Introducing proteins into cultured animal cells

Improved methods of protein purification and antibody production, together with a burgeoning array of techniques for single-cell analysis, have led increasing numbers of cell biologists to try microinjecting functional proteins, or antibodies raised against them, into living cells. This has highlighted the shortcomings of available techniques, and encouraged attempts to develop new methodologies. Here I present a brief appraisal of available microinjection procedures, and give examples of their application. I have limited my scope to the transfer of proteins into cultured animal cells.

The term microinjection is usually associated with direct pressure injection of proteins or other molecules into cells through glass microcapillaries (‘needle microinjection’). This is one of the simplest microinjection procedures, and the one most frequently used, but the term is commonly applied to other methods of introducing macromolecules into cells. Most of the other methods fall into two categories: (1) membrane-vesicle methods in which pre-loaded membrane vesicles (erythrocyte ghosts, liposomes, protoplasts) are caused to fuse with cultured cells and release their contents into the cytoplasm; and (2) physical methods, which rely on macromolecules entering cells by diffusion through holes transiently introduced in their plasma membranes by mechanical means (scraping from the substratum, agitating with glass beads).

In choosing a method, the following points should be considered. (1) How many cells must be injected? (2) How much disturbance to the cells can be tolerated? For example, what time interval may elapse between injection and the start of observations? (3) Is the method suitable for the particular protein of interest? Some proteins are degraded, for example, by components of erythrocyte membranes. (4) How much protein is available? Needle microinjection requires much less material than other methods. (5) How easy and reliable is the procedure, and is any special equipment required? One might also consider whether the method can be used for transferring nucleic acids rather than proteins, and whether the method can be applied to a variety of cell types without major modifications. No one method is universally applicable, but it is hoped that this review will help indicate the most suitable procedure for a given problem, and act as a starting point for further reading.

The most popular transfer method to date, apart from needle microinjection, uses erythrocyte (RBC) ghosts as convenient carrier vesicles for introducing exogenous material into cultured cells by cell fusion (Furusawa et al. 1974; Schlegel & Rechsteiner, 1975). The RBCs are loaded by causing them to lyse and reseal in a hypotonic solution of the protein to be transferred. During lysis, the RBC membranes become permeable, protein diffuses into the ghosts and haemoglobin diffuses out. Lysis and resealing is accomplished either by suspending RBCs directly in a hypotonic solution of the protein (Schlegel & Rechsteiner, 1978), or by dialysing RBCs in isotonic protein solution against hypotonic buffer (Yamaizumi et al. 1978). Concentrated saline is then added to restore isotonicity.

The RBC contents are transferred to the cytoplasm of cultured cells by inducing membrane fusion, using inactivated Sendai virus (Furusawa et al. 1974; Schlegel & Rechsteiner, 1975) or polyethylene glycol (PEG) (Schlegel & Mercer, 1980) as fusogen. Human RBCs fuse more efficiently, and are usually used in preference to RBCs of other species when using Sendai virus for fusion (Furusawa et al. 1974). PEG-mediated fusion has no such restrictions and has been used with guinea pig (Heumann et al. 1984), chicken (McClung & Kletzien, 1984) and human RBCs (McElligott & Dice, 1983). A potential advantage of using avian erythrocytes is that culture cells that have successfully fused with RBCs are marked by the distinctive, condensed erythrocyte nuclei (McClung & Kletzien, 1984).

A further refinement is to cause RBCs and culture cells to adhere during fusion, using phytohaemagglutinin (Schlegel & Mercer, 1980). This facilitates fusion, and allows fusion to occur while the cells are attached to the culture dish, rather than in suspension, thus ensuring minimum alteration to cell morphology. For example, McClung & Kletzien (1984), using phytohaemagglutinin and PEG, were able to fuse more than 90% of BHK cells growing in monolayer culture to at least one chicken RBC.

The RBC method has been applied to a wide variety of cell types, including haematopoietic stem cells (Hosoi et al. 1985), chromaffin cells (Kenigsberg & Trifaro, 1985), hepatoma cells (Katznelson & Kulka, 1985), PC12 cells (Heumann et al. 1984) and muscle cells (McElligott & Dice, 1983), and to cells at different stages of the mitotic cycle (Brown et al. 1985; McClung et al. 1984). It may also be possible, by incorporating antibodies to cell surface markers into RBC ghost membranes, to target specific cell types in a mixed population (Godfrey et al. 1983).

Certain proteins may be degraded by proteases resident in RBC ghost membranes (Netland & Dice, 1985). Other potential problems are: (1) binding of proteins to the ghost membranes during loading; and (2) leakage from RBC ghosts during exposure to the fusogen (Net-land & Dice, 1985). Thus, a certain amount of’setting up’ of the procedure is necessary for each new protein. The main drawback of the method is the relatively large amount of protein it consumes. For example, to load 10⁵ molecules per cell of IgG into 10⁶ cells would typically require 2×10⁷ RBC ghosts, loaded with about 0·5 mg ml⁻¹ of IgG. This would require at least 0·2 mg of pure IgG.

On the other hand, the RBC method enables large numbers of cells to be microinjected, so that they can be subjected to biochemical analysis. The research area that has principally benefitted from this approach is the study of the intracellular pathways of protein degradation (for references, see McElligott & Dice, 1983; Katznelson & Kulka, 1985; Doherty et al. 1987). Other applications are still uncommon, but examples include investigation of the role of cyclic AMP-dependent protein kinase in the cell growth cycle (McClung & Kleitzen, 1984), and the mechanism of action of Nerve Growth Factor (Heumann et al. 1984).

When amphipathic lipids such as phospholipids are dispersed in aqueous solution, multilamel-íar lipid vesicles (MLVs or liposomes) are formed spontaneously. These are formed of a series of completely closed sacs each bounded by a lipid bilayer, one within the other like the layers of an onion. MLVS can be converted to small unilamellar vesicles (SUVs) by sonication, and SUVs can in turn fuse to form large unilamellar vesicles (LUVs). The formation, structure and properties of each of these vesicles has been reviewed by Poste et al. (1976). Since soluble proteins and other macromolecules can be entrapped within lipid vesicles during their formation, they promised to be useful for transferring proteins into the cytosol of culture cells by membrane fusion. One of the problems of this approach has been the relatively small ratio of internal aqueous volume to lipid content of MLVs and SUVs (not much work has been done with LUVs), which limits the amount of material that can be transferred without greatly altering the cell’s membrane composition. Also, cells ingest a significant number of SUVs by endocytosis rather than membrane fusion, which further limits their value (Poste et al. 1976).

However, lipid vesicles may be uniquely suited to inserting integral membrane proteins into cells. Trans -membrane proteins can be inserted into the lipid bilayer of MLVs or SUVs by co-incubation in the presence of a non-ionic detergent (see references, quoted by Harris et al. 1984; Fernandez-Botran & Suzuki, 1986). After removal of the detergent, the lipid vesicles can be fused with culture cells using PEG (Fernandez-Botran & Suzuki, 1986), or by means of Sendai virus haemagglutinin/neuraminidase (HN) and fusion (F) proteins incorporated into the vesicle membranes (Harris et al. 1984). (The HN protein mediates cell attachment, and F protein catalyses fusion.) These general approaches have been used to transfer membrane receptor proteins between, for example, B and T lymphocytes (Jakobovits et al. 1982), cytolytic and non-cytolytic T lymphocytes (Harris et al. 1984) and macrophages and T lymphocytes (Fernandez-Botran & Suzuki, 1986). In each of these examples, the ability of donor cells to respond to a specific external stimulus was transferred to the normally unresponsive acceptor cells.

Bacteria from which the cell walls have been removed with lysozyme are known as protoplasts. Like other membrane vesicles, they can be fused to animal cells using PEG, and so proteins synthesized from bacterial expression vectors can be transferred directly into cultured cells. Most interest in protoplast fusion has, however, focussed on transferring plasmid DNA (Sandri-Goldin et al. 1983), and there are only a few examples of the intentional transfer of proteins. Waldman & Milman (1984) transferred functional thymidine kinase (TK), expressed in Escherichia coli, into a TK⁻ animal cell line, and Ferguson et al. (1986) introduced human adenovirus EIA proteins made in E. coli into cultured cells, and showed that the protein could rescue an adenovirus mutant that lacks EIA.

The transferred protein is necessarily contaminated with bacterial proteins and nucleic acids, although with current expression vectors, the polypeptide of interest may comprise as much as 50% of the total protein. Protoplast fusion appears to be rather toxic to mammalian cells; in the study by Ferguson et al. (1986) only 10-20% of cells survived the procedure. As recombinant DNA technology continues to develop, the future of protoplast fusion for transferring proteins is promising, but uncertain.

This appropriately named method, introduced by McNeil et al. (1984), is suitable for cells that grow as a monolayer in culture. The culture medium is replaced with a small volume of solution containing the protein to be transferred, and the cells are scraped off the dish with a rubber-tipped spatula (‘rubber policeman’). This probably tears holes in the plasma membrane at sites of attachment to the substratum, through which protein enters by diffusion. The cells are allowed to recover and are replated.

The method is exceptionally simple and easy to perform, can be applied to many cell types with minimal modification, and loads a high proportion (around 50%) of those cells that survive the procedure (also typically 50%). Like other bulk methods, however, it requires relatively large amounts of protein. Other drawbacks are: (1) cells need several hours to assume normal morphology after scraping/replating, so the method may be unsuitable for studying rapid events such as intracellular protein trafficking; and (2) direct exposure to solutions of some proteins or their analogues may be toxic to cells (e.g. fluoresceinated actin; McNeil et al. 1984).

There are still very few examples of the application of scrape-loading. It has been used alongside RBC-mediated microinjection, with similar results, to study intracellular protein turnover (Doherty et al. 1987). Also, Ortiz et al. (1987) showed that functional enzymes (HGPRT and TK) could be introduced into cell lines lacking these activities, by transferring whole cell extracts prepared from wild-type cells.

Recently, McNeil & Warder (1987) described a procedure, ‘bead-loading’, which attempts to circumvent some of the problems of methods surveyed so far. They covered monolayer cells growing on glass coverslips with a small volume (down to <20 μl) of protein solution, and sprinkled fine (75-500 μm) glass beads on to the monolayer, with gentle agitation. Protein entered the cells, presumably through holes punched in their plasma membranes by moving beads. Since beadloading can be performed on a small scale, it can use less protein than RBC fusion or scrape-loading. Also, cell survival is better than with scrape-loading, and cell morphology is unaltered. Under optimal conditions, nearly 100% of bovine aortic endothelial cells could be loaded with horseradish peroxidase or fluoresceinated dextrans. However, the amount of dextran loaded per cell was highly variable, and not all cell types were as amenable to loading. Until the method has been used more generally, its practical value is difficult to predict.

Microinjection of cells via fine glass needles (Graess-mann et al. 1980; Graessmann & Graessmann, 1983) is by far the most popular method, and has become standard in many laboratories. It has several distinct advantages over other methods. (1) The approach is direct, reproducible and requires a minimum of set-up time. (2) The operator has instant visual feedback on the success or otherwise of an injection, and the number and location of injected cells is known precisely. (3) Very small quantities of material are needed, 2-5 μl usually being sufficient for many experiments. (4) The method can be used equally well for RNA, DNA or proteins, and with a wide range of cell types. (5) Experimental observations can begin within seconds of, or even during microinjection. On the negative side, there is the need for specialized, expensive equipment (an inverted microscope, micromanipulator and needle-puller) and a certain amount of dedication (and practice!) to develop the necessary skills. The major limitation is the small number of cells that can be injected. A skilled practitioner can inject about 300 cells in 15 min, but maintaining this rate is tiring and tedious. This means that for most purposes analysis is restricted to the single-cell (microscopic) level.

This problem has been solved in a variety of ways. For studies of protein transport within the cell, or assembly into cytoskeletal elements, the subcellular redistribution of microinjected proteins can be followed in the light microscope by indirect immunofluorescence or immunocytochemical staining (Goldfarb et al. 1986; Richardson et al. 1988). Proteins bound to colloidal gold particles, either directly or indirectly via gold-conjugated antibodies, can be visualized in the light microscope after silver enhancement (Richardson et al. 1988) or directly in the electron microscope (Mitchison et al. 1986). Fluorescently coupled proteins can be followed in real-time in living cells by image-intensified fluorescence microscopy (Arnheiter et al. 1984; Wang, 1985; Kreis, 1986) and diffusion constants can be measured by fluorescence photobleaching recovery (Kreis et al. 1982). For studies of the effects of microinjected enzymes or antibodies on cell growth, DNA synthesis can be measured by [³H]thy-midine autoradiography (Matuoka et al. 1988; Yu et al. 1988) or more simply with an antibody against bromodeoxyuridine (Gratzner, 1982). Also, there are several morphological correlates of cell growth stimulation or oncogenic transformation that can be observed at the single cell level (Bar-Sagi et al. 1987; Yu et al. 1988).

In conclusion, needle microinjection is for many applications the method of choice, and its popularity is bound to spread with continuing development of microscopical technique. Other methods have their own special attributes. For example, RBC-mediated microinjection and scrape-loading will be useful to generate large numbers of injected cells; lipid-vesicle-mediated transfer may be the preferred method for incorporating membrane proteins into cells; protoplast fusion could have advantages for protein engineering. The only method at present that seems to come anywhere near needle microinjection for conservation of material, simplicity, and lack of trauma to the cell is the untested technique of bead-loading. Those would-be microinjectors, who dread the idea of spending hours wrestling with a micromanipulator, take hope!