The first application of colloid centrifugation to sperm selection was as a density gradient, in which two or more layers of colloids of different densities were used e.g. [19] . The degree of separation achieved depends on the difference in density of the colloids used in the gradient and the difference in densities of the cells to be separated. The technique gained popularity for preparing human semen in fertility clinics, since the volume of semen to be processed was generally small and the sperm concentration low. Typically the volume of colloid in each layer was 1 - 2 mL, depending on the number of layers used, and the volume of semen to be processed was approximately 1.5 mL. The technique had severe limitations when trying to process animal semen which typically has much larger volumes and higher sperm concentrations than human semen. Nevertheless, this technique was also used for preparing frozen animal semen (mostly bovine) for IVF e.g. [20] . Although sperm quality in animal semen could be improved by DGC in the laboratory, the consensus of opinion was that it was too impractical for processing whole ejaculates for AI in the field.

In the1990s, a debate occurred concerning whether Percoll^TM was harmful to spermatozoa, with some researchers reporting adverse effects from the use of PVP-coated silica (e.g. [21] . However, PVP has been used to facilitate the capture of bovine spermatozoa for ICSI without any detrimental effects on the spermatozoa [22] . Subsequently, there were reports of problems with mouse zygote development following exposure of spermatozoa to PVP [23] . These conflicting results may be due to species differences in susceptibility to PVP or may reflect differences in the PVP from various sources [24] . Subsequently, some batches of Percoll were found to have high endotoxin levels [25] , necessitating the testing of each batch to identify those suitable for use with sperm samples. Other problems potentially associated with the use of PVP-coated silica include structural alterations in human spermatozoa [26] , although these were observed after exposure to 10% PVP, a high concentration per se.

Commercially available colloid formulations for DGC of semen became available early in the 21^st century. Stallion sperm quality was improved in the selected samples [27] and the technique could be used to prepare small numbers of spermatozoa from problem ejaculates for low dose insemination [28] . However, sperm yields were too low and the method too impractical for routine application when preparing spermatozoa for conventional insemination in the field.

Towards the end of the 1990s, a modification of DCG was developed utilizing one layer of Percoll^TM for preparing human spermatozoa [29] . Good results were obtained with oligospermic ejaculates but the sperm quality of normospermic semen was not improved. Thus the technique was not considered to be useful at that time, since there was a preference to have a standard protocol for all sperm samples. Similar results were later reported when one layer of Percoll^TM was tested with stallion semen [30] , with sperm quality being improved only in poor quality ejaculates.

Single Later Centrifugation (SLC) using one layer of a species-specific colloid for stallion spermatozoa (Androcoll-E) was reported by Morrell et al. [31] . The advantage of using only one layer of colloid is that the protocol is less time-consuming and easier to use than DGC. The same authors described a series of studies in which “normal” stallion ejaculates were split and processed by DGC and SLC. Sperm quality was improved in both types of selected samples compared to the uncentrifuged samples, and there was no difference in sperm quality between the two colloid centrifugation groups [9] [14] . Results from 250 ejaculates processed by SLC showed that the technique was repeatable and reliable under field conditions for stallion ejaculates at commercial stud farms [16] . In addition, development of a different colloid formulation for use in 50 mL tubes (Androcoll-E Large) enabled SLC to be scaled-up to process whole stallion ejaculates in 4 - 6 of these 50 mL tubes [32] . It would be too laborious to attempt to use DGC in this size of tube. This scaled-up technique has been slowly gaining popularity around the world, particularly for processing problem stallion ejaculates.

It is also possible to use SLC to process 25 mL of extended semen in 100 mL tubes or even 150 mL extended semen in 500 mL tubes [33] . The latter is particularly useful for processing boar ejaculates, which are voluminous. Thus protocols and colloids have been developed for the following sizes of tubes: 15 mL, 50 mL, 100 mL, 500 mL [34] . For optimum results, the specific colloid formulation for a particular size of tube should be used. These modifications in the SLC methodology are shown in Table 1.

The SLC technique was also modified to process small volumes of thawed bull semen. Whereas some studies were reported in which 1 mL of colloid was used in a microcentrifuge tube with a small volume of semen e.g. [35] , it was found that a higher yield of spermatozoa cold be obtained using 1 ml of colloid in a 15 mL tube [36] .

^*Optimal sperm concentration of up to 100 × 10⁶ is appropriate for boar and stallion semen; for bull semen, 50 × 10⁶/mL should be used.

The silane-coated silica colloid is expensive, and therefore being able to use smaller volumes is attractive on economic grounds.

This method separates the spermatozoa from seminal plasma [39] and extender such as cryopreservation medium. Semen extender is added to the semen which is then centrifuged gently; the supernatant is removed and the sperm pellet resuspended in medium or fresh semen extender. Most of the spermatozoa that were present in the original sample are found in the sperm pellet, together with dead and abnormal spermatozoa and cellular debris [40] . However, there are some reports that the most motile spermatozoa may swim-up again from the sperm pellet while the centrifuge is decelerating, and are thus discarded with the supernatant. Up to 20% of the most motile spermatozoa may be lost through this route [41] . Sources of reactive oxygen species are not removed; these may be detrimental to sperm viability or may cause chromatin damage. In an effort to reduce the number of spermatozoa lost, a modification of the sperm washing technique is to use cushion fluid in the bottom of the centrifuge tube. A dense material is layered under the sperm sample before centrifugation, through which the spermatozoa cannot pass; thus the cushion fluid acts as a buffer between the spermatozoa and the walls of the centrifuge tube allowing the sample to be centrifuged at a much higher g force than where there is no cushion fluid. However, such samples also contain more damaged chromatin than samples prepared by other methods [42] , suggesting that increasing the centrifugal force is not necessarily problem-free.

In a modification of the washing technique, low-molecular weight components of seminal plasma have been removed from boar ejaculates by dialysis [43] . More recently, a sperm filter similar to a coffee filter was reported to separate spermatozoa from seminal plasma without the need for centrifugation [44] . The recovery rate after the sperm filter was greater than after sperm washing (89% versus 81%).

Migration techniques rely on the ability of spermatozoa to move from the extended ejaculate or washed sperm pellet into another medium [25] , thereby separating themselves from seminal plasma. The sperm sample can be underneath, on top of, or to one side of the second medium into which the spermatozoa will swim [25] , although the most common arrangement is to have the sperm sample underneath (hence the name “swim-up”). Selection is based on motility; there is no selection for other parameters of sperm quality such as head morphology, viability, acrosome status or chromatin integrity [45] . Sperm samples may exhibit better midpiece- and tail-morphology after swim-up than after sperm washing, e.g. [40] because tail abnormalities may hinder motility and the spermatozoa are therefore retained in the original sample. Migration through, media containing hyaluronic acid may select for spermatozoa with intact membranes [46] [47] . The major disadvantage of migration methods is the low recovery rate, e.g. 10% - 20% [40] thus making it impractical for preparing AI doses in most animal species, although it can be a useful technique for separating spermatozoa from seminal plasma prior to IVF.

Substances such as glass wool and Sephadex beads have been used to select motile, viable spermatozoa with intact acrosomes [48] . Non-viable spermatozoa tend to adhere to the matrix more than motile spermatozoa [49] . The mechanism of action is unclear but may involve adherence of abnormal spermatozoa to the matrix due to surface charges [50] , or proteins present on the sperm surface after capacitation [51] . Bull spermatozoa are reported to show improved freezability after Sephadex filtration [52] . Such methods help to eliminate leukocytes (and thus some sources of ROS) and fewer spermatozoa are lost than with other methods, e.g. a recovery rate of approximately 63% was reported by [53] . However, the filtrate is not as clean as with other sperm separation methods [53] , because not all of the seminal plasma and cellular debris is removed.

Newer technologies, such as microfluidics, electrophoresis, motile sperm organelle morphology examination (MSOME), and birefringence have been advocated for processing human semen [54] . One example is that the microfluidic device was used to select spermatozoa with better chromatin fragmentation, as evaluated by the DNA dispersion test, from oligospermic human samples [55] . However, such methods are too expensive and impractical to be considered for processing animal semen. In addition, the authors claim that eliminating the centrifugation steps should improve the selection of sperm with higher motility, normal morphology and DNA integrity, whereas colloid centrifugation achieves all these outcomes if performed correctly.

Pregnancy rates after AI are known to be linked to sperm quality in the ejaculate [10] ; therefore, selecting the most motile, morphologically normal spermatozoa with intact chromatin should, theoretically, result in a higher pregnancy rate after AI. This was found to be the case in an AI trial in horses, in which insemination of SLC- selected sperm samples resulted in significantly more pregnancies than unselected control samples [57] . The SLC-selected stallion sperm samples also retain motility, viability and chromatin integrity in storage at 6˚C, which unselected controls do not [58] . The SLC samples were still fertile 96 h after semen collection and selection, whereas unselected semen usually cannot be stored for more than 36 h after semen collection. The pattern of ROS-production is changed in the selected samples, with significantly less production of hydrogen peroxide in the SLC-selected samples than in unselected controls. This observation may help to explain the apparently longer life of the SLC-selected spermatozoa than unselected samples.

Fresh bull semen has also been prepared by SLC, using the scaled-up variant of a bovine-specific colloid in 50 mL tubes. Sperm chromatin integrity was reported to be improved in SLC samples compared to controls [59] , although sperm motility was only improved in a haemospermic sample [60] . In other studies, a higher proportion of spermatozoa with high mitochondrial membrane potential was observed in the SLC samples than in controls [61] . Superoxide production was also increased; these two parameters indicate that the metabolic activity of the spermatozoa in the SLC samples may be increased compared to controls [61] .

In some species, such as cattle, the AI industry depends on cryopreserved semen. Cryopreservation not only prolongs the life of the semen sample and enables it to be transported around the world for insemination into females in other countries; it also permits the semen to be “quarantined” until the donors have been shown to be free of disease at the time of semen collection. However, spermatozoa are easily damaged by the processes of freezing and thawing, despite the addition of cryoprotectants; thus the sperm quality of thawed semen may be poor and the resulting pregnancy rate low. At the present time, success with freezing stallion and boar semen is not as good as with bull semen, with the result that the equine and swine breeding industries rely mostly on fresh (usually cooled) semen.

Selection of stallion and boar spermatozoa by SLC prior to cryopreservation results in better sperm quality in the thawed samples than in unselected controls e.g. stallion [42] ; boar [62] [63] . The SLC-selected sperm samples survived longer after thawing than control samples [64] . However, using a different colloid, Mancill et al. [65] were unable to show a difference in pregnancy rates compared to controls when colloid-centrifuged stallion semen was frozen and subsequently used for AI. Clearly further research is needed to optimize the protocols.

The most frequently reported application for SLC in animal assisted reproduction is in preparing thawed sperm samples for IVF e.g. [66] . In this application, the spermatozoa have to be separated from the cryoextender in which the semen was frozen, and from seminal plasma that contains decapacitation factors. Selection of spermatozoa with normal morphology and intact membranes is associated with increased blastocyst formation and quality. It is not only the ability of the spermatozoa to penetrate and activate the oocyte that is important but also the ability of the zygote to continue to develop; chromatin integrity is implicated in this process. The advantages of SLC over “swim-up” are that the selection is not just based on motility, but also on other attributes of sperm quality, the procedure is quicker than for “swim-up” and a higher yield of spermatozoa is recovered. Stallion spermatozoa have been prepared for ICSI using SLC [67] .

Sperm quality may be a problem in rare breeds [68] , particularly since there are a limited number of males from which to choose, and they will not have been selected for sperm quality. Thus, SLC can be useful either in selecting good quality spermatozoa prior to cryopreservation or in processing the thawed sperm samples. There have been reports of the use of SLC to improve sperm quality in several rare species: e.g. goats [69] ; donkey [70] ; brown bear [71] , and domestic cat as a model for wild species [72] . Use of SLC with thawed sperm samples is more appropriate for IVF than for AI in the field but is theoretically possible for the latter if there is access to a centrifuge.

As mentioned previously, epididymal or testicular spermatozoa can be purified by SLC e.g. cat [73] ; stallion [5] . A similar method has been used to prepare wolf spermatozoa free of somatic cells for DNA extraction [74] .

Since silane-coated silica is non-cytotoxic, colloid centrifugation could potentially be used for other types of cells, such as stem cells and immune cells, as well as female reproductive cells. The composition of the salt solution used to dilute the colloid should be altered to suit the types of cells to be selected, and the pH and osmolarity adjusted. The density of the colloid should also be optimized for the particular cell type, depending on the application.