (meiotic cytoplasmic streaming; Kimura et al., 2017; McNally et al., 2010; Yang et al., 2003) and in Drosophila late-stage oocytes (ooplasmic streaming; Gutzeit, 1986; Gutzeit and Koppa, 1982; Palacios and St Johnston, 2002; Serbus et al., 2005). For both C. elegans meiotic embryos and Drosophila oocytes, cytoplasmic streaming is completely dependent on microtubules (Gutzeit, 1986; Yang et al., 2003) and the heavy chain of the major microtubule plus-end motor, kinesin-1 (Ganguly et al., 2012; Kimura et al., 2017; Lu et al., 2016; McNally et al., 2010; Palacios and St Johnston, 2002; Quinlan, 2016; Serbus et al., 2005). Kinesin-1 heavy chain (KHC; encoded by Khc in Drosophila and by unc-116 in C. elegans) contains an N-terminal ATPase hydrolysis-powered motor domain, a central coiled-coil region for dimerization, and a C-terminal tail that is responsible for cargo binding and autoinhibition (Fig. 1B), the activity of which is regulated by kinesin-1 light chain (KLC; encoded by Klc in Drosophila and by klc-1 and klc-2 in C. elegans) and/or other cargo-binding adaptors (Verhey and Hammond, 2009).

During cytoplasmic streaming in C. elegans meiotic embryos, yolk granules are transported by kinesin-1 along microtubules from the minus ends at the cortex towards the inward-pointing plus ends, resulting in accumulation of yolk granules in the center of the embryos – a process also known as yolk granule packing (McNally et al., 2010). Hence, it was initially proposed that the movement of yolk granules drives meiotic cytoplasmic streaming (McNally et al., 2010). However, a more recent study has shown that yolk granule packing is not sufficient to drive streaming; instead, ER integrity is essential for streaming, as fragmentation of the ER network using specific genetic mutations blocks cytoplasmic streaming without disrupting yolk granule transport (Kimura et al., 2017). Thus, according to the current model, streaming is largely driven by kinesin-1-mediated ER transport along microtubules, creating a local flow. Because of its network-like structure, the ER transmits the mechanical force to the neighboring cytoplasm and aligns microtubules towards the flow direction for collective movement of the cytoplasm (Kimura et al., 2017) (Fig. 1B).

A similar model involving transport of small individual cargoes along microtubules by kinesin-1 has been proposed for Drosophila ooplasmic streaming (Monteith et al., 2016; Quinlan, 2016; Serbus et al., 2005). However, unlike meiotic cytoplasmic streaming in C. elegans, which is completely dependent on the kinesin-1 cargo adaptor KCA-1 (which interacts with KHC via KLC; McNally et al., 2010; Yang et al., 2005), ooplasmic streaming in Drosophila is not abolished upon knockout of KLC or the kinesin-1 adaptor Pat1, which are typically involved in cargo binding (Loiseau et al., 2010; Palacios and St Johnston, 2002). A likely explanation is that from a hydrodynamic point of view, movement of spherical cargoes produces much less viscous drag on the surrounding cytoplasm than movement of the network-like structure of the ER in C. elegans, as the viscous drag is proportional to the surface area of the moving object. Therefore, although it is possible that the transport of small spherical vesicles synergistically contributes to the overall streaming movement (Fig. 1C), other mechanisms must exist for Drosophila ooplasmic streaming.

Intriguingly, in addition to its traditional cargoes, KHC can move microtubules as a special cargo via its ATP-independent microtubule-binding site at the KHC tail (Fig. 1C; Movie 1) (Lu and Gelfand, 2017). This additional C-terminal microtubule-binding site was identified soon after the discovery of kinesin-1 (Hackney and Stock, 2000; Navone et al., 1992; Seeger and Rice, 2010) and is highly conserved across species from worm to human (Lu and Gelfand, 2017; Lu et al., 2016), but was mostly ignored until kinesin-driven microtubule sliding was described and characterized (Barlan et al., 2013; Jolly et al., 2010; Lu et al., 2013; Lu and Gelfand, 2017). Mutations of the KHC tail decrease microtubule-binding affinity (to ∼30% of wild-type KHC tail binding affinity) and result in a significant reduction of cytoplasmic streaming (to ∼42% of wild-type streaming velocity), indicating that KHC-driven microtubule–microtubule sliding provides the main force that drives rotation of the cytoplasm (Lu et al., 2016; Winding et al., 2016). Previously, microtubule sliding by KHC has been shown to drive cell shape change of Drosophila S2 cells and neurons via microtubules directly pushing against the cell membrane (del Castillo et al., 2015a; Jolly et al., 2010; Lu et al., 2013, 2015; Lu and Gelfand, 2017; Winding et al., 2016). In the ooplasmic streaming model, free cytoplasmic microtubules are transported by KHC along cortically anchored microtubules (Fig. 1C; Movie 1). Because microtubules are long polymers that have a stiffness comparable to Plexiglass (Gittes et al., 1993), their movement can create high hydrodynamic forces via viscous drag to propel movement of the cytoplasm (Fig. 1C). A further notion supporting the role of microtubule transport in streaming is that oocytes display a classic pattern of bundled microtubules along the oocyte cortex (Lu et al., 2016; Serbus et al., 2005; Theurkauf et al., 1992). This specific microtubule pattern can be self-organized by kinesin-1 walking on the microtubules, which in turn spontaneously drives large-scale cytoplasmic movement (Monteith et al., 2016), as has been shown experimentally (Drechsler et al., 2020) and by computational simulations (Stein et al., 2021).

In addition to microtubule-based movement by kinesin-1, actin-walking myosin motors can drive several types of cytoplasmic streaming in large animal cells through actomyosin contraction (Table 1). Contraction of non-muscle myosin-II (which has a heavy chain encoded by nmy-2 in C. elegans and by zip in Drosophila) at the actin cortex drives both cortical flow and cytoplasmic flow in the C. elegans one-cell zygote (Golden, 2000; Gubieda et al., 2020; Munro et al., 2004; Shelton et al., 1999) and Drosophila syncytial embryos (Deneke et al., 2019; von Dassow and Schubiger, 1994) (Fig. 2B,C). In C. elegans embryos, a flow of actomyosin cortex moves asymmetrically towards the anterior side of the zygote, while a cytoplasmic flow in the center of the zygote moves towards the posterior end (Fig. 2B), resembling the pattern of reverse fountain streaming found in the pollen tubes of plants (Goldstein and van de Meent, 2015). The mechanism driving the anterior cortical flow is the release of Aurora A (also known as AIR-1 in C. elegans) from the sperm-derived centrosomes, which inhibits non-muscle myosin-II-dependent actomyosin contractility at the posterior cortex (Cowan and Hyman, 2004; Kapoor and Kotak, 2020; Zhao et al., 2019), thus resulting in asymmetric contractility of the cortex (Mayer et al., 2010; Munro et al., 2004) (Fig. 2B). This anterior-directed cortical flow generates and transmits enough force through the viscous cytoplasm to create a posterior-directed cytoplasmic flow in the center (Niwayama et al., 2011).

Drosophila syncytial embryos display cortical flow towards the center and cytoplasmic flow towards both poles (von Dassow and Schubiger, 1994) (Fig. 2C). Similar to C. elegans embryos, both types of flow are driven by non-muscle myosin-II contractility at the cortex. During cell cycles 4–6, mitotic exit allows the active phosphatase PP1 to recruit non-muscle myosin-II near the nuclei, creating a non-muscle myosin-II gradient with the highest concentration at the central cortex of the embryo (Deneke et al., 2019) (Fig. 2C). This non-muscle myosin-II gradient leads to polarized cortical contraction, thus creating the cortical flow. As the cytoplasm is an incompressible fluid, the inward cortical flow, in turn, drives the outward cytoplasmic flow (Deneke et al., 2019) (Fig. 2C).

In addition to C. elegans and Drosophila, the starfish oocyte is known for its cytoplasmic flow and large surface deformations during meiosis (Hamaguchi and Hiramoto, 1978), which are caused by the non-muscle myosin-II-dependent cortical actomyosin contraction waves (Klughammer et al., 2018) (Fig. 2D). In contrast to C. elegans and Drosophila embryos, which have specific patterns of actomyosin localization (Fig. 2B,C), the starfish oocyte has a dynamic accumulation of actomyosin, and the band of cortical contraction traveling across the oocyte is controlled by a Cdk1 gradient released from the asymmetrically localized nucleus (Bischof et al., 2017) (Fig. 2D). This bears a resemblance to the cytoplasmic flow, membrane deformation and cell shape oscillation caused by asymmetric polar contractions during cytokinesis in cultured mammalian cells (Sedzinski et al., 2011).

Cytoplasmic streaming is not limited to invertebrate oocytes and embryos. In zebrafish zygotes, cytoplasmic flow moves the ooplasm and centrally located yolk granules towards the animal pole, while peripherally located yolk granules are displaced towards the vegetal pole (Fuentes and Fernandez, 2010; Shamipour et al., 2019) (Fig. 2E, Table 1). Although the pattern of cytoplasmic movement in zebrafish embryos looks similar to the reverse fountain streaming pattern in invertebrate embryos, the underlying mechanism is quite different. Both animal pole-directed and vegetal pole-directed cytoplasmic movements are independent of cortical actomyosin contraction. Instead, the animal pole-directed movement is primarily caused by the displacement of an actin mesh network driven by actin polymerization along the animal-to-vegetal axis (Shamipour et al., 2019) (Fig. 2E). The vegetal pole-directed yolk granules are pushed by formin-dependent polymerization of actin comets on the surface of the individual yolk granules (Shamipour et al., 2019) (Fig. 2E). Interestingly, this bulk actin polymerization is also cell cycle-dependent: cell cycle entry triggers the actin polymerization waves promoting cytoplasmic movement (Shamipour et al., 2019). This is distinct from the cytoplasmic movement in the Drosophila syncytial embryo, which is initiated by cell cycle exit (Deneke et al., 2019).

In mouse prophase I oocytes, no coherent cytoplasmic flow pattern is observed; instead, a less directional, more diffuse cytoplasmic movement has been reported (Almonacid et al., 2015). This cytoplasmic movement is driven by active diffusion of actin-positive vesicles and is dependent on myosin-Vb (MYO5B), not myosin-II (Almonacid et al., 2015). The asymmetric actin vesicle activity creates a pressure gradient and a propulsion force that is sufficient to drive the centripetal movement of large objects (with a diameter greater than a few micrometers), including the nucleus (Almonacid et al., 2015; Colin et al., 2020) (Fig. 1D, Table 1).

In contrast to prophase I oocytes, bidirectional cytoplasmic movement occurs in mouse oocytes in meiosis II. Cytoplasmic particles stream along the cortex from the actin-rich cortical cap region and circulate back to the center towards the cap region, resembling the reverse fountain streaming pattern (Fig. 2F). This cytoplasmic movement is mainly driven by actin retrograde flow from active Arp2/3 complex at the actin-rich cortical cap (Yi et al., 2011). When Arp2/3 activity is inhibited, contraction of the non-muscle myosin-II ring flanking the actin cap results in reverse streaming, indicating that Arp2/3-driven actin dynamics provides sufficient force to antagonize non-muscle myosin-II contraction and drive the cytoplasmic movement (Yi et al., 2011) (Fig. 2F, Table 1).

Very few types of cells are greater than 100 µm in diameter, suggesting that physical constraints exist to prevent the formation of extremely large cells. Cytoplasmic streaming is frequently associated with cells larger than 100 µm (Table 1), and thus it has been proposed as a mechanism to overcome the challenges that exist in these large cells.

The most apparent function of cytoplasmic streaming is to mix cytoplasmic contents in large cells. For cells larger than 100 µm, diffusion and individual directional transport over a long distance in the crowded, highly viscous cytoplasm become increasingly difficult. Therefore, cytoplasmic streaming in large plant cells provides a perfect solution to mix and deliver nutrients, large molecules and vesicles essential for metabolism, homeostasis and growth (Goldstein et al., 2008; Verchot-Lubicz and Goldstein, 2010). In Arabidopsis, cytoplasmic streaming is dependent on myosin-XI. Replacing the endogenous myosin-XI with a faster or a slower chimeric myosin-XI motor induces an increase and decrease in plant size, correlated with streaming acceleration and deceleration, respectively (Tominaga et al., 2013). Furthermore, in Arabidopsis roots, streaming velocity is highest in the actively growing zone and is positively correlated with cell length (Verchot-Lubicz and Goldstein, 2010). This suggests that cytoplasmic streaming velocity controls plant size by limiting the rate at which nutrients, metabolites, hormones and organelles are mixed throughout the cell.

Another example of cytoplasm mixing occurs in Drosophila late-stage oocytes, where fast bulk cytoplasmic streaming mixes the yolk-containing ooplasm with the yolk-free cytoplasm supplied by the connected sister cells, known as nurse cells (Fig. 1C). Inhibition of cytoplasmic streaming results in stratification of oocytes, causing the failure of future embryogenesis (Gaspar and Janos, 2009; Lu et al., 2018, 2016; Serbus et al., 2005).

For cells of a smaller size (less than 100 µm), the most common cargo transport mechanism is mediated by molecular motors carrying the attached cargoes along either microtubules or actin filaments (Barlan and Gelfand, 2017). However, when the cell size increases to more than 100 µm, active transport has to reach a velocity range of 10 µm/s for efficient distribution, which becomes increasingly challenging in the crowded cytoplasm (Goldstein et al., 2008; Goldstein and van de Meent, 2015). Instead, cytoplasmic streaming provides an efficient way to deliver cargoes over large distances in a synchronized manner (Verchot-Lubicz and Goldstein, 2010). For example, in the algae Chara, chloroplasts rotate in the internodal cells and move in and out of the light-illuminated area (Fig. 1A). This increases photosynthetic efficiency by avoiding photon oversaturation (Dodonova and Bulychev, 2012). In C. elegans, cortical granules are carried by cytoplasmic streaming to the cortex of the meiotic embryos for efficient exocytosis, which is vital to protect the embryos from osmotic and mechanical stresses (Kimura et al., 2017) (Fig. 1B). In addition, mitochondria churn within the Drosophila late oocytes by ooplasmic streaming and are captured at the posterior end for future germline inheritance (Hurd et al., 2016) (Fig. 1C).

The shuttle streaming occurring in the plasmodium of the slime mold P. polycephalum has been hypothesized to drive the mass transport of nutrients and signaling molecules throughout the entire tubular network via Taylor dispersion (Alim et al., 2013; Iima and Nakagaki, 2012). As a single-celled protist relying on phagocytosis for food intake, this slime mold employs cytoplasmic streaming for food searching (Fig. 2A). A food stimulus is proposed to trigger a signaling molecule that increases local contraction, which translates to a higher cytoplasmic flow in the shortest tube connecting two food sources. The tube with an increased cytoplasmic flow grows at the expense of the tubes with lower flow rates (Alim et al., 2017). Thus, the famous intelligence of the slime mold in navigating the food maze and finding the shortest route between food sources (Nakagaki et al., 2000; Tero et al., 2010) relies on its ability to locally regulate cytoplasmic streaming. Furthermore, fragments of the plasmodium are known to display directional flow-dependent amoeboid locomotion behaviors. The shuttle streaming resembles a peristalsis-like movement and leads to asymmetric cytoplasmic movement and a net displacement of cell fragments (Matsumoto et al., 2008; Rieu et al., 2015; Zhang et al., 2019).

The establishment of polarity in embryos involves positioning cell fate determinants in specific areas to set a blueprint for future embryonic development. This process often involves cytoplasmic streaming (Illukkumbura et al., 2020).

In C.elegans one-cell zygotes, after symmetry breaking by sperm-derived centrosomes, asymmetric actomyosin contraction drives cortical flow towards the anterior, which is essential for localizing the polarity determinants partitioning defective protein-3 and -6 (PAR-3 and PAR-6), as well as atypical protein kinase C (aPKC, also known as PKC-3), at the anterior cortex, and PAR-1 and PAR-2 at the posterior cortex (Goehring et al., 2011; Munro et al., 2004; Nance and Zallen, 2011) (Fig. 2B). Actomyosin-dependent cortical tension leads to the clustering of PAR complexes via PAR-3 oligomerization at the cortex, which facilitates advection transport by anterior-directed cortical flow (Dickinson et al., 2017; Rodriguez et al., 2017; Wang et al., 2017). Consequently, due to mutual exclusion mechanisms, PAR-1 and PAR-2 occupy the posterior cortex, establishing a reciprocal localization pattern with PAR-3–PAR-6–aPKC (Fig. 2B) (Hao et al., 2006; Motegi et al., 2011). Importantly, induced intracellular flow by focused laser beam-driven thermal expansion is sufficient to enhance, reposition or reverse PAR protein localization and thus alter the polarity of the entire embryo (Mittasch et al., 2018).

Unlike C.elegans embryos, the anterior–posterior axis in Drosophila embryos is pre-determined and maintained in the oocytes via localization of anterior and posterior polarity determinants – bicoid (bcd) mRNA and oskar (osk) mRNA, respectively – at opposite poles of the oocyte (Lasko, 2020). Cytoplasmic streaming is known to transport and localize the maternally-loaded osk mRNA-containing ribonucleoproteins (RNPs) (Glotzer et al., 1997; Lu et al., 2018, 2016; Sinsimer et al., 2011) (Fig. 1C) and germ plasm components, such as nanos mRNA and Vasa protein (Forrest and Gavis, 2003; Little et al., 2015; Sinsimer et al., 2013), at the posterior pole. The requirement for streaming in posterior axis determination may be explained by the long distance (∼100–300 µm) these components have to travel from the anterior side, where they initially enter the oocyte, to their final destination at the posterior pole. This precise positioning seems to contradict the non-specific nature of global cytoplasmic streaming involved in their delivery. Therefore, local anchorage mechanisms at the posterior cortex are necessary to achieve this posterior localization of polarity determinants following streaming-dependent delivery. Actin reorganization, the actin-based motor myosin-V, RNA-binding proteins (such as Osk, Rump and Lost), cortical microtubules and microtubule-based motors (kinesin-1 and dynein) are all implicated in capturing and accumulating specific mRNAs and proteins at the posterior pole (Lu et al., 2018; Sinsimer et al., 2011, 2013; Tanaka et al., 2011; Tanaka and Nakamura, 2011).

In vertebrate zygotes (including those of birds, reptiles and fish), ooplasmic reorganization subdivides the zygote into a yolk granule-free animal pole and a yolk granule-enriched vegetal pole (Fuentes et al., 2018) (Fig. 2E). For example, zebrafish embryos undergo extensive cytoplasmic streaming to achieve the ooplasm-yolk granule separation that is essential for the redistribution of maternal determinants (Fuentes and Fernandez, 2010). During this process, bulk actin polymerization leads to the ooplasm moving towards the animal pole via viscous drag of the actin mesh network, while yolk granules are pushed down towards the vegetal pole via the polymerization of actin comets, leading to phase segregation and separation of maternal determinants (Shamipour et al., 2019) (Fig. 2E).

Besides organelles, mRNAs and proteins, cytoplasmic movement can also provide mechanical forces to translocate large intracellular structures, including the spindle and the nucleus, and thus is essential for proper cell division.

In C. elegans meiotic embryos, cytoplasmic streaming is sufficient but not necessary to push the meiotic spindle into the cortex for polar body extrusion, suggesting that it is redundant with other mechanisms in controlling and anchoring meiotic spindle position (Yang et al., 2003) (Fig. 1B). Furthermore, this meiotic cytoplasmic streaming also influences the position of the sperm-derived pronucleus/centrosome complex (SPCC) in zygotes, which serves as a spatial cue for symmetry breaking and anterior–posterior axis specification (Kimura and Kimura, 2020) (Fig. 1B). In Drosophila syncytial embryos, cytoplasmic streaming spreads nuclei along the anterior–posterior axis after fast synchronized divisions to ensure accurate nuclear positioning and mitotic synchrony before cellularization (Deneke et al., 2019; von Dassow and Schubiger, 1994) (Fig. 2C). In mouse prophase I oocytes, cytoplasmic movement driven by active diffusion of actin vesicles positions the nucleus, as well as other large objects, in the center of the oocyte (Almonacid et al., 2015; Colin et al., 2020) (Fig. 1D). When mouse oocytes enter meiosis II, cytoplasmic streaming driven by Arp2/3 activity at the actin cap maintains an asymmetric meiotic spindle position close to the cortex for polar body extrusion (Yi et al., 2011) (Fig. 2F).

Altogether, cytoplasmic movement in large cells provides a powerful way to redistribute, position and maintain the distributions of both large organelles (for example, the spindle and the nucleus) and small molecules (such as mRNAs and proteins) over extensive distances. This ‘go-with-the-flow’ bulk transport presents an efficient way to overcome the physical constraints of large cells.

Intracellular cytoplasmic movement is usually associated with oocytes, because they are the largest cells of the animal body and can grow to more than 100 µm in diameter. Instead of using de novo synthesis of biological materials to achieve such a massive size, oocytes can take advantage of intercellular cytoplasmic movements to obtain materials from interconnected sister cells. In metazoans ranging from insects to humans, oocytes are often connected to a group of germline cells, known as germline cysts (Pepling et al., 1999). Stable intercellular bridges, called ring canals, form within germline cysts as a result of incomplete cytokinesis and have been documented in many organisms, including Drosophila, Xenopus, chicken, mouse, rat, hamster, rabbit and humans (Haglund et al., 2011). It has been proposed that ring canals provide a channel for directional transport of organelles, nutrients, proteins and mRNAs to the growing oocytes (Haglund et al., 2011).

In C. elegans gonads, germline cells are organized slightly differently from those in higher organisms: the plasma membranes of all germline cells are incomplete, making a gonad syncytium. Enlarged, transcriptionally silent oocytes and transcriptionally active younger germline cells are connected to a shared region called the gonad core via cytoplasmic bridges (Fig. 3A). Bulk proximal flow in the gonad core carries cytoplasmic materials, including mitochondria and germline-specific vesicles, into the growing oocytes. This proximal streaming is dependent on actin and non-muscle myosin-II contraction in or near the oocyte areas and thus ‘pulls’ cytoplasmic contents from the gonad core area (Nadarajan et al., 2009; Priti et al., 2018; Wolke et al., 2007) (Fig. 3A, Table 1). In addition to actomyosin contraction, it has recently been shown that hydraulic instability can also drive large germ cells to grow and small ones to shrink, resembling the two-balloon experiment in which the smaller balloon pushes air into the larger connected balloon due to a difference in pressure (Chartier et al., 2021) (Fig. 3A).

In Drosophila ovaries, within an interconnected 16-cell germline cyst, one cell is specified as an oocyte, while the rest of the 15 sister cells differentiate into nurse cells (Hinnant et al., 2020). The oocyte is directly connected with the four largest nurse cells via four ring canals (Doherty et al., 2021; Imran Alsous et al., 2017). Intriguingly, these nurse cell–oocyte ring canals display the highest asymmetry in the density of actin fibers, harboring more actin fibers on the nurse cell side than on the oocyte side (Lu et al., 2021; Nicolas et al., 2009; Riparbelli and Callaini, 1995). Cytoplasmic flow brings organelles, mRNAs and proteins from the nurse cells to the growing oocyte via the ring canals during both mid-oogenesis (Lu et al., 2022) (Fig. 3B) and late oogenesis (Buszczak and Cooley, 2000; Mahajan-Miklos and Cooley, 1994) (Fig. 3C).

Despite the similarity of these two types of cytoplasmic flows, the underlying molecular mechanisms are quite different (Table 1). The cytoplasmic flow in mid-oogenesis is driven by the microtubule minus-end-directed motor cytoplasmic dynein, which facilitates microtubule gliding at the nurse cell cortex (Lu et al., 2022) (Fig. 3B; Movie 2). This is reminiscent of the cortical dynein gliding activity for spindle positioning in mitosis (Laan et al., 2012; Raaijmakers and Medema, 2014; Vaughan, 2012) and microtubule organization in axons (Del Castillo et al., 2015b; Rao et al., 2017). In this process, moving microtubules ‘pull’ cytoplasmic contents, including mitochondria, Golgi units and even inert particles, through the ring canals via viscous drag (Movie 2). In contrast, the late-stage cytoplasmic flow, also known as nurse cell dumping, is mainly driven by the non-muscle myosin-II contraction of nurse cells (Edwards and Kiehart, 1996; Jordan and Karess, 1997; Wheatley et al., 1995). Contraction at the nurse cell cortex increases the surface pressure and thus ‘squeezes’ all the remaining nurse cell cytoplasm into the oocyte (Jordan and Karess, 1997; Mahajan-Miklos and Cooley, 1994) (Fig. 3C). Recently, it has been shown that rather than constant ‘squeezing’, dynamic actomyosin-driven surface contraction waves are essential to complete nurse cell dumping, while cell size-dependent hydraulic transport, reminiscent of the two-balloon experiment, is required for the initial stage of nurse cell dumping (Imran Alsous et al., 2021) (Fig. 3C).

In the mouse ovary, primordial germ cells undergo synchronous division and give rise to germline cysts interconnected with the ring canals from incomplete cytokinesis (Pepling and Spradling, 1998). Following fragmentation of the interconnected cysts, a primary oocyte survives and grows, while the rest of the connected germ cells shrink in size and undergo programmed death, sharing many key characteristics of Drosophila oocyte specification (Lei and Spradling, 2016; Niu and Spradling, 2022; Pepling and Spradling, 2001). During this process, multiple cytoplasmic components, including centrosomes, mitochondria and Golgi material, are transferred from the sister germ cells to the growing oocyte, analogous to the Drosophila nurse cell-to-oocyte transfer (Lei and Spradling, 2016) (Fig. 3D). Remarkably, the transfer of cytoplasmic components to mouse oocytes is dependent on the microtubule network and the minus-end-directed motor dynein (Lei and Spradling, 2016), similar to the dynein-dependent cytoplasmic flow observed during Drosophila mid-oogenesis (Lu et al., 2022) (Fig. 3B). However, unlike the Drosophila ring canals, which increase in size during development to prepare for a high volume of cytoplasmic passage from nurse cells to the oocyte (Cooley, 1998), the ring canals in mouse oocytes remain constant in size. To allow for the passage of cytoplasm, a large membrane gap replaces the intercellular ring canal (Lei and Spradling, 2016), and some germ cells transfer most of their cytoplasm through the large membrane gaps before the onset of programmed death (Lei and Spradling, 2016). This resembles nurse cell dumping in Drosophila late oogenesis (Buszczak and Cooley, 2000; Mahajan-Miklos and Cooley, 1994) (Fig. 3C), a process during which large membrane gaps have been reported recently (Ali-Murthy et al., 2021).

It is tempting to propose that the mechanism of ‘nursing the oocyte’ via intercellular flow is highly conserved across species in metazoans. Nevertheless, at this moment, it remains unclear how the germ cells transfer cytoplasmic contents to mouse oocytes, as the mouse cysts are highly packed, and it is therefore very challenging to perform live imaging on them. Future studies leveraging high-resolution live-cell microscopy, ex vivo culture and 3D organoid technology are needed to determine the exact type of cytoplasmic transfer occurring during mammalian oocyte growth and gain a better understanding of oocyte development in mammals.

Bulk intracellular and intercellular cytoplasmic movements are most prevalent in extremely large cells, such as algae internodal cells, animal oocytes and zygotes. Actin filaments and the actin-based motor myosin-XI provide the driving force for plant cytoplasmic streaming, whereas in animal oocytes and zygotes, various mechanisms involving the actin-based motors myosin-II and myosin-V, and the microtubule-based motors kinesin-1 and cytoplasmic dynein, generate different cytoplasmic movement patterns. Intracellular cytoplasmic movement helps extremely large cells to overcome the physical constraints of their size and plays a role in mixing cytoplasm, delivering cargoes, regulating cell growth and migration, establishing cell polarity, and positioning the nucleus and the spindle. Intriguingly, intercellular cytoplasmic flow has been shown to be essential for oocyte growth in the invertebrates C. elegans and Drosophila. Recent studies suggest that mammalian oocytes might also employ an evolutionarily conserved strategy to acquire large amounts of cytoplasmic contents from their sister germ cells.

First observed in 1774, cytoplasmic streaming has been a focus of scientists' attention for over two centuries. Dozens of streaming types, underlying mechanisms and related biological functions have been identified over the years (Table 1). Undoubtedly, molecular motors and their cytoskeletal tracks are at the center of these amazing self-organized movements. Future studies using quantitative measurements and analysis, optogenetic manipulation, in vitro reconstitution and computational modeling will further advance our understating of how molecular motors drive and regulate cytoplasmic movements.

We thank our reviewers and the editor of this Review for their positive and constructive comments.