Vol. 172, No. 5
JOURNAL OF BACTERIOLOGY, May 1990, p. 2521-2526 0021-9193/90/052521-06$02.00/0 Copyright X) 1990, American Society for Microbiology
Three Separate Classes of Bacterial Ice Nucleation Structures M. A. TURNER, F. ARELLANO, AND L. M. KOZLOFF*
Department of Microbiology, University of California, San Francisco, California 94143-0404 Received 19 October 1989/Accepted 16 January 1990
Studies of the properties of the ice nucleation structure exposed on the surfaces of various bacteria such as Pseudomonas syringae, Erwinia herbicola, or various strains of Ice' recombinant Escherichia coli have shown that there are clearly three major related but chemically distinct types of structures on these cells. First, the ability of Ice' cells to nucleate super-cooled D20 has been examined, and it has been found that this ability (relative to the ability of the same cells to nucleate super-cooled H20) exhibited three characteristic nucleating patterns. The rarest structure, called class A, is found on only a small fraction of cells in a culture, nucleates H20 at temperatures above -4.4°C, and is an effective nucleator of super-cooled D20. A second class of structure, called class B, is found on a larger portion of the cells, nucleates H20 between -4.8 and -5.7°C, and is a relatively poor nucleator of super-cooled D20. The class C structure is found on almost all cells and nucleates at -7.6°C or colder. These three classes of structures were also differentiated by their sensitivities to low concentrations of water-miscible organic solvents such as dioxane or dimethyl sulfoxide. Depending on the specific bacterial strain, the addition of these solvents to bacterial suspensions lowered the nucleation activity of the class A structure by 1,000-fold or more. The nucleation activities of class B structures in the same culture were highly resistant to these compounds and were lowered only by 20 to 40%. The class C structures were more sensitive than Class B structures were, and the nucleation activities decreased 70 to 90%. Finally, the pH sensitivity of these three classes of structures was examined. The class A structure was destroyed in buffers at pH 4.5 or lower but was stable in buffers at higher pHs. The class B structure was less sensitive to acidic buffers but was destroyed at pH 5.5 or lower and was stable at higher pHs. However, the class C structure was unaffected by incubation in buffers with pHs of 3.5 to 9.0. Suggestions for the actual nucleation structures of the three classes are proposed.
When bacterial cultures which can nucleate super-cooled H20 are examined for their ice nucleating activities at various temperatures, it has been invariably noticed that only an extremely small fraction of cells in the cultures have ice nucleation activities at temperatures above -5C. At temperatures below -10°C, essentially all cells exhibit nucleation activities. A plot of the number of freezing nucleus units per cell versus temperature gives what has been called a cumulative nucleus spectrum (9). After examining cumulative nucleus spectra, several laboratories (7, 12) have proposed that, for convenience, it could be inferred that there were three types of nucleating activity on the bacterial surfaces. Cells active at -5°C or warmer, -5 to -8°C, or -10°C were arbitrarily assigned type I, type II, or type III activity, respectively. Since these spectra did not exhibit any inflection points or irregularities, it was not apparent what determined the nucleation activity and whether there was a chemical or quantitative difference or both a qualitative and quantitative difference in the nature of the nucleating materials on the cell surface at the different temperatures. One major hypothesis was that cell nucleating activity was proportional to the size of the nucleating structure and that different cells expressed different amounts of the ice gene product. It is known that the ice nucleation gene codes for a protein of 120 kilodaltons (3, 4). Govindarajan and Lindow (2) have shown that the target size of the ice nucleation structure was roughly proportional to the ability of cells to nucleate at the warmest temperature and that nucleation at -3°C required a structure with an apparent size of 700 kilodaltons. Similarly, Southworth et al. (8) reported a nonlinear but positive relationship between the concentra*
tion of the ice nucleation gene product and the activity of a bacterial culture. Both these studies offered no evidence of any chemical heterogeneity between structures active at various temperatures but implied that there were larger aggregates of interacting ice nucleation proteins at the warmest temperatures. No attempt was made to distinguish the structures chemically, and indeed this was difficult because the most active nucleation structures were the rarest. However, the possibility that heterogeneity existed was still expressed clearly by Wolber and Warren (10). In this paper we will present three independent lines of evidence which demonstrate that the nucleation structures are chemically heterogeneous and that three classes of nucleating structure can be precisely defined. We have preferred to describe these structures as classes A, B, and C instead of using the older terminology of types I, II, and III. Although there is considerable correspondence, for example, between type I and class A structures, the ability to specify precise temperature limits for those nucleating structures and even to define cells exhibiting transition activities between classes argued for a new but easily understood terminology. It should also be noted that the common practice of using a freezing bath at -5°C to define type I structures involves the examination of bacterial cultures which have properties midway between those of class A and class B structures. Finally, the use of the term class does not argue against the quantitative relationship postulated by others (2, 8). This paper does not chemically identify the nucleating structures of classes A, B, and C but does show that these classes can be differentiated by their properties. The major properties examined include a comparison of the nucleation of super-cooled D20 versus super-cooled H20, varying
Corresponding author. 2521
2522
J. BACTERIOL.
TURNER ET AL.
sensitivities to low concentrations of organic solvents, and sensitivities to mildly acidic buffers. (A brief summary of some D20 experiments was presented at the Federation of American Societies for Experimental Biology meeting, San Francisco, Calif., February 1989, and at the 4th International Conference on Ice Nucleation, Saskatoon, Saskatchewan, Canada, June 1989.) MATERIALS AND METHODS The ice-nucleating bacteria used were Pseudomonas syringae C-9 (5) and Ice' Escherichia coli strains containing plasmids with various amounts of P. syringae DNA. E. coli C9la was constructed in this laboratory and contained 15 kilobases of P. syringae C9 DNA (4). E. coli AGS335 obtained from G. Warren of DNA Plant Technology contains a plasmid with 3.4 kilobases of P. syringae DNA (3, 11). In addition, Erwinia herbicola (7) and Ice' E. coli rec 151 containing a plasmid with 8 to 9 kilobases of Erwinia herbicola DNA (from S. Yankofsky [12]) were also examined. Finally, the strain Snomax, a commercial preparation of P. syringae used for artificial snow making, was a gift from Richard LeDuca of Eastman Kodak Co. Cumulative nucleus spectra were determined, using a drop freezing nucleus spectrometer, and the freezing nucleus units were calculated either per viable cell or for comparison per unit volume (9). Bacterial cultures were grown on various media as described below. Other reagents were of analytical grade, and the D20 was obtained from Sigma Chemical Co.
RESULTS Bacterial nucleation of super-cooled D20. The ability of Ice' cells to nucleate super-cooled D20 was examined because the chemical properties of D20 liquid and solid are similar but not identical to H20 liquid and ice. For example, given that 100% D20 melts at 3.8°C, it might be expected that bacterial ice nucleation for varying concentrations of cells per milliliter would occur at a fixed temperature that is higher for super-cooled D20 than for super-cooled H20. The cumulative nucleus spectra in these two media for P. syringae is shown in Fig. 1. Bacterial cells, either washed off a nutrient agar slant or grown in a defined glycerol salts-amino acid medium, were divided into two equal portions. The cells were centrifuged at 10,000 x g for 10 min at 5°C, and the cell pellets were suspended in the same volume of either ordinary Tris buffer (pH 7.5) or Tris buffer (pH 7.5) in 90% D20. The 90% D20-Tris buffer was used instead of 100% D20 because the cell pellet already contained water which could not be removed. According to R. E. Feeney, 90%o D20 plus 10% H20 melts at 3.4°C (personal communication). These suspensions containing 2 x 109 cells per ml were incubated at 23°C for 1 h, and the ice nucleation activity was determined. Ice' strains exhibited typical spectra in which nucleation began at about -3°C for H20 and 0°C for 90% D20. Relatively few of the 2 x 109 bacterial cells per ml acted as nucleating centers at these warm temperatures, but at -8 to -10°C, essentially every cell acted as a nucleating center. It can be seen, however, in Fig. 1 that the freezing spectra in H20 and D20 for P. syringae C-9 do not give curves with fixed differences. A difference freezing spectrum was calculated by subtracting the freezing temperature for cells in H20 from the freezing temperatures of the cells in D20 at the same number of freezing nucleus units per milliliter. When this difference was plotted against the number of freezing
z
t
g
11
H20
106
0
-2.0
-4.0
-6.0
-8.0
-10.0
Temperature FIG. 1. Freezing spectra for suspensions of P. syringae C-9 in super-cooled H20 and D20. The procedure for these measurements is given in the text. FNU/ml, the number of freezing nucleus units per milliliter of the suspension.
nucleus units per milliliter for H20, an unexpected bowlshaped curve was found (Fig. 1 and 2a). At warm temperatures above -4°C, the difference was more than 3°C; at temperatures below -5°C, the difference was less than 1°C; and below -8°C, the difference was again 3°C. Almost identical bowl-shaped curves were obtained with cells from agar slants or from rich or minimal liquid media. With some exceptions, most of the Ice' bacteria strains gave quite similar results (Fig. 2a to e and 3a and b). We postulate that these difference spectra indicate that there are three classes (A, B, and C) of nucleating structures on the bacterial cell surfaces. Since it is the D20 cumulative freezing spectrum which apparently varies, we believe this indicates that the three classes of structures differ markedly in their abilities to nucleate super-cooled D20. These data are summarized in Table 1 in which, for each bacterial strain, the temperature limits for each class of nucleating structure are given. In the transition areas between different classes, it appears that the cell population is mixed and possibly contains different cells (each with different freezing structures) or individual cells with mixed freezing structures. Finally, given that every cell expresses class C nucleation activity, these data suggest that class C structures are converted to class B and then to class A structures. This offers a possible explanation for the properties of Snomax
2523
CLASSES OF BACTERIAL ICE NUCLEATION STRUCTURES
VOL. 172, 1990
TABLE 1. Temperature limits for each class of nucleating structure for various bacterial strains Temp limit (°C) for nucleating structure class: B
Strain A
0 I
.cm 0
a)
Pseudomonas syringae C-9 Pseudomonas syringae Snomax Ice' Escherichia coli C9la Ice' Escherichia coli AGS335 Erwinia herbicola Ice' Escherichia coli rec 151
-8.5
-5.6 >-7.6
+
0.9
E
C?Nm 0
o
U
30
t -
a
Ice+ E. coli
strain AGS 335
9
(d) 30 | /
s
aLE
Ice+Erwinia herbicola
>
$ 0) 10
nucleation structures. Since
U-
40
and Ice' E. coli rec 151 containing Erwinia herbicola DNA. P. syringae Snomax is an extremely efficient ice nucleator, and most of the class C structures are apparently converted to class B and A structures. This is shown in Fig. 3a in which the difference spectra are plotted as shown in Fig. 2; in Fig. 3b, the difference is also plotted versus the freezing temperature of H20. On the other hand, E. coli rec 151 is a poor nucleator (Fig. 2e and Table 1) and appears to form the class A structure from the class B structure only poorly. Effect of dilute organic solvents on the three classes of
(e)
30° 9
/
20
Ice+ E. coli strain rec 151 10
104
i07
lo'
1o5
108
10'
FNU/ml
FIG. 2. Freezing difference spectra in D20 versus H20 at the same number of freezing nucleus units (FNU) per milliliter for five different bacteria (a through d). The difference spectra was calculated from spectra such as those shown in Fig. 1 for each bacterial suspension.
D20 2
is known to react more
strongly than H20 does with the hydrophobic domains of proteins (1), it seemed possible that class A and C structures were much more hydrophobic than were class B structures. We examined the effect of dioxane and dimethyl sulfoxide (both water miscible), which would be expected to react with the hydrophobic domains of proteins on these nucleating structures. The concentrations of solvents which had the least effect on class B structures were selected. This was 3 to 5% (vol/vol) both for dioxane and for dimethyl sulfoxide. to the Theconcentration, to buffered cell were solvent was added in and serial final dilutions suspensions performed the same 0.05 M Tris (pH 7.0) buffer containing the same concentration of the solvent. The results are shown in Fig. 4. Class A nucleating structures were largely and ireversibly destroyed by exposure to both solvents. No classrrA activity was regained upon removal of the solvent. Class B structures were resistant to these organic compounds, and class C structures, depending on the strain, were sensitive (but
0 3~~~~~~~~~~~~~~~~ I
N~~~~~~~~~
0
x~~~~~~~~~~~~~~~~~~ 0 g.~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
co
2t
107
o10
109 0 FNU/ml
10)0
10'
~ ~~~~~0
-4
-5
-6
H20 Freezing Temperature FIG. 3. Freezing difference spectra in D20 versus H20 for P. syringae Snomax. (a) The difference in freezing temperatures at the same number of freezing nucleus units (FNU) per milliliter; (b) the difference in freezing temperatures, D20-H20, at the same nuclear freezing units per milliliter as that of the freezing temperature in H20.
2524
TURNER ET AL.
J. BACTERIOL.
) 5% Doxane 3% Doxare -0-5% Dioxane _ 1e --5% DMSO 5% DMSO v / ,5% DMSO > I I I I...l l 1I I, 1 .0
w
0~~~~~~~~~~~~
cc
102~~~~~~~~~~~~~~~~~~~~~~~~~~
E. col,
E. herbicola
°0-5% Dioxane _ -&5% DMSO I
-4.0
I
-6.0
l
l
-8.0
l
l -10.0 -4.0
l
/
-6.0
rain rec 151
P. syringae, strain Snomax
3% Dioxane -0_ 4% Dioxane *3% DMSO --4% DMSO 1 -10.0 -4.0 -6.0 -8.0 -8.0 -10.0
TEMPERATURE
FIG. 4. Effect of dioxane and dimethyl sulfoxide on ice nucleation by various bacteria. The relative ice nucleation activity (INA) is the ratio of the ice nucleation activity in freezing nucleus units (FNU) per milliliter at a given temperature in the presence of organic compound to the ice nucleation activity at the same temperature in the absence of the organic compound. DMSO, Dimethyl sulfoxide.
considerably less so than the class A structures were). These observations support the view that bacterial nucleating structures are chemically distinct and support the conclusion from the D20 experiments that class A and C structures are more hydrophobic than class B structures are. This was true for all strains, including Ice' E. coli rec 151, which has only a small amount of class A structures, and for P. syringae Snomax, in which most of the cells have the class A and B structures but few have only the class C structure (Fig. 3). The pH activity profiles of the three classes of ice nucleation structures. Earlier studies of the effect of various pHs on bacterial ice nucleation activity were carried out by an insensitive method of measuring ice nucleation (5). In this previous work, only the temperature at which 50% of the cells were nucleation active was determined, and this crude measurement obscured any of the effects of any different classes of nucleation structures. In the current experiments, a bacterial suspension in water was diluted into 0.1 M acetate buffers (pH 3.5, 4.0, 4.5, and 5.0), 0.1 M phosphate buffers (pH 5.5, 6.0, 6.5, 7.0, or 7.5), or 0.1 M Tris buffers (pH 8.0, 8.5, 9.0, and 9.5). Serial dilutions were carried out in the same buffer. The nucleation activity at pH 6.0 to 6.5 for all classes was found to be the highest for all strains and was considered the maximal value for all strains. Class A structures were found to be highly sensitive to low pH buffers, while class B and C structures were more resistant. These results are shown in Fig. 5. The relative decrease in ice nucleation activity as a function of pH is shown for class A, B, and C structures measured at -4, -6, and -9°C, respectively. These results support the conclusion that the
three classes defined by the D20 experiments have distinct chemical properties. Class A structures were highly sensitive at pH 5.5 or below, and the loss of activity could not be regained by raising the pH. Class B structures were stable down to about pH 4.5, and below this pH the structure was also irreversibly inactivated. Both class A and B structures were largely stable up to pH 9.5. Class C structures were almost completely stable from pH 3.5 to 9.5 DISCUSSION The ice nucleating structures are largely localized in the bacterial outer membrane (1), and we believe that the nucleating structures all contain the ice nucleating protein to which additional elements are added during posttranslational modification. Furthermore, these structures may be aggregated, a situation which increases the nucleation activity. The D20-H20 difference freezing spectra would not necessarily be sensitive to differences in aggregation. While a number of differences between H20 and D20 are known, perhaps most pertinent for this system is the relative increased ability of D20 to interact with the hydrophobic domains of proteins (1). The class A structure, which reacts well with D20, is apparently more hydrophobic than the class B structure is. The class B structure only poorly nucleates super-cooled D20, the difference in the D20-H20 spectra is quite small, and the class B structure appears to be much more hydrophilic than class A is. On the other hand, class C structures can again effectively nucleate D20 and appear to be more hydrophobic than class B structures are.
CLASSES OF BACTERIAL ICE NUCLEATION STRUCTURES
VOL. 172, 1990
2525
z
>wi a:
10'
10-6
X.
1.0
,I
-A
10-20 Ice'
E. herbicola
P. syringae, strain
E. Coll, strain rec 151
~~~~~~~~~~~~~~Snomax
I
:class A (40C) I
_ /
10
r
4 0
|
-o
classB(-60C)
-/,,. class C (-90C)
pH FIG. 5. Sensitivity of the three different classes of ice nucleation structures on various bacteria to buffers at various pHs. The relative ice nucleation activity (INA) is the ratio of the activity at a given pH to the activity at pH 6.5. Classes A, B, and C were measured at -4, -6, and -9°C, respectively.
While the difference spectra of the nucleation of supercooled D20 versus the nucleation of super-cooled H20 does not distinguish class A from class C structures, the quite different properties of class B nucleation structures, the organic solvent sensitivity, and the low pH sensitivity support the conclusion that the three proposed structures are quite different. Finally, one alternative explanation for the varying properties of the three classes is that the structures themselves are not different but their individual environments differ. For example, it is possible (but unlikely) that the class B structure, which is resistant to organic solvents and which reacts poorly with D20, is somehow shielded. All the characteristic properties differentiating the three classes argue against this possibility, but further characterization of these structures is clearly necessary. We suggest that the class C structure is the protein product of the inaZ gene and is expressed on most cells. This protein, based on translation of the gene, should have a molecular mass of 120 kilodaltons (3, 9). It has been purified, and it is generally agreed that the protein provides the critical template necessary for the formation of ice crystals (9). However, the purified protein initiates ice nucleation only at around -6°C, where class C-initiated nucleation begins. Furthermore, the apparent molecular mass of the isolated protein as observed by sodium dodecyl sulfatepolyacrylamide gel electrophoresis is 150kd. According to Wolber et al. (11), this apparent discrepancy may be caused by any of several factors, including posttranslational modification by covalent attachment of lipid or carbohydrates. On
the basis of the triphasic D20-H20 freezing spectra, organic solvent sensitivity, and pH sensitivity, we believe that posttranslational modification is most likely and accounts for the observed properties. We propose, in accordance with the discussion above, that the class C structure is modified to produce the class B structure and that this in turn is converted into the class A structure. The three structures must all contain the inaZ gene product, and on the basis of the relative frequencies of the three structures plus the transitional activities, it appears that class C -- class B class A in a regular reaction sequence. The chemical nature of the class A structure must account for its hydrophobicity, including its sensitivity to organic solvents. We have previously proposed (4) that inositol as phosphatidylinositol, a phospholipid found in these Ice' bacteria but not' in Ice- bacteria, plays a role in ice nucleation. Of special importance is the observation that phospholipase CII, a lipase which specifically cleaves the inositol-phosphate bond in proteins anchored to cell membranes (6), reduces the nucleation activity of most Ice' bacteria (4, 6). Although not appreciated at that time, these earlier results indicate the specific destruction of the class A structure by the CII lipase. While the evidence that phosphatidylinositol anchors the InaZ protein is not complete, this hypothesis is in accord with all the results found in this and earlier work (4, 5). ACKNOWLEDGMENT This work was supported by grant no. DCB-8509759 from the National Science Foundation to L. M. Kozloff.
2526
TURNER ET AL.
LITERATURE CITED 1. Ahmed, A. I., D. T. Osuga, and R. E. Feeney. 1980. Anti-freeze protein from fishes: freezing behavior in H20 and D20. Biochem. Int. 1:41-46. 2. Govindarajan, A. J., and S. E. Lindow. 1988. Size of bacterial ice-nucleation sites measured in situ by radiation inactivation analysis. Proc. Natl. Acad. Sci. USA 85:1334-1338. 3. Green, R. L., and G. J. Warren. 1985. Physical and functional repetition in a bacterial ice gene. Nature (London) 317:645-648. 4. Kozloff, L. M., M. Lute, and David Westaway. 1984. Phosphatidylinositol as a component of the ice nucleating site of Pseudomonas syringae and Erwinia herbicola. Science 226:845-846. 5. Kozloff, L. M., M. A. Schofield, and M. Lute. 1983. Ice nucleating activity of Pseudomonas syringae and Erwinia herbicola. J. Bacteriol. 153:222-231. 6. Low, Martin G. 1989. Glycosyl-phosphatidylinositol: a versatile anchor for cell surface proteins. Fed. Am. Soc. Exp. Biol. 3:1600-1608. 7. Phelps, P. A., T. H. Giddings, M. Prochoda, and R. Fall. 1986.
J. BACTERIOL.
8.
9. 10. 11.
12.
Release of cell-free nuclei by Erwinia herbicola. J. Bacteriol. 167:496-502. Southworth, M. W., P. K. Wolber, and G. J. Warren. 1988. Nonlinear relationship between concentration and activity of a bacterial ice nucleation protein. J. Biol. Chem. 263:1521115216. Valli, G., and E. J. Stansbury. 1966. Time dependent characteristics of the heterogeneous nucleation of ice. Can. J. Phys. 44:477-502. Wolber, P., and G. Warren. 1989. Bacterial ice-nucleation proteins. Trends Biochem. Sci. 14:179-182. Wolber, P. K., C. A. Deininger, M. K. Southworth, J. Vandekerckhove, M. Van Montague, and G. J. Warren. 1986. Identification and purification of a bacterial ice-nucleation protein. Proc. Natl. Acad. Sci. USA 83:7256-7260. Yankofsky, S. A., Z. Levin, T. Bertold, and N. Sandlerman. 1981. Some basic characteristics of bacterial freezing nuclei. J. Appl. Meteorol. 20:1013-1019.
JOURNAL OF BACTERIOLOGY, May 1990, p. 2521-2526 0021-9193/90/052521-06$02.00/0 Copyright X) 1990, American Society for Microbiology
Three Separate Classes of Bacterial Ice Nucleation Structures M. A. TURNER, F. ARELLANO, AND L. M. KOZLOFF*
Department of Microbiology, University of California, San Francisco, California 94143-0404 Received 19 October 1989/Accepted 16 January 1990
Studies of the properties of the ice nucleation structure exposed on the surfaces of various bacteria such as Pseudomonas syringae, Erwinia herbicola, or various strains of Ice' recombinant Escherichia coli have shown that there are clearly three major related but chemically distinct types of structures on these cells. First, the ability of Ice' cells to nucleate super-cooled D20 has been examined, and it has been found that this ability (relative to the ability of the same cells to nucleate super-cooled H20) exhibited three characteristic nucleating patterns. The rarest structure, called class A, is found on only a small fraction of cells in a culture, nucleates H20 at temperatures above -4.4°C, and is an effective nucleator of super-cooled D20. A second class of structure, called class B, is found on a larger portion of the cells, nucleates H20 between -4.8 and -5.7°C, and is a relatively poor nucleator of super-cooled D20. The class C structure is found on almost all cells and nucleates at -7.6°C or colder. These three classes of structures were also differentiated by their sensitivities to low concentrations of water-miscible organic solvents such as dioxane or dimethyl sulfoxide. Depending on the specific bacterial strain, the addition of these solvents to bacterial suspensions lowered the nucleation activity of the class A structure by 1,000-fold or more. The nucleation activities of class B structures in the same culture were highly resistant to these compounds and were lowered only by 20 to 40%. The class C structures were more sensitive than Class B structures were, and the nucleation activities decreased 70 to 90%. Finally, the pH sensitivity of these three classes of structures was examined. The class A structure was destroyed in buffers at pH 4.5 or lower but was stable in buffers at higher pHs. The class B structure was less sensitive to acidic buffers but was destroyed at pH 5.5 or lower and was stable at higher pHs. However, the class C structure was unaffected by incubation in buffers with pHs of 3.5 to 9.0. Suggestions for the actual nucleation structures of the three classes are proposed.
When bacterial cultures which can nucleate super-cooled H20 are examined for their ice nucleating activities at various temperatures, it has been invariably noticed that only an extremely small fraction of cells in the cultures have ice nucleation activities at temperatures above -5C. At temperatures below -10°C, essentially all cells exhibit nucleation activities. A plot of the number of freezing nucleus units per cell versus temperature gives what has been called a cumulative nucleus spectrum (9). After examining cumulative nucleus spectra, several laboratories (7, 12) have proposed that, for convenience, it could be inferred that there were three types of nucleating activity on the bacterial surfaces. Cells active at -5°C or warmer, -5 to -8°C, or -10°C were arbitrarily assigned type I, type II, or type III activity, respectively. Since these spectra did not exhibit any inflection points or irregularities, it was not apparent what determined the nucleation activity and whether there was a chemical or quantitative difference or both a qualitative and quantitative difference in the nature of the nucleating materials on the cell surface at the different temperatures. One major hypothesis was that cell nucleating activity was proportional to the size of the nucleating structure and that different cells expressed different amounts of the ice gene product. It is known that the ice nucleation gene codes for a protein of 120 kilodaltons (3, 4). Govindarajan and Lindow (2) have shown that the target size of the ice nucleation structure was roughly proportional to the ability of cells to nucleate at the warmest temperature and that nucleation at -3°C required a structure with an apparent size of 700 kilodaltons. Similarly, Southworth et al. (8) reported a nonlinear but positive relationship between the concentra*
tion of the ice nucleation gene product and the activity of a bacterial culture. Both these studies offered no evidence of any chemical heterogeneity between structures active at various temperatures but implied that there were larger aggregates of interacting ice nucleation proteins at the warmest temperatures. No attempt was made to distinguish the structures chemically, and indeed this was difficult because the most active nucleation structures were the rarest. However, the possibility that heterogeneity existed was still expressed clearly by Wolber and Warren (10). In this paper we will present three independent lines of evidence which demonstrate that the nucleation structures are chemically heterogeneous and that three classes of nucleating structure can be precisely defined. We have preferred to describe these structures as classes A, B, and C instead of using the older terminology of types I, II, and III. Although there is considerable correspondence, for example, between type I and class A structures, the ability to specify precise temperature limits for those nucleating structures and even to define cells exhibiting transition activities between classes argued for a new but easily understood terminology. It should also be noted that the common practice of using a freezing bath at -5°C to define type I structures involves the examination of bacterial cultures which have properties midway between those of class A and class B structures. Finally, the use of the term class does not argue against the quantitative relationship postulated by others (2, 8). This paper does not chemically identify the nucleating structures of classes A, B, and C but does show that these classes can be differentiated by their properties. The major properties examined include a comparison of the nucleation of super-cooled D20 versus super-cooled H20, varying
Corresponding author. 2521
2522
J. BACTERIOL.
TURNER ET AL.
sensitivities to low concentrations of organic solvents, and sensitivities to mildly acidic buffers. (A brief summary of some D20 experiments was presented at the Federation of American Societies for Experimental Biology meeting, San Francisco, Calif., February 1989, and at the 4th International Conference on Ice Nucleation, Saskatoon, Saskatchewan, Canada, June 1989.) MATERIALS AND METHODS The ice-nucleating bacteria used were Pseudomonas syringae C-9 (5) and Ice' Escherichia coli strains containing plasmids with various amounts of P. syringae DNA. E. coli C9la was constructed in this laboratory and contained 15 kilobases of P. syringae C9 DNA (4). E. coli AGS335 obtained from G. Warren of DNA Plant Technology contains a plasmid with 3.4 kilobases of P. syringae DNA (3, 11). In addition, Erwinia herbicola (7) and Ice' E. coli rec 151 containing a plasmid with 8 to 9 kilobases of Erwinia herbicola DNA (from S. Yankofsky [12]) were also examined. Finally, the strain Snomax, a commercial preparation of P. syringae used for artificial snow making, was a gift from Richard LeDuca of Eastman Kodak Co. Cumulative nucleus spectra were determined, using a drop freezing nucleus spectrometer, and the freezing nucleus units were calculated either per viable cell or for comparison per unit volume (9). Bacterial cultures were grown on various media as described below. Other reagents were of analytical grade, and the D20 was obtained from Sigma Chemical Co.
RESULTS Bacterial nucleation of super-cooled D20. The ability of Ice' cells to nucleate super-cooled D20 was examined because the chemical properties of D20 liquid and solid are similar but not identical to H20 liquid and ice. For example, given that 100% D20 melts at 3.8°C, it might be expected that bacterial ice nucleation for varying concentrations of cells per milliliter would occur at a fixed temperature that is higher for super-cooled D20 than for super-cooled H20. The cumulative nucleus spectra in these two media for P. syringae is shown in Fig. 1. Bacterial cells, either washed off a nutrient agar slant or grown in a defined glycerol salts-amino acid medium, were divided into two equal portions. The cells were centrifuged at 10,000 x g for 10 min at 5°C, and the cell pellets were suspended in the same volume of either ordinary Tris buffer (pH 7.5) or Tris buffer (pH 7.5) in 90% D20. The 90% D20-Tris buffer was used instead of 100% D20 because the cell pellet already contained water which could not be removed. According to R. E. Feeney, 90%o D20 plus 10% H20 melts at 3.4°C (personal communication). These suspensions containing 2 x 109 cells per ml were incubated at 23°C for 1 h, and the ice nucleation activity was determined. Ice' strains exhibited typical spectra in which nucleation began at about -3°C for H20 and 0°C for 90% D20. Relatively few of the 2 x 109 bacterial cells per ml acted as nucleating centers at these warm temperatures, but at -8 to -10°C, essentially every cell acted as a nucleating center. It can be seen, however, in Fig. 1 that the freezing spectra in H20 and D20 for P. syringae C-9 do not give curves with fixed differences. A difference freezing spectrum was calculated by subtracting the freezing temperature for cells in H20 from the freezing temperatures of the cells in D20 at the same number of freezing nucleus units per milliliter. When this difference was plotted against the number of freezing
z
t
g
11
H20
106
0
-2.0
-4.0
-6.0
-8.0
-10.0
Temperature FIG. 1. Freezing spectra for suspensions of P. syringae C-9 in super-cooled H20 and D20. The procedure for these measurements is given in the text. FNU/ml, the number of freezing nucleus units per milliliter of the suspension.
nucleus units per milliliter for H20, an unexpected bowlshaped curve was found (Fig. 1 and 2a). At warm temperatures above -4°C, the difference was more than 3°C; at temperatures below -5°C, the difference was less than 1°C; and below -8°C, the difference was again 3°C. Almost identical bowl-shaped curves were obtained with cells from agar slants or from rich or minimal liquid media. With some exceptions, most of the Ice' bacteria strains gave quite similar results (Fig. 2a to e and 3a and b). We postulate that these difference spectra indicate that there are three classes (A, B, and C) of nucleating structures on the bacterial cell surfaces. Since it is the D20 cumulative freezing spectrum which apparently varies, we believe this indicates that the three classes of structures differ markedly in their abilities to nucleate super-cooled D20. These data are summarized in Table 1 in which, for each bacterial strain, the temperature limits for each class of nucleating structure are given. In the transition areas between different classes, it appears that the cell population is mixed and possibly contains different cells (each with different freezing structures) or individual cells with mixed freezing structures. Finally, given that every cell expresses class C nucleation activity, these data suggest that class C structures are converted to class B and then to class A structures. This offers a possible explanation for the properties of Snomax
2523
CLASSES OF BACTERIAL ICE NUCLEATION STRUCTURES
VOL. 172, 1990
TABLE 1. Temperature limits for each class of nucleating structure for various bacterial strains Temp limit (°C) for nucleating structure class: B
Strain A
0 I
.cm 0
a)
Pseudomonas syringae C-9 Pseudomonas syringae Snomax Ice' Escherichia coli C9la Ice' Escherichia coli AGS335 Erwinia herbicola Ice' Escherichia coli rec 151
-8.5
-5.6 >-7.6
+
0.9
E
C?Nm 0
o
U
30
t -
a
Ice+ E. coli
strain AGS 335
9
(d) 30 | /
s
aLE
Ice+Erwinia herbicola
>
$ 0) 10
nucleation structures. Since
U-
40
and Ice' E. coli rec 151 containing Erwinia herbicola DNA. P. syringae Snomax is an extremely efficient ice nucleator, and most of the class C structures are apparently converted to class B and A structures. This is shown in Fig. 3a in which the difference spectra are plotted as shown in Fig. 2; in Fig. 3b, the difference is also plotted versus the freezing temperature of H20. On the other hand, E. coli rec 151 is a poor nucleator (Fig. 2e and Table 1) and appears to form the class A structure from the class B structure only poorly. Effect of dilute organic solvents on the three classes of
(e)
30° 9
/
20
Ice+ E. coli strain rec 151 10
104
i07
lo'
1o5
108
10'
FNU/ml
FIG. 2. Freezing difference spectra in D20 versus H20 at the same number of freezing nucleus units (FNU) per milliliter for five different bacteria (a through d). The difference spectra was calculated from spectra such as those shown in Fig. 1 for each bacterial suspension.
D20 2
is known to react more
strongly than H20 does with the hydrophobic domains of proteins (1), it seemed possible that class A and C structures were much more hydrophobic than were class B structures. We examined the effect of dioxane and dimethyl sulfoxide (both water miscible), which would be expected to react with the hydrophobic domains of proteins on these nucleating structures. The concentrations of solvents which had the least effect on class B structures were selected. This was 3 to 5% (vol/vol) both for dioxane and for dimethyl sulfoxide. to the Theconcentration, to buffered cell were solvent was added in and serial final dilutions suspensions performed the same 0.05 M Tris (pH 7.0) buffer containing the same concentration of the solvent. The results are shown in Fig. 4. Class A nucleating structures were largely and ireversibly destroyed by exposure to both solvents. No classrrA activity was regained upon removal of the solvent. Class B structures were resistant to these organic compounds, and class C structures, depending on the strain, were sensitive (but
0 3~~~~~~~~~~~~~~~~ I
N~~~~~~~~~
0
x~~~~~~~~~~~~~~~~~~ 0 g.~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
co
2t
107
o10
109 0 FNU/ml
10)0
10'
~ ~~~~~0
-4
-5
-6
H20 Freezing Temperature FIG. 3. Freezing difference spectra in D20 versus H20 for P. syringae Snomax. (a) The difference in freezing temperatures at the same number of freezing nucleus units (FNU) per milliliter; (b) the difference in freezing temperatures, D20-H20, at the same nuclear freezing units per milliliter as that of the freezing temperature in H20.
2524
TURNER ET AL.
J. BACTERIOL.
) 5% Doxane 3% Doxare -0-5% Dioxane _ 1e --5% DMSO 5% DMSO v / ,5% DMSO > I I I I...l l 1I I, 1 .0
w
0~~~~~~~~~~~~
cc
102~~~~~~~~~~~~~~~~~~~~~~~~~~
E. col,
E. herbicola
°0-5% Dioxane _ -&5% DMSO I
-4.0
I
-6.0
l
l
-8.0
l
l -10.0 -4.0
l
/
-6.0
rain rec 151
P. syringae, strain Snomax
3% Dioxane -0_ 4% Dioxane *3% DMSO --4% DMSO 1 -10.0 -4.0 -6.0 -8.0 -8.0 -10.0
TEMPERATURE
FIG. 4. Effect of dioxane and dimethyl sulfoxide on ice nucleation by various bacteria. The relative ice nucleation activity (INA) is the ratio of the ice nucleation activity in freezing nucleus units (FNU) per milliliter at a given temperature in the presence of organic compound to the ice nucleation activity at the same temperature in the absence of the organic compound. DMSO, Dimethyl sulfoxide.
considerably less so than the class A structures were). These observations support the view that bacterial nucleating structures are chemically distinct and support the conclusion from the D20 experiments that class A and C structures are more hydrophobic than class B structures are. This was true for all strains, including Ice' E. coli rec 151, which has only a small amount of class A structures, and for P. syringae Snomax, in which most of the cells have the class A and B structures but few have only the class C structure (Fig. 3). The pH activity profiles of the three classes of ice nucleation structures. Earlier studies of the effect of various pHs on bacterial ice nucleation activity were carried out by an insensitive method of measuring ice nucleation (5). In this previous work, only the temperature at which 50% of the cells were nucleation active was determined, and this crude measurement obscured any of the effects of any different classes of nucleation structures. In the current experiments, a bacterial suspension in water was diluted into 0.1 M acetate buffers (pH 3.5, 4.0, 4.5, and 5.0), 0.1 M phosphate buffers (pH 5.5, 6.0, 6.5, 7.0, or 7.5), or 0.1 M Tris buffers (pH 8.0, 8.5, 9.0, and 9.5). Serial dilutions were carried out in the same buffer. The nucleation activity at pH 6.0 to 6.5 for all classes was found to be the highest for all strains and was considered the maximal value for all strains. Class A structures were found to be highly sensitive to low pH buffers, while class B and C structures were more resistant. These results are shown in Fig. 5. The relative decrease in ice nucleation activity as a function of pH is shown for class A, B, and C structures measured at -4, -6, and -9°C, respectively. These results support the conclusion that the
three classes defined by the D20 experiments have distinct chemical properties. Class A structures were highly sensitive at pH 5.5 or below, and the loss of activity could not be regained by raising the pH. Class B structures were stable down to about pH 4.5, and below this pH the structure was also irreversibly inactivated. Both class A and B structures were largely stable up to pH 9.5. Class C structures were almost completely stable from pH 3.5 to 9.5 DISCUSSION The ice nucleating structures are largely localized in the bacterial outer membrane (1), and we believe that the nucleating structures all contain the ice nucleating protein to which additional elements are added during posttranslational modification. Furthermore, these structures may be aggregated, a situation which increases the nucleation activity. The D20-H20 difference freezing spectra would not necessarily be sensitive to differences in aggregation. While a number of differences between H20 and D20 are known, perhaps most pertinent for this system is the relative increased ability of D20 to interact with the hydrophobic domains of proteins (1). The class A structure, which reacts well with D20, is apparently more hydrophobic than the class B structure is. The class B structure only poorly nucleates super-cooled D20, the difference in the D20-H20 spectra is quite small, and the class B structure appears to be much more hydrophilic than class A is. On the other hand, class C structures can again effectively nucleate D20 and appear to be more hydrophobic than class B structures are.
CLASSES OF BACTERIAL ICE NUCLEATION STRUCTURES
VOL. 172, 1990
2525
z
>wi a:
10'
10-6
X.
1.0
,I
-A
10-20 Ice'
E. herbicola
P. syringae, strain
E. Coll, strain rec 151
~~~~~~~~~~~~~~Snomax
I
:class A (40C) I
_ /
10
r
4 0
|
-o
classB(-60C)
-/,,. class C (-90C)
pH FIG. 5. Sensitivity of the three different classes of ice nucleation structures on various bacteria to buffers at various pHs. The relative ice nucleation activity (INA) is the ratio of the activity at a given pH to the activity at pH 6.5. Classes A, B, and C were measured at -4, -6, and -9°C, respectively.
While the difference spectra of the nucleation of supercooled D20 versus the nucleation of super-cooled H20 does not distinguish class A from class C structures, the quite different properties of class B nucleation structures, the organic solvent sensitivity, and the low pH sensitivity support the conclusion that the three proposed structures are quite different. Finally, one alternative explanation for the varying properties of the three classes is that the structures themselves are not different but their individual environments differ. For example, it is possible (but unlikely) that the class B structure, which is resistant to organic solvents and which reacts poorly with D20, is somehow shielded. All the characteristic properties differentiating the three classes argue against this possibility, but further characterization of these structures is clearly necessary. We suggest that the class C structure is the protein product of the inaZ gene and is expressed on most cells. This protein, based on translation of the gene, should have a molecular mass of 120 kilodaltons (3, 9). It has been purified, and it is generally agreed that the protein provides the critical template necessary for the formation of ice crystals (9). However, the purified protein initiates ice nucleation only at around -6°C, where class C-initiated nucleation begins. Furthermore, the apparent molecular mass of the isolated protein as observed by sodium dodecyl sulfatepolyacrylamide gel electrophoresis is 150kd. According to Wolber et al. (11), this apparent discrepancy may be caused by any of several factors, including posttranslational modification by covalent attachment of lipid or carbohydrates. On
the basis of the triphasic D20-H20 freezing spectra, organic solvent sensitivity, and pH sensitivity, we believe that posttranslational modification is most likely and accounts for the observed properties. We propose, in accordance with the discussion above, that the class C structure is modified to produce the class B structure and that this in turn is converted into the class A structure. The three structures must all contain the inaZ gene product, and on the basis of the relative frequencies of the three structures plus the transitional activities, it appears that class C -- class B class A in a regular reaction sequence. The chemical nature of the class A structure must account for its hydrophobicity, including its sensitivity to organic solvents. We have previously proposed (4) that inositol as phosphatidylinositol, a phospholipid found in these Ice' bacteria but not' in Ice- bacteria, plays a role in ice nucleation. Of special importance is the observation that phospholipase CII, a lipase which specifically cleaves the inositol-phosphate bond in proteins anchored to cell membranes (6), reduces the nucleation activity of most Ice' bacteria (4, 6). Although not appreciated at that time, these earlier results indicate the specific destruction of the class A structure by the CII lipase. While the evidence that phosphatidylinositol anchors the InaZ protein is not complete, this hypothesis is in accord with all the results found in this and earlier work (4, 5). ACKNOWLEDGMENT This work was supported by grant no. DCB-8509759 from the National Science Foundation to L. M. Kozloff.
2526
TURNER ET AL.
LITERATURE CITED 1. Ahmed, A. I., D. T. Osuga, and R. E. Feeney. 1980. Anti-freeze protein from fishes: freezing behavior in H20 and D20. Biochem. Int. 1:41-46. 2. Govindarajan, A. J., and S. E. Lindow. 1988. Size of bacterial ice-nucleation sites measured in situ by radiation inactivation analysis. Proc. Natl. Acad. Sci. USA 85:1334-1338. 3. Green, R. L., and G. J. Warren. 1985. Physical and functional repetition in a bacterial ice gene. Nature (London) 317:645-648. 4. Kozloff, L. M., M. Lute, and David Westaway. 1984. Phosphatidylinositol as a component of the ice nucleating site of Pseudomonas syringae and Erwinia herbicola. Science 226:845-846. 5. Kozloff, L. M., M. A. Schofield, and M. Lute. 1983. Ice nucleating activity of Pseudomonas syringae and Erwinia herbicola. J. Bacteriol. 153:222-231. 6. Low, Martin G. 1989. Glycosyl-phosphatidylinositol: a versatile anchor for cell surface proteins. Fed. Am. Soc. Exp. Biol. 3:1600-1608. 7. Phelps, P. A., T. H. Giddings, M. Prochoda, and R. Fall. 1986.
J. BACTERIOL.
8.
9. 10. 11.
12.
Release of cell-free nuclei by Erwinia herbicola. J. Bacteriol. 167:496-502. Southworth, M. W., P. K. Wolber, and G. J. Warren. 1988. Nonlinear relationship between concentration and activity of a bacterial ice nucleation protein. J. Biol. Chem. 263:1521115216. Valli, G., and E. J. Stansbury. 1966. Time dependent characteristics of the heterogeneous nucleation of ice. Can. J. Phys. 44:477-502. Wolber, P., and G. Warren. 1989. Bacterial ice-nucleation proteins. Trends Biochem. Sci. 14:179-182. Wolber, P. K., C. A. Deininger, M. K. Southworth, J. Vandekerckhove, M. Van Montague, and G. J. Warren. 1986. Identification and purification of a bacterial ice-nucleation protein. Proc. Natl. Acad. Sci. USA 83:7256-7260. Yankofsky, S. A., Z. Levin, T. Bertold, and N. Sandlerman. 1981. Some basic characteristics of bacterial freezing nuclei. J. Appl. Meteorol. 20:1013-1019.
