Review of the Literature

The Development of a Novel Antifungal Silage Inoculant

Doctoral Research Thesis
Tomas James Rees, Cranfield University Biotechnology Centre, UK
In Collaboration with the Ecosyl Products Ltd (formerly Zeneca Bioproducts)

©February 1997

Return to the full index to 'Development of a Novel Antifungal Silage Inoculant'
Visitor number: .

Silage making, an overview
Composition of silage
Lactic acid bacteria, taxonomy and physiology
Lactic acid bacteria as silage inoculants
The aerobic deterioration of silage
    Methods for studying the aerobic deterioration of silage
    Microbial causes of the aerobic deterioration of silage
    Preventing the aerobic deterioration of silage
Yeast in mature silage
Antifungal capabilities of lactic acid bacteria
Methods for assessing antifungal activity
The aerobic production of acetate by Lactobacillus plantarum
    The aerobic growth of L. plantarum
    Mechanism for the aerobic production of acetate
    Regulation of the pyruvate oxidase pathway
Aims of this study
A note on the structure of this thesis

Silage making, an overview

Silage is the acid fermented product of anaerobic fodder fermentation, the aim being to preserve summer crops for winter feeding. Silage manufacture has been practised since ancient times, with archaeological evidence for silos and silage making stretching back into the second millennium BC (Woolford, 1984a), but the modern era of silage manufacture was ushered in by a French farmer, A. Goffart. According to Henderson (1987) he, in 1877, advocated a technique involving the rapid filling and sealing of the silo to generate anaerobic conditions, which is still the basis of silage manufacture today, because the principal objective of the modern ensilage process is to rapidly achieve anaerobic conditions.

The exposure of forage crops to air for any appreciable period of time results in a large increase in the activity of aerobic micro-organisms, leading to the loss of nutritive value and eventually an inedible, possibly toxic, product. By excluding air the growth of anaerobic organisms is facilitated, resulting in the production of ethanol and a variety of organic acids such as acetic, butyric and lactic acid. The objective of silage fermentation is to encourage the growth of lactic acid producing bacteria^† and to inhibit undesirable and potentially hazardous organisms. These organisms, principally clostridia, listeria and enterobacteria, degrade amino acids and produce unpalatable butyric acid (Lindgren et al., 1987), as well as causing diseases such as botulism and listeriosis. In well made silage, the production of lactic acid will result in a pH of around 4, which is usually sufficient to inhibit most organisms, partly by the direct effect of hydrogen ions but mainly by enhancing the toxic effect of organic acids. The result is a sharp reduction in metabolic activity and, so long as air continues to be excluded, a potential storage time of many months.

There are several techniques whereby anaerobic conditions may be achieved, but they all involve sealing the crop in an airtight silo. The construction of these silos is very varied, with the most popular forms being the clamp silo, effectively a three walled bunker, and the tower silo. Tower silos are usually used with dryer crops since the higher pressures exerted can lead to excessive effluent production. An alternative technique which has been gaining popularity in recent years is the preservation of silage in big bales, which are wrapped in plastic sheets or sealed in plastic bags and then stacked for the fermentation period. This technique has the advantage of lower fixed overheads and decreased effluent production, but imperfect wraps and the dryer fodder used can leave the bales susceptible to aerobic spoilage.

Any produce which has sufficient fermentable carbohydrate may be ensiled, but the most popular raw materials are grasses and maize (Woolford, 1984a), with other crops including sorghum, barley and lucerne. The choice of crops for ensilage depends primarily on the local environment, but sometimes waste products such as the fibrous residues left after juice extraction from fruit, or sugar extraction from beet or cane, can be used. In 1992, 49.1 million tonnes of grass or legumes and 2.2 million tonnes of arable crops were ensiled in the UK (Anon, 1993).

The fermentation of crops to produce silage is a complex process with many uncontrolled environmental factors, and if the silage upon opening is found to be of poor nutritional value then this will affect the milk yield, weight gain and general health of the animals (Brookes & Buckle, 1992). For this reason there is a strong commercial pressure to develop a technique which produces consistently high quality silage. One method which has been successfully used is to inoculate the silo with a specially selected lactic acid bacterium (LAB) which is known to possess good silage making qualities. The criteria by which a LAB is selected will be discussed later (Section 1.4), but it will suffice to say here that the use of the right strain can result in silage in which the overall metabolic activity is reduced, the production of undesirable compounds is restricted, and the growth of hazardous organisms is inhibited.

Composition of silage

The composition of silage depends upon the crop ensiled, but the most important controllable factor determining silage quality is the water content. It is usually referred to indirectly as the dry matter (DM), which is defined as the sum total of the other constituents (including volatile organic components) – after wilting, the DM content will be around 15–40%. The DM is the parameter most frequently quoted when discussing silage quality. It influences the microflora directly, selecting for those organisms best able to survive in wetter or dryer environments, and indirectly, because dried fodder will contain a higher concentration of sugar and other soluble components. It also affects the amount of effluent produced.

The DM content is primarily manipulated by a period of wilting, which occurs after the crop is cut, but before it is ensiled. The crop is left for a period of time in the field to dry, and the degree of wilting is controlled by the length of this period (usually one or two days). The rate at which wilting occurs depends on the prevailing weather conditions, and also on the crop species being wilted. Charmley & Veira (1991) wilted lucerne for twenty-four hours, and the DM content was found to increase from 17% to 28%. By contrast Rooke et al. (1985) found that the DM of ryegrass, a comparatively hardy crop, increased from 17% to only 23% after a 48 hour wilt. As well as the effects of wilting on DM, the DM also increases with the age of crop in the field, with crops cut late in the season tending to be dryer.

The dry matter portion of fodder at ensiling consists of soluble and insoluble carbohydrates, protein and soluble nitrogenous compounds, and ash (mineral components remaining after complete combustion), all in variable proportions. A synopsis of the constituents of fresh herbage and silage, as reported by a number of independent researchers, is presented in Table 1.1. The reports are highly variable and this reflects the problems encountered in silage research – in that the material under study varies from season to season, and from location to location. Some of this variability can be ascribed to operator effects or to the specific mechanics of harvesting and ensiling the crop, but a great deal is due to factors which are beyond control, such as the weather, the soil conditions, and the microflora of the standing crop.

The dry matter and protein content are largely unchanged by ensilage, but water-soluble carbohydrates (WSC) are consistently degraded. This can be attributed to the action of micro-organisms, with the missing WSC either converted to organic acids or fully oxidised.

Table 1.1(A). Composition of fodder from a range of sources immediately before ensiling. Values are a percentage of total weight (for DM), or a percentage of dry matter (other values). References cited are listed below the table.

Key:

- Data not available
DM: Dry Matter
ADF: Acid Detergent Fibre
WSC:        Water Soluble Carbohydrate
Soluble N:    Soluble Nitrogen
Ref:            Reference

Table 1.1(B). Composition of silage from a range of sources immediately after opening the silo. Values are a percentage of total weight (for DM), or a percentage of dry matter (other values). References cited are listed below the table.

References: 1, Charmley & Veira, 1991. 2, De Figuieredo & Marais, 1994. 3, Haigh & Peers, 1992. 4, Haigh, 1995. 5, Jonsson et al., 1990. 6, Lunden Pettersson & Lindgren, 1990. 7, McDonald et al., 1960. 8, Mir et al., 1994. 9, Muck & O'Kiely, 1994. 10, Rooke et al., 1988. 11, Rooke et al., 1985. 12, Sharp et al., 1994. 13, Wyman et al., 1995. 14, Ashbell & Lisker, 1988. 15, Rust et al., 1989.

Lactic acid bacteria, taxonomy and physiology

Lactic acid bacteria have been used for centuries in the preparation and processing of foods and beverages, and are nowadays used in numerous fermentation processes. As well as silage manufacture, they are used in the manufacture of fermented dairy products, in the production and preservation of sausages and meat, in the fermentation of olives and vegetables, and in baking.

They are a diverse group of genera which can be characterised as Gram-positive, catalase negative, non-sporulating, non-pigmented mesophils (Brookes & Buckle, 1992). The tolerated temperature range is generally between 5 and 50° C, with the optimum for most strains being about 30° C (McDonald et al., 1991). Shape is variable, from cocci through to elongated rods. Although metabolically similar, there is a lack of DNA homology between them. The most important genera in terms of silage microbiology are the lactobacilli, which are non-motile, Gram-positive, obligate saccharolytic fermenters. They lack the pathways for nitrate reduction, for the production of catalase, and for the production of cytochromes and other pigments. They also have a complex and variable nutritional requirement, differing according to species (Bottazzi, 1988).

LAB species can be subdivided on metabolic grounds into three groups according to Kandler & Weiss (1984). These subdivisions are based on the principal saccharolytic pathway employed by the species:

• Group I, obligate homofermentative (Figure 1.1); these convert hexoses into lactic acid via the Embden-Meyerhof pathway, but they are unable to ferment pentoses or gluconate.

• Group II, facultative heterofermentative; usually ferment hexoses homofermentatively into lactic acid but, in some strains and under some conditions, heterofermentative metabolism (Figure 1.2) into lactic acid, carbon dioxide and ethanol (or acetic acid) occurs. Acetic acid production occurs under conditions where NAD⁺ can be regenerated without the formation of ethanol, for example through the reduction of fructose or molecular oxygen. Pentoses are fermented into lactic and acetic acid via a phosphoketolase. This group includes several organisms important in silage, especially Lactobacillus plantarum.

• Group III, obligate heterofermentative (Figure 1.2); hexoses are fermented to lactic acid, carbon dioxide and ethanol (or acetic acid in the presence of an alternative electron acceptor). Pentoses are converted to lactic and acetic acids.

Demonstrating the classification of a species into either group I or group III can be problematic, since the circumstances which cause a group II organism to switch between homofermentative and heterofermentative pathways are variable and not well understood. Species originally classified as group I or group III have subsequently been found to belong to group II, and it may be hypothesised that, in fact, most LAB species fall into group II.

A variety of strain and species typing methods have been developed for use with LAB (for a review see Dykes & von Holy, 1994). Phenotypic typing methods have been the most commonly used, and a dedicated kit for use with LAB, the API 50 CHL system, has been developed by BioMerieux of France. This kit is mainly useful for identification to species level, and its precision can be greatly improved by the computerised application of Bayes’s theorem (Cox & Thomsen, 1990). More recently genotypic techniques, such as plasmid profiling and ribotyping, have been developed, which allow a more consistent and accurate identification of individual strains.

The diversity of the LAB group is reflected in mature silage, and examples of species which have been isolated from silage is shown in Table 1.2.

Figure 1.1. The homofermentative pathway of lactic acid bacteria. Adapted from McDonald et al. (1991).

Figure 1.2. The heterofermentative pathway of lactic acid bacteria. Adapted from McDonald et al. (1991)

Good silage depends upon a rapid drop in pH to prevent the growth of clostridia and enterobacteria, which in turn depends upon a rapid and effective fermentation. The speed of the fermentation may be improved by adding selected LAB, in sufficient numbers to dominate the epiphytic flora, to the fodder before ensilage. Ideally, the inoculum should provide 10⁶ CFU g^-1 of fresh crop (McDonald, 1991) although, in a recent survey (Cunningham, 1994), 19 out of 101 inoculant samples tested failed to provide this number.

The criteria which the ideal silage inoculant would meet were first formulated by Whittenbury in 1961 (Seale, 1986). These original criteria have been reformulated several times (for example Brookes & Buckle, 1992, McDonald, 1991, Seale, 1986, Woolford & Sawczyc, 1984), but the core features are generally considered to be the following:

1. Rapid growth and successful competition with the natural microflora.

2. Homofermentation of sugars and quick production of lactic acid.

3. Acid toleration (to approximately pH 4).

4. Fermentation of a wide range of sugars.

5. No production of unfermentable dextran from sucrose

6. No production of mannitol from fructose.

7. No degradation of organic acids.

8. Growth or at least survival at temperatures up to 50° C.

9. Good growth on wilted grass with low moisture content.

10. Must be possible to formulate in powdered/granular form, and must remain stable in this form during storage.

In addition, it has been suggested that, in order to avoid lactate acidosis, strains which produce purely the L(+) lactate isomer should be employed (Hellings et al., 1985). For specialised uses, other criteria are also important. Ohmomo et al. (1995) reported the selection of a strain of a heat-tolerant Lactobacillus, with potential use as an inoculant in tropical regions.

Strains of LAB which fulfil all these criteria are rare. Woolford & Sawczyc (1984) examined, under laboratory conditions, a selection of twenty-one LAB strains and found that none fulfilled them all. They did, however, find three strains which fulfilled most of them – one of Streptococcus durans, one of Lactobacillus acidophilus, and one of Lactobacillus plantarum.

In practice, most silage inoculants contain at least L. plantarum, but other Lactobacillus species are also often used. The range of inoculant products is large – in 1988 there were at least 40 on the market, and details of the strains found in them have been published (Anon, 1988). L. plantarum growth is often poor at a pH above 5 and in aerobic conditions and, because of this, inoculants are often formed from a mixture of L. plantarum complemented with some other species. A popular choice is Enterococcus faecalis, which is better suited to the initial conditions found in the silo (Seale, 1986). Pediococcus spp. are also popular as co-inoculants with L. plantarum.

Inoculation with a strain selected according to the above criteria will result in silage with more lactic acid bacteria, and a greater proportion of homofermentative LAB. The result is a faster pH decline and more digestible silage, as demonstrated in grass silage by Harrison et al. (1989). These effects have also been found with alfalfa, corn, sorghum and wheat silage (Ely et al., 1981).

An inoculum is most useful in cases where the fodder is low in water soluble carbohydrate (WSC), or has a high buffering capacity because, in these cases, the most efficient possible use of the available carbohydrate must be made (McDonald et al., 1991). Classic examples of crops that are low in WSC and have a high buffering capacity are clovers and lucerne, both of which are difficult to ensile without the use of an inoculant.

The aerobic deterioration of silage

Ensilage depends upon the successful exclusion of air, and in a well sealed silo less than 0.5% of atmospheric oxygen is left after 30 minutes (Woolford, 1990). However the amount of oxygen required to allow maximal aerobic spoilage in the conditions found in silage is very low – possibly less than 1% v/v (Lisker et al., 1989) – so even this low level of oxygen is sufficient to allow some growth of aerobic micro-organisms, and the problem is exacerbated in cases where the silo has been imperfectly sealed. The modern technique of big-bale silage is particularly prone to spoilage, because of the high surface to-area to volume ration of the bale, the ease with which the cover can be damaged and the lower packing density of the ensiled grass. The growth and metabolism of aerobic organisms will result in a silage with lower lactate and higher acetate content, and with higher pH and volatile nitrogen values (Ruxton & McDonald, 1974). This poor quality silage will then allow the growth of non-acid tolerant pathogens like clostridia and listeria (Ruxton & Gibson, 1995), which are potentially deadly.

When the silo is opened to remove stored material the ingress of air is inevitable. The pH and temperature of the silo rises as organic acids and residual WSC are degraded, and there is an increase in volatile basic nitrogen (Ohyama et al., 1975). There is also a loss of dry matter content. This aerobic growth rapidly degrades the energy content of the silage, and will often decrease palatability and reduce voluntary intake (Davies, 1993). Severe deterioration will allow the growth of filamentous fungi, which may produce dangerous mycotoxins (Nout et al., 1993, Ohmomo et al., 1994). As well as being directly toxic to cattle, such toxins can also inhibit rumen fermentation (von Maiworm et al., 1995). With aerobic spoilage causing an average loss of 5% of ensiled material, the cost to UK agriculture is estimated at about £110 million (Woolford, 1990).

The chemical composition of a silage, most importantly its pH, can be important in predicting the rate of spoilage. Good quality silage, with a pH of below 4, tends to be unstable in air (Woolford, 1978, O'Kiely, 1989b). This is because an effective and rapid fermentation to a low pH will restrict the growth of organisms that produce short chain fatty acids (especially butyric acid produced principally by Clostridia spp). These SCFAs have a wide-ranging antimicrobial effect, and will inhibit the organisms responsible for aerobic deterioration. Where butyric acid is present in silage at more than 0.5% w/w, for example, it can effectively halt aerobic activity (Ohyama et al., 1975). On the other hand, high WSC exacerbates the problem of aerobic spoilage. Although it does not bring forward the onset of aerobic deterioration, as soon as aerobic growth does begin, the losses in silos with a high WSC are particularly severe (Ohyama et al., 1975). This may be because the higher concentrations of sugars found in these silos stimulate the activity of aerobic organisms and facilitate the degradation of lactic acid.

The primary factor, then, in determining the aerobic stability of a silo is the spectrum and quantity of organic acids it contains. Poor quality silage, with a large amount of SCFA, will be stable in air, whereas good, high quality silage will not. This is the predicament facing silage manufacturers. Fortunately, this rule is not hard and fast, and sometimes well preserved silage does remain stable in air, for reasons which are currently unknown. For example, O'Kiely (1989a), found that the bulk chemical composition of well preserved silage is not related to the degree of aerobic spoilage they suffer and elsewhere (O'Kiely, 1989b) he suggests that this stability may be ascribed to small amounts of unidentified chemicals formed within the silage.

Methods for studying the aerobic deterioration of silage

The most realistic, but also the most poorly controlled and time consuming method, for studying the aerobic deterioration of silage, is to repeatedly sample actual silos on the farm. The sheer size of a farm silo makes it difficult to perform experiments in replicate, and the lack of environmental control can make the results difficult to interpret.

For this reason, the most common method for investigating the aerobic spoilage of silage is under laboratory conditions. The silage can be made in the laboratory, usually in sealed plastic tubes, or it can be removed from farmyard silos (Henderson et al., 1979). Although fresh fodder is usually used in laboratory scale silos, frozen material can also be used. As might be expected, freezing and thawing will affect both the chemical composition and the microbial populations of the fodder but Wyman et al. (1995) have concluded that, when used judiciously, the ensilage of thawed fodder can be a useful experimental tool.

Usually, samples to be studied are placed in containers exposed to the atmosphere and the temperature, microbiological and chemical composition measured at intervals.

When comparing two or more treatments for aerobic stability, an objective end point must be used , but the criteria which have been put forward are almost as numerous as the researchers that have investigated the problem. This is a reflection of the complex nature of aerobic spoilage, which may manifest itself in different ways depending on the silage studied. Methods used depend on the utilisation of substrate by the aerobic organisms, and their consequent caloric production, and on microbial analysis.

In theoretical terms, the total thermal output produced by aerobically respiring organisms is the best indication of aerobic spoilage, since it is the loss of energy from the silage that is usually of most concern to the farmer. This can be estimated by the time taken for the silo to reach a certain fixed temperature increase over the ambient, as was done by Rooke (1990a). To do this the sample should be kept adiabatically in polystyrene boxes to simulate the low thermal transfer through a large mass of silage (Ohyama et al., 1975). If the thermal transfer from the aerobically-spoiling sample is low, its temperature will be approximately equal to a factor of its thermal output and its heat capacity. In closely controlled circumstances where the heat capacity of the replicate silos is constant this method will produce a good estimate of heat production.

Thermal output can be difficult to estimate in practice, and so a variety of chemical and microbiological endpoints are often used instead. An effective method to integrate aerobic metabolism is to measure CO₂ production, which can be done on a laboratory scale by allowing it to flow freely into a caustic pool at the bottom of the fermenter. The CO₂ is trapped and may be titrated at the end of the experiment (Ashbell et al., 1991, Crawshaw et al., 1980). CO₂ production can also be measured on-line, by flushing the samples with a constant flow of CO₂-free air and measuring CO₂ in the effluent with infra-red gas analysis (Brookes, 1990). A better measurement of respiratory activity is the biological oxygen demand (BOD), which has been used by Pahlow (1981). This technique avoids any confusion caused by the release of CO₂ from anaerobic metabolism.

Other techniques used include measuring loss of WSC, or increase in aerobic organisms. These methods are less satisfactory since they offer only indirect measures of aerobic spoilage, but they can be useful when combined with other endpoints. For example Brookes (1990) used a yeast count of greater than 10⁵ combined with a raised CO₂ production to indicate aerobic deterioration. This, of course, assumes that yeast numbers are closely linked to aerobic spoilage, but Brookes’ method does have the advantage of being an integration of chemical and microbial endpoints. Due to the multivariate nature of aerobic spoilage, if thermal output is not to be estimated then a suite of endpoint criteria should be used.

Microbial causes of the aerobic deterioration of silage

The relationship between the composition of a silage and its subsequent rate of spoilage seems to be a very loose one (Henderson et al., 1979). The reasons for this loose relationship are unknown, but it may be due to the variability of microbial population – in most studies of aerobic spoilage, the microbes present are only identified to a very basic level. The relationship between the precise microbial population and the chemical composition of the silage is complex and poorly understood, but the composition determines both the organisms present, and their subsequent rate of growth on exposure to air. Because of this, the organisms responsible for aerobic spoilage will vary from one silage to another, sometimes even when they appear superficially similar. The nature of the organisms responsible for aerobic spoilage is hence still the subject of some debate.

The microflora of the standing crop is very sparse, and it consists mainly of strict aerobes (McDonald et al., 1991). Amongst the anaerobes, coliforms are the most abundant. Also present are smaller numbers of Bacillus, Clostridium, LAB and yeasts. Lactic acid bacteria are present in numbers of around 10⁴ to 10⁵ per gram fresh weight (Moran et al., 1990, Rooke, 1990b), but the epiphytic numbers of other organisms are less well known. In addition to the epiphytic microbes, passage through harvesting machinery can serve to inoculate the crop with a variety of organisms (Fenton, 1987).

When the crop is cut for harvesting, breakdown of the plant tissue stimulates the growth of both aerobic and anaerobic organisms, which can result in a considerable increase in numbers in the time before ensilage, especially if the crop is left to wilt. After the silo has been sealed oxygen is rapidly depleted, which favours the growth of anaerobic and micro-aerophilic organisms. The rapid production of lactic and other acids inhibits the growth of non-acidophiles, among the most important being pathogens such as Listeria monocytogenes (Donald et al., 1993, Grant et al., 1995), and by the end of the fermentation the microflora is dominated by species of lactobacilli.

When the silage is opened for feed-out, the microflora are once more given a supply of oxygen, and aerophilic organisms begin to proliferate. Initially, growth is slow, as the pH is still very low, but the pH rises as lactic acid is catabolised (Ohyama et al., 1975). Other organic acids are also degraded, most notably butyrate and propionate, (which have a wide ranging antimicrobial effect at low pH), but it is only after most lactic acid is degraded that the majority of aerobic fungi and bacteria can thrive, and an explosion of growth ensues (Woolford & Wilkie, 1984). The increase in pH and loss of lactic acid is not as rapid as might be expected because, as the pH rises, lactic acid bacteria also resume growth on the residual sugars (Moon et al., 1980).

It seems, then, that the oxidation of lactic acid is the all-important initial step in aerobic deterioration. The species which are responsible for its catabolism are still the subject of some debate, with the prime candidates being species of yeast and acetic acid bacteria (AAB). In 1978 Woolford was the first to suggest that yeast are responsible for deterioration in grass whereas AAB are responsible for deterioration in maize. Although by 1990 the message had changed slightly ("The consensus of opinion is in favour of fungi, and yeasts in particular, as having a large, if not exclusive role in the deterioration process in silage made from a variety of forage crops" [Woolford, 1990]), the evidence does suggest that AAB are very important in non-grass crops.

Evidence that yeast play an important role in the initial stages of lactate assimilation is convincing. They are the principal acid-tolerant fungal species, and aerobic stability has been found to be inversely related to the numbers of lactate-assimilating yeast in ryegrass, alfalfa and maize silage (Rooke, 1990a, Selmer-Olsen, 1990, O'Kiely et al., 1987).

Their effect has been investigated by inoculating yeast strains into the fodder before ensilage, or into the silage itself. When fodder is inoculated with yeast, the result is a silage in which stability is significantly decreased (Moon et al., 1980, Woolford & Wilkie, 1984). This reduction in stability is at least partly due to the reduced production of lactic acid in such silage, rather than the increased numbers of yeast present at opening. Because of this uncertainty, it is better to inoculate the silage after opening. This technique was used by Holden & Blackburn (1987), who inoculated radiation-sterilised grass silage with yeast, and found that the pattern of deterioration was identical to unsterilised silage – very convincing evidence that yeast play a major role in the onset of deterioration.

Another way to analyse the importance of different microbes in the deterioration process is to selectively eliminate them using antimicrobials, which can be added either to the fodder or to the mature silage. When grass is treated in this way to eliminate yeast the resulting silage is more stable in air: Pahlow (1981) found an increased stability in grass silage containing an antimycotic agent, but not in silage containing antibacterial antibiotics. di Menna et al. (1981) also found that grass silage treated with benzoate, an antimycotic, had a lower yeast population and was more stable – despite normal levels of aerobic bacteria. However, when Woolford et al. (1980) applied the antimycotic pimaricin to silage made from ryegrass, red clover, maize, lucerne and tall fescue they found that, although the stability of the silages was slightly greater, the only substantial improvement was with lucerne and tall fescue.

The equivocal results of Woolford et al. (1980) are echoed elsewhere and it appears that, when crops other than grass are used, bacteria may have a greater influence on the rate of aerobic spoilage. When Kung et al. (1991) produced legume silage from fodder treated with the antifungal vancomycin, they found it to actually be less stable in air, despite having lower yeast and mould counts than the control silos, and Pahlow (1981) was not able to increase the stability of maize silage with an antimycotic, as he was with grass silage. He found that maize silage could be stabilised – but only by using both an antibacterial and an antimycotic agent.

This suggests that bacteria are also important in reducing aerobic stability in some cases, especially with maize. In fact in maize silage, Crawshaw et al. (1980) found that CO₂ production was most closely correlated to bacterial numbers, and not at all with yeast numbers. Bacterial species which can cause aerobic spoilage are rare – most bacteria are unable to thrive in the hostile environment found in well preserved silage. In an attempt to find bacteria which can live in such an environment, Aries et al. (1982) investigated a variety of sites (including a silage tower and a silage pit) for bacteria able to use lactic acid at a low pH, and they found that it could only be done by members of the genus Acetobacter. These organisms preferentially oxidise ethanol to acetic acid, but in the absence of ethanol metabolism is switched to the oxidation of lactic and acetic acids.

Muck & Pitt (1993) conducted six trials of a laboratory based study of the aerobic deterioration of maize, and AAB appeared to initiate heating in all six cases – but yeast also appeared to play an important role in three of them. Twenty out of twenty-four strains of bacteria they isolated from aerobically spoiled maize silage were Acetobacter strains (the others appeared to be LAB). Furthermore Spoesltra et al. (1988) found that inoculation of maize silage with strains of AAB isolated from other silos gives a dramatic reduction in aerobic stability. This effect is not seen when grass silage is inoculated with Acetobacter aceti (Rooke, 1990a).

So it appears that, in general, grass silage is spoiled by yeast whereas maize silage is spoiled by AAB (although yeast can play a role), with other silage crops falling somewhere between the two. The chemical and microbiological factors which determine whether silage succumbs to yeast spoilage or AAB spoilage are largely unknown, but the recent development of a number of mathematical models of aerobic spoilage has provided an useful insight into its microbiology (if only by demonstrating how limited our knowledge of these organisms’ physiology is, and how difficult it is to make hard, practical predictions based on current theories). The initial work done in this area was by Courtin & Spoelstra (1990), and by Muck et al. (1991). Their work has been built upon and improved with the model described by Williams et al. (1995), but the usefulness of the Williams model and the Muck model, with regard to determining the microbial causes of aerobic spoilage, is negated by their assumption that AAB are not present in grass silage. Courtin & Spoelstra’s model does include AAB, and it does a reasonable job of predicting the relative stability of different silages – although it does consistently overestimate stability.

In their model, both yeast and AAB were assumed to metabolise lactate and acetate, whereas AAB could also grow by converting ethanol to acetate. AAB were given a much lower growth rate, and were also repressed by a high dry matter. AAB were inhibited by the total lactic acid concentration, but acetic acid was only inhibitory in its undissociated form (making the inhibitory effect pH dependent). Yeast were inhibited only by the undissociated form of both acids. With these characteristics, they found that the stability of a silage is largely dependent on numbers of yeasts and the level of organic acids, but when yeast counts and DM are low, then AAB will predominate.

The model of Courtin & Spoelstra has been criticised by Ruxton & Gibson (1993) for the omission of WSC and the inclusion of only one yeast strain, and these omissions seem likely, if anything, to underestimate the importance of yeast in aerobic deterioration. The characteristics of yeast growth in the model are also hypersensitive to changes in lactic acid concentration at certain pH values (Williams et al., 1995).

Pitt et al. (1991), using the model of Muck et al. (1991), which does not include AAB, also found that aerobic stability was, usually, inversely related to yeast population. They suggested that restricting counts to lactate-assimilating yeast would have increased the correlation with yeast count.

There is also direct experimental evidence which helps to understand why some silos are spoiled by yeast and some by AAB. For the most part, experimental work has been directed towards an understanding of what facilitates yeast growth in silage. Yeast are encouraged by the slow diffusion of air through plastic covers, which can happen with bale silage (Lindgren et al., 1985). Yeast growth is also encouraged by sugars, and inhibited by a high DM content and higher temperatures (Ohyama et al., 1981a, 1981b). The effect that these factors have on the growth of AAB is not known. However, it is known that the advantage swings to bacteria when the silage contains high levels of acetic acid (Ashbell & Lisker, 1988). Bacterial spoilage is also more likely in silos in which the pH drop has not been great (Lindgren et al., 1985). In these poorly preserved silos, secondary (or clostridial) fermentation can occur resulting in the production of short chain fatty acids which have antifungal activity (Woolford, 1975, 1984b, Ohyama & Hara, 1975). In addition, poorly preserved silage tends to have a low DM content and this further inhibits the growth of yeast (Beuchat, 1983, Ohyama et al., 1980).

But these factors do not help us to understand why maize silage should succumb to AAB spoilage, whilst grass is spoilt by yeast. Maize silage tends to have a higher sugar content, higher dry matter, and to be generally well preserved – all characteristics which should favour yeast growth. A simple explanation is that the standing maize crop supports few yeast but many AAB, but this has not been investigated. Thus the problem is, at present, unresolved.

Preventing the aerobic deterioration of silage

Currently, the best way to prevent or minimise aerobic spoilage is to handle the silage in such a way that exposure to air is minimised (O'Kiely 1989a). Air infiltration during storage occurs by two mechanisms, simple diffusion (Muck & Pitt, 1993), and a hydrostatic pumping effect as dense CO₂ sinks out at the bottom to be replaced by fresh air at the top (McGechan & Williams, 1994). Infiltration can be minimised by careful sealing but, in a standard pit clamp, this is not usually practicable, and a more realistic technique to prevent oxygen influx is to produce well consolidated silage. Unfortunately, the results of Vreman & Bosma (1987) indicate that even short chopped and heavily compacted silos are not much better preserved than poorly consolidated ones.

Once the silo is opened total exclusion of air becomes impossible. Using specialised block cutting machinery to take silage from the silo can help to reduce the penetration of air, but the work of Muck & Huhnke (1995), suggests that the benefits of this technique are limited. The problems of aerobic metabolism are exacerbated in the modern technique of big-bale silage, which uses high dry matter forage wrapped in a plastic covering. The use of HDM enables the bales to be manipulated more easily, but precludes good consolidation, and the low water content encourages yeast growth over bacterial growth (Beuchat, 1983, Ohyama et al., 1981b). In addition, the bag used in this technique is partially permeable to oxygen and is also sensitive to mechanical damage (Jonsson et al., 1990). Preventing aerophilic growth by restricting oxygen influx is, in these circumstances, almost impossible

The alternative is to minimise aerobic spoilage by chemical or microbiological means. In general this can be summarised as either artificially acidifying the silage, applying antibiotics which kill some or all of the microflora, or using an inoculant with an antimicrobial capacity.

Treatment of crops with acids prior to ensilage has been used as a way to produce a consistently low pH without fermentation, but a low pH in itself is not sufficient to prevent aerobic deterioration, as many yeast are able to withstand the lowest pH which can be achieved in practice. In contrast to inorganic acids, organic acids all have an antimicrobial activity which is independent of their effect on pH. By choosing the right one, it is possible to achieve a reduction in pH, and to selectively inhibit those organisms responsible for aerobic spoilage.

Formic, acetic and propionic acids, used alone and in combination, have all been found to inhibit aerobic spoilage (Crawshaw et al., 1980, Lindgren et al., 1988, Sebastian et al., 1996). Formic acid has primarily an antibacterial effect whereas propionic acid also has an antimycotic effect and so, although it has a higher pKa, propionic acid is generally more effective at preventing aerobic spoilage. When it is added to fodder at a concentration of 1%, the resulting silage does not undergo an increase in fungal growth, and it is generally more stable than untreated silage or silage treated with formic acid (Britt et al., 1975). Acetic acid has a somewhat lesser antibiotic effect, and it is only effective at stabilising silage when added at a concentration of around 4% (Woolford 1978).

The main problem with using acids as a preservative is their hazardous nature, and the safety precautions which must be observed with their use have been largely responsible for a decrease in the use of acid additives in recent years. In addition, an acid additive will increase the volume of effluent produced by the silage, and the presence of acetic, butyric, caproic and valeric acids in grass silage has a deleterious effect on intake (Gill et al., 1988).

Formaldehyde also has well known antibiotic affects, but Barry et al. (1980) found that the use of formaldehyde as an additive aggravates the problem of aerobic spoilage. This is because it inhibits fermentation, resulting in a high pH which facilitates the growth of spoilage organisms. They also found that formaldehyde appears to push the balance in favour of bacteria as a cause of aerobic spoilage. Because formic acid both lowers the pH of the silage and selectively inhibits bacteria, a formaldehyde/formic acid additive might be expected to produce a stable silage, and this has indeed been demonstrated by Davies (1991).

More complex antibiotics have also been tested for their ability to prevent aerobic spoilage, but their use is rare because of their extremely high cost when compared to other potential additives. One antimycotic, pimaricin (an antimycotic produced by Streptomyces notalensis), has been tested. Woolford et al. (1980) found that pimaricin was effective in increasing the aerobic stability of silage, even though very little pimaricin could be recovered from the silos (and the recovery rates decreased sharply during storage – presumably due to microbial degradation). Benzoate and vancomycin have also been used, with mixed results (di Menna et al., 1981, Kung et al., 1991).

The use of inoculants to retard aerobic spoilage has many attractions – not least its convenience. The standard inoculants which are currently used are an effective way to reduce yeast and mould numbers within the anaerobic silo (O'Leary & Hemken, 1985, McAllister et al., 1995), and well-produced silage will have, on opening, low counts of aerobic organisms. This protective effect of LAB is still evident where small amounts of air are suffused into the silo (Pahlow & Zimmer, 1985). Unfortunately, once the silo is opened, aerobic deterioration appears more prevalent in silage which is of high quality (Woolford, 1990), and the use of homofermentative LAB inoculants often decreases aerobic stability (Kung et al., 1991, Rust, et al., 1989, Weinberg et al., 1993). This is because the rapid reduction in pH which takes place in high quality silage will inhibit the growth of Clostridia – beneficial in terms of silage quality, but the short chain fatty acids which are produced by Clostridia are very effective in preventing aerobic fungal growth.

An ideal inoculant would be able to both produce good quality silage by rapidly decreasing the pH, and produce an agent which will minimise aerobic spoilage. The antifungal capabilities of LAB are discussed in Section 1.7, but it will be sufficient to say here that they have not been widely explored.

A possible alternative is to use a dual inoculant, containing both lactic acid bacteria and another species which is capable of producing an antifungal agent. A wide variety of microbes are known to have an antifungal action, and the incorporation of suitable strains can significantly reduce mould growth on exposure to air (Moran et al., 1993). A number of Bacillus species are known to produce antifungal compounds (Katz & Demain, 1977). Unfortunately, they will not grow anaerobically and they will not tolerate a low pH, and so they will only activate after aerobic spoilage has already set in. In addition, as aerobic organisms they will actually contribute to the aerobic spoilage process, and these species also produce spores which can cause contamination problems on dairy farms.

Given the retarding effect of propionic acid on aerobic spoilage (see above), it has been suggested that the use of an inoculant capable of producing propionic acid would be a useful alternative to the direct application of acid (Dawson et al., 1993). Propionibacteria are microaerophilic bacteria which, in anaerobic conditions, grow preferentially by the conversion of lactic acid to propionic acid (Gribbon et al., 1994, Hettinga & Reinbold, 1972, Liu & Moon, 1982). The preservative effect of Propionibacterium propionici has been demonstrated in high moisture corn by Dawson et al. (1994), in which it was effective in reducing both yeasts, moulds and Acetobacter populations. However, propionibacteria cannot tolerate a low pH, which lessens their usefulness as an inoculant (Lee et al., 1976). Weinberg, Ashbell, Hen & Azrieli (1995) found that, in maize, Propionibacterium shermanii only gave a slight increase in aerobic stability, mainly because rapid production of a low pH inhibited its growth. It was later shown (Weinberg, Ashbell, Bolsen, et al., 1995) that, in maize silage, propionibacterium are only effective when the reduction in pH is very slow, and the final pH of the silage relatively high. Similarly, a propionic acid-producing bacteria was unable to improve aerobic stability in orange pulp silage (Alio et al., 1994).

Yeast in mature silage

As discussed in Section 1.5.2, yeasts are the most important organisms implicated in silage deterioration. Silage immediately prior to opening (mature silage) is home to a large range of yeasts, which Beck (1978) has classified into two physiological groups:

• The ground-growing or sediment yeasts, which have a high fermentative ability for sugars but variable ability to assimilate lactate.

• The top-growing or pellicle yeasts, which have a weaker fermentation capacity but a high respiration capacity for lactic acid.

The fate of these two yeast groups during the fermentation process has been followed by Jonsson & Pahlow (1984). They found that, in strictly anaerobic conditions, there is a continuous decrease in the number of lactate assimilating yeasts to a level below the limit of detection (around 100 CFU/g DM). The lactate assimilating yeasts constituted around 15% of the final yeast population, and were mainly members of the species C. lambica and S. exiguus. The remainder of the yeast population was found to be made up of Saccharomyces species which were fermentative but could not oxidise lactate.

The exact composition of the yeast flora is highly variable between reports and depends, presumably, on the crop ensiled and the period of ensilage. An overview of the species which have been found in different silages has been provided by Middlehoven & Franzen (1986), and their findings are presented in Table 1.3. The actual contribution of each of these species to aerobic spoilage is not known, but other work has shown that the yeast which are responsible for the onset of aerobic deterioration are only a subset of the yeast species which can be found in mature silage, as suggested by the work of Jonsson & Pahlow (1984).

For example, Candida krusei, Candida lambica, Saccharomyces exiguus, Candida holmii, and Candida milleri are all capable of rapid growth at pH 4.0, and they can all metabolise acetic and lactic acids (Middlehoven & Franzen 1986). Saccharomyces dairensis, however, provides an example of a species which cannot metabolise organic acids under these conditions.

Holden (1987) found that in grass silage S. cerevisiae, C. lambica and other Candida species represented the majority of the yeast population after ensilage, but that C. lambica declined rapidly on aerobic exposure and could not be linked to the loss of lactate. On the other hand Middlehoven & van Baalen (1988) found that the strains primarily responsible for the aerobic deterioration of maize silage were C. lambica, C. holmii, and C. milleri. Occasionally considerable numbers of Candida famata, Geotrichum candidum, and Hansenula anomola were also found but, although all three were found to be capable of assimilating lactic acid at pH 4.0, the contribution of H. anomola to aerobic deterioration was found to be small (the contributions of C. famata and G. candidum were not analysed). Inoculation of maize silage with Saccharomyces exiguus resulted in a large decrease in aerobic stability (Spoelstra et al., 1988), and it may be that further work with yeast inocula will be required to determine which of the ‘lactate assimilating’yeast are actually capable of initiating aerobic spoilage.

It is worth noting that the taxonomy of many yeasts is ambiguous, with the asexual and sexual (imperfect and perfect) forms being known under different generic names. Thus it is believed (Kreger-van Rij, 1984) that all Candida spp. are in fact the imperfect forms of other species. With regard to the species found in silage, the perfect forms of C. lambica C. famata and C. holmii are respectively known as Pichia fermentans, Debaromyces hansenii and Saccharomyces exiguus. In addition, H. anomola is the perfect form of Candida pelliculosa. G. candidum is an imperfect species which, although resembling the yeasts, is not actually classified amongst them.

Antifungal capabilities of lactic acid bacteria

As discussed in Section 1.7, aerobic spoilage might be minimised or prevented by the use of a lactic acid bacteria inoculum which could inhibit yeast growth. Studies on the effect of LAB on fungi are complicated by the fact that fungi are sensitive to the normal by-products of LAB metabolism, most notably acetic and lactic acids (Bonestroo et al., 1993). Since the primary function of a silage inoculant is to maximise lactic acid and minimise acetic acid production, it is important that any observed antifungal effect is additinal to, and not simply a function of, these acids.

Diacetyl is another metabolite commonly produced by LAB which has a wide ranging antimicrobial effect (Jay, 1982). Unfortunately it has a sharp butter-like aroma, and the concentration required (~200 m g/l) is likely to produce unpalatable silage, as well as being difficult to achieve in practice (Piard & Desmazeaud, 1991a).

There are several reports of LAB producing low molecular weight compounds which are capable of inhibiting filamentous fungi. A patent application by King et al. (1986) describes the production, in cucumber juice, of a putative antifungal substance by Lactobacillus casei var. rahmnosus, active against a variety of moulds. However the control experiments were poor, since the control organism (a commercial LAB inoculant, Pediococcus pentosaceus) grew less rapidly and produced less acid. Another patent, by Hill (1989) also describes the discovery of an LAB strain (this time of Lactobacillus plantarum) which will inhibit the growth of an unspecified silage spoilage mould. The nature of the inhibition was not investigated.

Coallier-Ascah & Idziak (1985) found a strain of Streptococcus lactis which would inhibit the growth of Aspergillus flavus when the two were grown in co-culture. This happened even when S. lactis was grown within a dialysis sack, suggesting that the active agent was a low molecular weight compound which was secreted into the medium. They found that lactic acid was not able to inhibit A. flavus but, besides stating that it had been partially purified and lost activity on storage, they did not investigate the nature of the inhibitor further.

Karunaratne et al. (1990) found that a commercial silage inoculant consisting of three Lactobacillus species (Lactobacillus acidophilus, Lactobacillus bulgaricus, and Lactobacillus plantarum) was able to inhibit the growth of Aspergillus flavus, but felt that this effect was due to a combination of acidity and microbial competition. Gourama & Bullerman (1995) further clarified this work using the same inoculant strains. Growth of the Lactobacillus spp. within a dialysis sack was found to inhibit A. flavus spore viability, spore germination, growth and aflatoxin production when the MWCO was 6,000 to 8,000, but not when the MWCO was 1,000. The cell free supernatant was also able to inhibit aflatoxin production, but not spore viability or germination. This effect could not be overcome with the addition of glucose, nor was it due to lactic acid.

Haikara et al. (1994) also reported the discovery of a strain of Lactobacillus plantarum (L. plantarum 601) with antifungal capabilities towards Fusarium spp. and Aspergillus niger. The activity was enhanced by cellulolytic and pectinolytic enzymes, and by siderophores. In related work (Haikara & Niku-Paavola, 1994), L. plantarum VTT-E-78076 was found to produce a substance, active against Fusarium spp, which was purified by vacuum evaporation, gel chromatography, and by ion exchange. The active agent proved to be an oxygen rich organic compound containing the structures CH₃CO and CH₃O. A further, very recent report on L. plantarum VTT-E-78076 (Niku-Paavola et al., 1996) describes a series of culture products active against Fusarium avenaceum and the bacterium Pantoea agglomerans. These compounds co-purified with lactic acid in a gel chromatograph, and were characterised by GC/MS as low molecular weight cyclic organic compounds (Figure 1.3). They were most active at pH 4 and below, and acted synergistically with each other and with lactic acid.

Figure 1.3. Low molecular weight antifungal compounds from Lactobacillus plantarum VTT E-78076. From Niku-Paavola et al. (1996).

Whilst there is good evidence that some strains of LAB produce metabolites, other than acetic and lactic acid, which are capable of inhibiting filamentous fungi, the evidence for such activity towards yeast is more scant. LAB strains with antibacterial activity are well known and widespread (for reviews see Abee, 1995, Holzapfel et al., 1995, Muller et al., 1996, Piard & Desmazeaud, 1991b), but these strains have rarely been tested for activity towards yeast. When this is done, the results are usually negative.

For example, Marth & Hussong (1963) tested cell free filtrates of Leuconostoc citrovorum (with antibacterial activity) for activity towards four yeast species (Saccharomyces cerevisiae, Saccharomyces fragilis, Torula glutinis, and Mycotorula lipolytica), but all the filtrates were found to be inactive. Similarly, Batish et al. (1989) screened the supernatants of 19 LAB cultures using a well diffusion assay, and found that 5 were able to inhibit mould growth. None, however, were able to inhibit any of the 6 yeast species studied (Saccharomyces cerevisiae strains 522, SCB, SC-1; Saccharomyces fragilis strain 3465; Candida guillermondia strain 3124 and Rhodotorula glutinis strain RG). Gavrilova & Zakharenko (1987) briefly described testing a series of organisms used as silage inoculants in Kazakhstan (amylolytic lactic acid streptococci, pentose-fermenting LAB and propionic acid bacteria). They too were unable to find any inhibition of yeast growth.

Whilst these reports failed to find any evidence of anti-yeast effect due to LAB, other work claims to demonstrate such activity, although some of the evidence is equivocal. A patent application (Mäyrä-Mäkinen et al., 1994) claims the use of a strain of Lactobacillus casei ssp. rhamnosus (LC-705) as a broad spectrum antifungal. The effect was only seen when L. casei LC-705 was grown in direct competition with the target organisms – the presence of lactic acid and the effect of pH was not controlled for, leaving open the question of whether a specific antifungal toxin is being produced. The effect was also marginal without the presence of Propionibacterium shermanii, which produces the toxic metabolite propionic acid. The supernatant was unable to inhibit any yeast or mould. A later report (Suomalainen & Mäyrä-Mäkinen, 1995) claims that, when Lactobacillus rhamnosus LC-705 is used as a starter culture with Propionibacterium freundenreichii ssp. shermanii JS (as the commercial formulation Bio profit-preparate) for the production of quark or yoghurt, the growth of added yeast is inhibited. Again, the lack of adequate controls militates against any conclusion regarding the nature of the inhibition.

Collins & Hardt (1980) found that the sterilised filtrate from a culture of Lactobacillus acidophilus was able to slightly retard the growth of Candida albicans when compared to the freshly prepared broth. This result seems unremarkable when consideration is made of the fact that Collins & Hardt did not control for lactic acid levels, nor for the likelihood that the spent medium would contain reduced levels of fermentable carbohydrate.

Other studies provide more definitive evidence for the existence of certain LAB strains with the power to inhibit yeast growth. Makanjoula et al. (1992) examined several strains of Lactobacillus spp. and Leuconostoc spp. for their ability to inhibit the growth and ethanol production of an unidentified distillers yeast. They found that, although all strains tested produced some reduction in yield, there were two strains (one of Lactobacillus brevis and one of Lactobacillus plantarum) which were able to substantially depress ethanol production. The L. plantarum strain caused a marked flocculation of the mixed culture, and it is likely that this played a large role in its inhibitory activity. The L. brevis strain caused no such flocculation, and the mechanism of its effect remained unresolved.

Microgard^TM, a commercially available fermented milk product, can be used to preserve dairy products such as cottage cheese. Al-Zoreky et al. (1990) demonstrated its antimicrobial properties against a wide range of microorganisms, including the partial inhibition of Kluveromyces marxianus and an unidentified black yeast. However Microgard^TM actually stimulated the growth of an unidentified spoilage yeast and Aspergillus niger. Al-Zoreky et al. claimed that the inhibitory activity can be increased by varying the fermentation conditions under which Microgard^TM is produced, but did not provide any details.

Dicks (1994) reported a Lactobacillus sp. which released an antifungal into the supernatant. It was described as being effective against Monilia, which is apparently a loosely defined term referring to a variety of yeast species (including Hansenula anomola and Candida krusei [Kreger-van Rij, 1984]). The agent was of low molecular weight (smaller than 10kDa), and probably acted by permeabilising the cell membrane. It resisted degradation by proteolytic enzymes, and was effective within a pH range of 3 to 7.

Vandenburgh & Kanka (1989) have patented an antifungal product from a lactic acid bacterium, this time from Pediococcus spp, especially Pediococcus acidilacti. The product was a small peptide (400–500 MW), which was heat stable and could be extracted with butanol. The purified compound contained lactic acid and valine, and the activity was susceptible to an unidentified protease, but not to lipase, lysozyme, DNAse or RNAse. It was active against a wide range of filamentous fungi, but the effect on yeast was variable and less well characterised – it inhibits Saccharomyces cerevisiae but not Candida albicans. A brief report by Nout et al. (1987) also described the inhibition of a wide variety of fungi and bacteria by Pediococcus spp, but the nature of the agent is not known.

Perhaps the best characterised antifungal product from an LAB species is reuterin (an equilibrium mixture of monomeric, hydrated monomeric, and cyclic dimeric forms of 3-hydroxypropionaldehyde), produced by Lactobacillus reuteri (Figure 1.4 [Talarico & Dobrogosz, 1989]). Reuterin is a by-product of the reduction of glycerol (for which reuterin production has an absolute requirement), which is an energy yielding pathway (Talarico et al., 1990). As well as toxicity towards a wide range of Gram-negative and Gram-positive bacteria, it has been claimed to be equally effective against lower eucaryotic genera of yeasts and fungi such as Candida, Torulopsis, Saccharomyces, Saccharamycoides, Aspergillus and Fusarium (Chung et al., 1989), although no data to support this claim was provided.

Figure 1.4. Reuterin, an antimicrobial from Lactobacillus reuteri, exists in three forms in aqueous solution (from Talarico & Dobrogosz, 1989)

Methods for assessing antifungal activity

The observed effectiveness of an antifungal agent is critically dependent on the test protocol used to study it. For example, a study involving seven laboratories, each using the techniques with which they were most familiar, found the minimum inhibitory concentration (MIC) for three antifungal agents to vary by as much as 50,000-fold (Rex et al., 1993). The main problems are that the sensitivity of the target strain is dependent on both the media and the conditions (Piddock, 1990), and on the numbers of organisms used. Any variation in test protocol will vary the conditions and so affect the observed antifungal effect. The techniques used to assay antifungal agents are varied, but they can be broadly categorised into two fundamental approaches – using liquid and solid media.

Techniques using liquid media are the most effective way of assessing the potency of antimicrobial compounds – al-Hiti & Gilbert (1982) found that liquid methods were both more sensitive and more reproducible. A nutrient solution containing the agent of interest is inoculated with the indicator organism, and subsequent growth is analysed, usually by turbidometry, but a more recent development measures changes in culture conductivity (Hogg et al., 1987). Liquid media methods are most suitable when the active agent is readily available as a sterile extract from its producing organism. In a primary screen for an antimicrobial-producing organism, the uncertainty introduced by using only an extract of the culture could easily lead to false negatives, because such an extract may miss an active agent which is only transiently present, or the agent may be lost during processing. In this case, it would be preferable to test the activity of an actively growing culture of the test organism, which renders turbidometric methods unsuitable since growth of the test organism tends to mask the inhibition in growth of the indicator organism (for example see Toba et al., 1991) In addition, the two species may flocculate, which will not only affect the optical density of the culture, but also the growth characteristics of the cultures (Makanjuola et al., 1992).

The other methods rely on the diffusion of the antifungal substance through an agar gel, resulting in the inhibition of colony growth in the indicator organism. They are less quantitative than liquid methods, but this is not a severe problem in a primary screen where only a binary outcome is desired. Agar-diffusion methods are also handicapped by the fact that antimicrobial synthesis is sometimes dependent on direct contact between the indicator and test strain. An example of this is in the production of reuterin by Lactobacillus reuteri, where synthesis is markedly decreased even when producer and target strains are separated only by a dialysis membrane (Axelsson et al., 1989). Despite these drawbacks, the simplicity of agar techniques has encouraged their widespread use. Three fundamental techniques have been commonly used; the flip streak method, the well-diffusion method, and the spot-plate method.

In the flip streak method (Lewus & Montville, 1991), the surface of an agar plate is streaked with the test culture. The plate is then flipped, and perpendicular streaks of the indicator organism are placed on the opposite surface. Where the streaks cross each other, colony growth is restricted, and the length of the inhibition zone is proportional to the antagonistic effect. This method is less quantitative than the other two methods, but it is useful for detecting interactions between two strains of which either may be inhibitory (or stimulatory) to the other. The drawback is that, if the indicator is a vigourous competitor, the test organism may not be given sufficient time to express its antimicrobial capacity. This is a particular hazard when screening for the activity of LAB towards yeast. For example Kennes et al. (1991) found that, in a medium simulating orange juice, yeast will always outcompete lactobacilli. Yeast are also known to be overtly toxic towards LAB, mainly due to the production of ethanol, but also as a result of sulphur dioxide production (Daeschel et al., 1987). Another hazard with this method is that agar is brittle, and so is inclined to break when flipped. In addition, the action of replacing the inverted agar tends to smear the streak of the test culture.

In the well-diffusion method, an agar plate is prepared from agar seeded with the indicator organism. A hole (or well) is cut into the agar and filled with the test substance or culture. If the substance inhibits the indicator, then a zone around the well will be clear of colonies, and the size of the zone will be approximately proportionate to the toxicity of the test substance – but zone size will also depend on the diffusion rate. A similar technique is the cylinder-plate diffusion test, where cylinders are placed on the plates and filled with the test substance (Brady & Katz, 1990)

The technique of well diffusion is an improvement over the flip-streak method, in that a fully grown culture can be tested in the well, but the sensitivity can still be restricted in cases where the indicator organism is inhibitory to the test organism. These problems can be overcome, to an extent, with the spot-plate method. This involves making a suspension of the test organism in molten agar, and placing a spot of this onto an agar plate. This is then overlaid with fresh agar seeded with the indicator organism. Antifungal activity is measured by zones of inhibition above the spot, in a similar way to the well diffusion method. Because fresh agar is supplied for the indicator organism, problems due to direct inhibition are minimised, thus increasing sensitivity. The two methods were compared by Lewus & Montville (1991) who found that, of the bacteriocin-producing LAB they found to give a positive result using a spot-plate method, only a few also tested positive in a well-diffusion assay.

The aerobic production of acetate by Lactobacillus plantarum

The production of lactate by lactic acid bacteria is a consequence of their anaerobic growth. Electrons, absorbed by NAD⁺ during glycolysis, are disposed of by the reduction of pyruvate, with its consequent conversion to lactate. If, instead of pyruvate, an alternative electron receiver is provided, pyruvate is freed to be used in other metabolic roles. One possible fate is decarboxylation to form acetate, through one of several alternative pathways, with the concomitant production of ATP. Acetate can also be formed by the direct oxidation of lactate.

Undissociated acetic acid is more strongly antifungal than lactic acid, and so its production in silage may be useful in preventing silage deterioration, especially if it can be limited to the aerobic face. The pathway of acetate formation by Lactobacillus plantarum, and the method of its regulation, means that there is a real possibility for this to happen. The evidence for the aerobic production of acetate, and the mechanism of its regulation, will now be discussed.

The aerobic growth of L. plantarum

In conditions of glucose limitation, the cell yield of L. plantarum is increased by aerobic conditions but the growth rate is unchanged. (Yousten et al., 1975). At higher glucose levels, the picture is a little more complicated, and several studies have shown that growth rate is retarded by oxygen (Archibald & Fridovich, 1981, Murphy & Condon, 1984b, Tseng & Montville, 1992). This inhibition of growth occurs before the increase in H₂O₂ production, and so it is probably a direct effect of oxygen toxicity. However, even at high glucose concentrations, the cell yield is still higher under aerobic conditions, and a higher cell yield in aerobic conditions has been found to be associated with an increase in acetate production (Sedewitz et al., 1984b). These findings are consistent with the hypothesis that acetate and ATP are being generated from pyruvate.

Lactate can also serve as an energy source in aerobic conditions. Murphy et al. (1985) found that the addition of lactate to L. plantarum cultures in the absence of other energy sources would support modest growth in aerobic but not anaerobic conditions. The picture is somewhat confused by their discovery that 4 moles of acetate are produced for every mole of lactate provided – a peculiar discrepancy which was not addressed (in a previous experiment [Murphy & Condon, 1984b] using glucose, 1.3 moles of acetate per 1 mole glucose was produced – a result more in line with expectations).

Mechanism for the aerobic production of acetate

There are three mechanisms whereby acetate may be produced from glucose in homofermentative lactic acid bacteria: by the direct oxidation of lactate; by the oxidation of pyruvate via the intermediary of acetyl CoA; and by the direct oxidation of pyruvate. These pathways are shown in Figure 1.5.

Figure 1.5. The production of acetate by homofermentative lactic acid bacteria. Under aerobic conditions, it is the central pathway (via pyruvate oxidase and acetate kinase) that is used by Lactobacillus plantarum.

i) Oxidation of lactate

The direct oxidation of lactate depends upon a functional lactate oxidase (LO), and Kandler (1983) reported that L. plantarum lacks this enzyme. Cell free extracts can oxidase lactate, but according to Murphy et al. (1985) this occurs indirectly as a result of its reduction to pyruvate by LDH, which is subsequently oxidised by pyruvate oxidase. As evidence for this, they found that lactate oxidation could be prevented by dialysis, but that it could be restored by that addition of DCPIP (which can be used by LDH as a substitute for NAD⁺). But the hypothesis of Murphy et al. cannot account for the observations of Götz et al. (1980), in which lactate oxidation was observed in the absence of peroxide formation – both pyruvate oxidase and lactate oxidase produce H₂O₂. The mechanism behind the observations of Götz et al. remain unexplained.

ii) Oxidation of pyruvate via acetyl CoA.

Acetyl CoA can be phosphorylated, by phosphate-acetyl transferase, to acetyl phosphate. It is then dephosphorylated by acetate kinase, with the concomitant synthesis of ATP. This is a common mechanism for the formation of ATP in a diverse range of lactic acid bacteria, but it is unlikely that this pathway is responsible for the aerobic formation of acetate in L. plantarum, for reasons which are discussed below.

There are two mechanisms for the synthesis of acetyl CoA from pyruvate. The first, catalysed by pyruvate-formate lyase, has been found in some strains of L. plantarum (Lindgren et al., 1990), but not in others (Tseng & Montville, 1990). Even when it occurs, this enzyme is inhibited by oxygen, and so it cannot be responsible for the formation of acetate in aerobic conditions.

The second mechanism, catalysed by pyruvate dehydrogenase, is activated by aerobic conditions (Condon, 1987). Pyruvate dehydrogenase has never been directly demonstrated in L. plantarum (Dirar & Collins, 1973, Tseng & Montville, 1990), but Kennes et al. (1991) provide some circumstantial evidence for its existence. They found that acetate and CO₂ were the sole major end products from anaerobic citrate fermentation by two strains of L. plantarum. Since the production of acetate and CO₂ as sole metabolites is characteristic of pyruvate dehydrogenase, this suggests that at least some strains must possess this enzyme. However, pyruvate dehydrogenase is only active aerobically, and it also requires an external electron acceptor to regenerate NAD⁺. Normally, this occurs by the reduction of pyruvate, acetyl CoA, or molecular oxygen to form lactate, ethanol or hydrogen peroxide, respectively. Their failure to detect either lactate or ethanol must raise doubts about the mechanism involved, but heterofermentative metabolism is an obvious possiblity.

iii) Direct oxidation of pyruvate

In the presence of oxygen, pyruvate oxidation to acetyl phosphate can occur directly. As with indirect oxidation, the acetyl phosphate so formed can be dephosphorylated by acetate kinase. This direct oxidation has been demonstrated in L. plantarum on several occasions (Dirar & Collins, 1973, Götz et al., 1980, Murphy & Condon, 1984a, 1984b, Tseng & Montville, 1992, Frey & Hubert, 1993). The enzyme responsible, pyruvate oxidase, has been investigated on both a physiological (Sedewitz et al., 1984a, 1984b) and a molecular basis (Miller & Shulz, 1993).

Of the three possible mechanisms of aerobic acetate formation, only the direct oxidation of pyruvate has been adequately demonstrated in L. plantarum. Where lactate oxidation has been observed, this must occur through an indirect route, involving as a first step the conversion of lactate to pyruvate by NAD⁺-dependent lactate dehydrogenase (nLDH). This requires an adequate supply of NAD⁺, and in aerobic conditions this can be provided by the action of NADH oxidase and NADH peroxidase, both of which are found in L. plantarum (Götz et al., 1980).

Regulation of the pyruvate oxidase pathway

The pathway leading to the aerobic formation of acetate is sensitive to oxygen (Tseng & Montville, 1992), pH (McFall & Montville, 1989, Sedewitz et al., 1984a, 1984b) and sugar, but, for the purposes of this study, it is the nature of sugar repression which is of most interest.

Glucose will repress acetate formation, but the pathway is unaffected by either galactose or fructose. Murphy et al. (1985) has shown that the oxygen-dependent metabolism of lactate to acetate is suppressed by glucose. When L. plantarum is grown aerobically on galactose, acetate forms 93% of the end products (Dirar & Collins, 1973), but when glucose is the energy source, acetate forms only 5% of the end products (Yousten et al., 1975). Lactose has, as might be expected, an intermediate effect, with acetate forming around 50% of the acid end products when L. plantarum grows aerobically on it (Sedewitz et al., 1984b). Fructose will actually stimulate acetate production when L. plantarum is grown on a mixed carbon source, overcoming glucose repression (Martínez-Anaya et al., 1994).

Which of the enzymes are subject to glucose repression is not known with certainty, and Murphy et al. (1985) has suggested that all three of the enzymes required for this pathway, NADH oxidase, pyruvate oxidase and acetate kinase, are repressed by glucose. However the consumption of oxygen by L. plantarum is not suppressed by glucose (Götz et al., 1980), which suggests that at least either NADH oxidase or pyruvate oxidase (and, possibly, NADH peroxidase) are outside its control. Pyruvate oxidase activity has been found by Sedewitz et al. (1984b) to be sharply inhibited by the presence of glucose. Their work showed that NADH oxidase activity is unaffected by glucose but, in contrast to the work of Tseng & Montville (1992) and Murphy & Condon (1984a), they also found that oxygen had no effect on NADH oxidase activity. Some caution must therefore be used in interpreting these results.

The mechanism of glucose repression in lactobacilli has been the subject of several recent reviews (Saier et al. 1995, Saier et al., 1996, Saier & Ramseier, 1996). The key protein, HPr, is activated by serine phosphorylation (Figure 1.6). In its active form it will inhibit the sugar-transporting phosphotransferase system, allosterically activate proteins such as sugar-P phosphatase and, in combination with transcription factor CcpA, it will inhibit transcription of many genes involved in carbohydrate metabolism. The kinase which mediates HPr phosphorylation is activated by key metabolites such as FBP and gluconate 6-phosphate.

Figure 1.6. HPr dependent catabolite repression. HPr kinase is activated by metabolites such as FBP, resulting in the phosphorylation of HPr on Ser46. HPr(Ser-P) has wide ranging effects, including the inhibition of PTS mediated sugar transport, allosteric activation of enzymes or, as shown here, the formation of a ternary complex with CcpA leading to the inhibition of transcription. Adapted from Saier et al. (1995).

Aims of the study

The introduction of inoculants has greatly facilitated the manufacture of well-preserved, high-quality silage, and a large inoculant industry has developed to supply this market. Unfortunately, this high-quality silage is particularly prone to aerobic spoilage, and the modern techniques of HDM (high dry-matter) and big-bale silage exacerbate this problem. In addition, HDM silage is often made in the belief that inoculation with additional LAB is not required, since the higher sugar content will stimulate the epiphytic LAB to sufficient levels. If a trend towards HDM silage manufacture is realised, the makers of silage inoculants will be confronted with a shrinking market.

Therefore the development of a novel inoculant with the ability to stabilise aerobically exposed silage will appeal to two markets. Firstly, those farmers using inoculants and LDM silage, for whom the quandary of higher quality silage being associated with increased instability will be resolved. Secondly, those farmers making HDM silage and for whom the benefits of an inoculant no longer outweigh the costs. These farmers may be persuaded to continue using an inoculant if it was also able to enhance aerobic stability.

The aim of this study, then, was firstly to characterise the need for an anti-spoilage inoculant. This entailed a survey of potential customers, identifying their silage making practice, and also their beliefs regarding aerobic spoilage. Once this was done, a screening program to select a organism suitable for an anti-spoilage inoculant was to be undertaken. This organism was then to be tested in laboratory scale silos, and to be further characterised in regard to its mode of action.

A second approach to solving the problem of aerobic spoilage was also investigated. This concerned the production, by L. plantarum, of acetic acid under aerobic conditions. This ability may desirable since, although acetic acid is not beneficial within the bulk of the silo, it has antifungal properties and so may help to prevent spoilage at the aerobic silo face. Acetic acid production is, in most silage, repressed by residual hexoses, and so the creation of a non-repressed mutant of L. plantarum was attempted.

A note on the structure of this thesis

The studies made for this thesis are presented as discrete chapters which encapsulate a specific stage or goal of the overall objective. Each chapter contains an introduction, and is then broken down into descriptions of individual experiments containing results and, where these results raise issues of immediate pertinence, a brief discussion. In each chapter, an overview of the methods used is given although, in each case, a detailed exposition of materials and methods used is presented in the first two chapters. Finally, in Chapter 10, an overview of the study is made, where results are summarised, connected, and discussed in a global context, and where the final conclusions may be found.