INFECrION AND IMMUNITY, Nov. 1994, P. 4848-48540019-9567/94/$04.00+OCopyright C 1994, American Society for Microbiology

Protein D, the Glycerophosphodiester Phosphodiesterase fromHaemophilus influenzae with Affinity for Human

Immunoglobulin D, Influences Virulence in a Rat Otitis ModelHAKAN JANSON,' ASA MELHUS,' ANN HERMANSSON,2 AND ARNE FORSGRENl*

Department of Medical Microbiology, Lund University, University Hospital, MASS-214 01 Malmojand Departmient of Oto-rhino-laryngology, Lund University, S-220 00 Lund,2 Sweden

Received 21 March 1994/Returned for modification 7 July 1994/Accepted 19 August 1994

A mutant lacking the ability to express the surface-exposed lipoprotein protein D was constructed by linkerinsertion and deletion mutagenesis of a cloned DNA insert containing the protein D structural gene from a

nontypeable Haemophilus influenzae strain (NTHi). An isogenic NTHi mutant was isolated after transformationof genetically competent bacteria. The transformant was unreactive to a protein D-specific monoclonalantibody in a colony immunoassay. In addition, the mutant lacked the ability to synthesize detectable levels ofprotein D by protein staining, immunoblot methods, glycerophosphodiester phosphodiesterase activity, andbinding studies of radiolabelled immunoglobulin D. The isogenic protein D-deficient mutant was comparedwith its parental strain for its ability to induce experimental otitis media in rats challenged with bacteria. Anapproximately 100-times-higher concentration of the mutant compared with that of the wild-type strain was

required in order to cause otitis among all rats challenged with that given dose. The protein D mutant exhibiteda generation time that was equal to that of the wild-type strain in complex broth medium. No difference inlipopolysaccharide expression was found between the mutant and the parental strain. These results suggestthat protein D may influence the pathogenesis of NTHi in the upper respiratory tract.

Haemophilus influenzae is an important pathogen in thehuman respiratory tract and a frequent cause of severe dis-eases of childhood, including septicemia and meningitis. Themolecular basis for the pathogenesis of these invasive diseases,which most frequently are caused by H. influenzae type b (Hib),has revealed that the capsular polysaccharide of this organismis an important virulence factor (29, 47). Another importantvirulence determinant is the lipopolysaccharide (LPS) of theHib outer membrane since alterations in LPS expression havebeen shown to reduce the virulence potential of this bacterium(20, 21, 50). No protein has been shown to be essential for thevirulence of Hib, but at least a 28-kDa lipoprotein has beensuggested to aid transepithelial invasion (8).

Nontypeable H. influenzae (NTHi) is a frequent colonizer ofthe upper respiratory tract and a leading cause of human otitismedia and respiratory infections, including pneumonia andsinusitis (31, 34, 44). The pathogenesis of NTHi disease isbelieved to begin with colonization of the nasopharynx. Sub-sequently, there is a contiguous spread within the respiratorytract to the middle ear, the sinuses, the conjunctiva, or thelungs. H. influenzae expresses adhesive pili that are likely to beinvolved in colonization of the host (3, 19, 45). However,nonpiliated NTHi organisms are also capable of adhering toeukaryotic cells via high-molecular-weight nonpilus adhesins(41).

Colonization of mucosal membranes by H. influenzae mayalso involve interaction with the local mucosal immune system.H. influenzae is one of a group of pathogenic microorganismsthat produce an extracellular immunoglobulin Al (IgAl)protease that specifically cleaves the heavy chain of humanIgAl (35). Another interesting immunological interaction is

* Corresponding author. Mailing address: Department of MedicalMicrobiology, Lund University, University Hospital, MASS-214 01Malmo, Sweden. Phone: 46/40-331341. Fax: 46/40-336234. Electronicmail address: mblafn@maja.mas.malmo.se.

the IgD binding properties of H. influenzae and Moraxella(Branhamella) catarrhalis (13). A 42-kDa lipoprotein, namedprotein D, which is exposed on the surfaces of all H. influenzaestrains, was first identified as the protein responsible for thisinteraction (2, 16, 36). However, it was later discovered thatprotein D did not bind to the majority of IgD myelomas testedand that Hib strains express an additional IgD-binding factoror cofactor that is lacking in NTHi (37). Despite this finding,protein D is still an interesting molecule since it is highlyconserved among all H. influenzae strains (2, 18), a featurewhich is relatively rarely found among NTHi outer membraneproteins (4, 32, 33). A possible explanation for this conserva-tion might be that protein D has an important enzyme activityto fulfill, since it was recently shown that protein D is 67%identical to the periplasmic glycerophosphodiester phospho-diesterase of Escherichia coli (30). Extracts of E. coli expressingrecombinant protein D had increased specific activity of glyc-erophosphodiester phosphodiesterase compared with that ofextracts of E. coli not expressing protein D (30).We sought to investigate whether protein D is involved in

the pathogenesis of NTHi. An isogenic mutant lacking theability to express protein D was constructed. The mutant was

compared with the wild-type strain for its ability to causeinfection in the middle ears of rats in an otitis model originallydeveloped for pneumococcal otitis (15). This rat model was

recently modified for studies of otitis media caused by H.influenzae (27).

MATERIALS AND METHODS

Bacterial strains, plasmids, and culture conditions. E. coliJM83 (49) was used as a recipient for plasmids pUC4K(Pharmacia Biotech, Sollentuna, Sweden) and pHIC348 (17)and derivatives thereof. H. influenzae 772, a nontypeableclinical isolate of biotype II, has been described previously(16-18, 36). H. influenzae strains were grown on chocolate agar

4848

Vol. 62, No. 11

on July 21, 2019 by guesthttp://iai.asm

.org/D

ownloaded from

PROTEIN D AFFECTS VIRULENCE IN A RAT OTITIS MODEL 4849

plates, in brain heart infusion broth (Difco Lab., Inc., Detroit,Mich.) supplemented with NAD and hemin (Sigma ChemicalCo., St. Louis, Mo.) each at 10 ,ug/ml (sBHI), or on sBHIplates. E. coli JM83 was grown in LB broth (24) supplementedwhere necessary with ampicillin (100 ,ug/ml) or kanamycin (50,ug/ml). Antibiotics were purchased from Sigma. LB agar alsocontained 1.5% (wt/vol) agar. All cultures were initially inoc-ulated from stocks stored in glycerol broth at -70°C onto solidmedia and grown at 37°C. For comparison of mutant NTHiand wild-type NTHi, we used glycerol broth stocks of bacteriaisolated from the blood of mice that had received high doses ofNTHi intraperitoneally. This was done to avoid the loss ofvirulence after repeated subcultivation in vitro (26).

Antibodies. Monoclonal antibody (MAb) 19B4 is reactiveagainst protein D from all 127 H. influenzae strains tested byWestern blot (immunoblot) analysis (2). A panel of MAbsrecognizing different epitopes of H. influenzae LPS were kindlyprovided by Alf Lindberg. MAbs were used in the form of cellculture supernatants. The raising and purification of polyclonalanti-protein D immune serum will be described in detailelsewhere (1). Briefly, a nonacylated, recombinant form ofprotein D (16) was purified from periplasmic extracts of E. coliexpressing this protein. Rabbits were immunized with thepurified protein in order to raise polyclonal antiserum againstprotein D. Immune serum was passed through a column withprotein D coupled with CNBr-activated Sepharose 4B. Boundmaterials, containing anti-protein D antibodies, were elutedfrom the column after being washed, and eluted material wasdialyzed against phosphate-balanced saline (PBS).

Construction of protein D mutants. The 1.9-kbp H. influen-zae insert from pHIC348 (17) encoding protein D was excisedfrom the pUC18 vector (49) by BamHI and HindIII digestion(see Fig. 1). Bands corresponding to the vector and insert wereexcised from a preparative agarose gel and extracted withGeneclean (Bio 101, Inc., La Jolla, Calif.). The 1.9-kbp insertwas digested with HaeIII and then with NlaIV to cleave a534-bp internal HaeIII fragment in half. An 816-bp BamHI-HaeIII fragment and a 544-bp HindIII-HaeIII fragment wereexcised from an agarose gel after electrophoresis. The isolated816- and 544-bp fragments were dephosphorylated with calfintestine alkaline phosphatase (24). A corresponding amountof pUC4K was digested with HincII, and the 1.3-kbp bandcorresponding to the kanamycin resistance cassette was excisedfrom the gel and extracted with Geneclean. DNA fragmentswere ligated to the 2.7-kbp BamHI-HindIII fragment frompHIC348, i.e., the pUC18 vector, and transformed into E. coliJM83 (28). Kanamycin-resistant transformants were pickedand checked for lack of protein D expression in a colonyimmunoassay with MAb 19B4 (17, 36). Plasmids from proteinD-negative transformants were prepared with Magic Miniprep(Promega Corp., Madison, Wis.) and analyzed on 0.8% aga-rose gels after restriction endonuclease digestions. One plas-mid (pHIK3) was chosen for further experiments. All enzymeswere from Boehringer Mannheim GmbH, Mannheim, Ger-many.

Genetic transformation of H. influenzae. Cells made compe-tent by the aerobic-anaerobic incubation procedure of Good-gal and Herriot (14) as modified by Cameron (5) weretransformed with pHIK3 DNA, linearized with BamHI. Trans-formants were selected on sBHI plates containing 15 ,ug ofkanamycin per ml. Transformants were checked for a lack ofreactivity to MAb 19B4 in a colony immunoassay (17, 36).

Confirmation that the appropriate chromosomal rearrange-ment had taken place was obtained by Southern hybridizationexperiments. A 403-bp probe for the detection of proteinD-specific sequences was constructed by using oligonucleotides

5'-GTCGCCTFITFT`GTAAC-3' and 5'-GTAAAGTCGATGACATAGTA-3' (-68 to -55 and 315 to 334 bp from thetranslational start of the protein D gene [hpd]) for PCR withdigoxigenin-labelled dATP (Boehringer Mannheim) includedin the reaction mixture (10). A 32P-labelled 1.3-kbp PstIfragment corresponding to the kanamycin linker of pUC4Kwas used for the detection of DNA fragments hybridizing tothis fragment. Hybridization conditions were described previ-ously (18).

Triton X-114 extraction of bacteria. H. influenzae 772 and aselected protein D-negative mutant designated 772Ahpdl,grown to mid-logarithmic phase, were extracted with TritonX-114 as previously described (16, 42). Triton X-114-insolubleresidue was suspended in 20 mM Tris-HCl (pH 8.0)-10 mMEDTA-1% sodium dodecyl sulfate (SDS) and boiled for 5 min.Proteins were precipitated from lysates or Triton X-114 ex-tracts by the addition of 10 volumes of acetone at -20°C.LPS preparation and analysis. LPS was purified from H.

influenzae strains grown overnight on chocolate agar plates bya small-scale modification (12) of the hot phenol-water extrac-tion method (48).A whole-cell enzyme-linked immunosorbent assay (ELISA)

was used for the determination of immunological properties ofLPS from H. influenzae 772 and the protein D-deficientmutant. Bacteria grown overnight were harvested from choc-olate agar plates and suspended in PBS, pH 7.2, at an opticaldensity at 600 nm of 0.2. Flat-bottom microtiter plates (Nunc)were coated with 0.1 ml of this suspension per well at 4°Covernight. Residual protein binding sites were blocked for 2 hat room temperature with 0.1 ml of 1.5% ovalbumin in PBS perwell. Plates were washed four times with PBS containing 0.05%Tween 20 before 0.1 ml of MAb culture supernatant, diluted1:10 (vol/vol) in PBS containing 1.5% ovalbumin, was added.Plates were incubated for 2 h at room temperature and washedas described above. Wells were incubated with 0.1 ml ofhorseradish peroxidase-conjugated rabbit anti-mouse immuno-globulins (Dako a/s, Glostrup, Denmark), diluted 1:1,000(vol/vol) in PBS-1.5% ovalbumin, for 2 h at room temperature.After a final wash, 0.1 ml of chromogenic substrate (tetra-methylbenzidine) was added to each well. Absorbance at 450nm was read after 30 min and the addition of 50 ,ul of 1 MH2SO4 per well.SDS-PAGE and Western blot analysis. SDS-polyacrylamide

gel electrophoresis (PAGE) of protein extracts on 10.6%modified Laemmli gels and the subsequent transfer of proteinsto Immobilon polyvinylidene difluoride membranes (Millipore,Bedford, Mass.) were performed as described previously (16).LPS analysis was performed with SDS-10% polyacrylamidegels by using the Tricine buffer system (38).

Direct binding of radiolabelled IgD to intact bacteria. Thebinding of 1251I-labelled human IgD myeloma 4490 to intactbacteria was performed as described earlier (2).Enzyme assay. Sonic extracts of H. influenzae grown to

mid-logarithmic phase in 35 ml of sBHI broth were preparedas described previously (30) except that cells were not frozenprior to sonication. The glycerophosphodiester phosphodies-terase activities of sonicate supernatants were determined by acoupled spectrophotometric assay using sn-glycerol-3-phos-phate dehydrogenase and NAD with glycerophosphocholine asthe substrate (7, 22, 23).

Experimental rat otitis model. Bacteria were inoculated intobroth and allowed to grow to a density of approximately 2 x108 CFU per ml. Suspensions were then diluted with sBHI orconcentrated by centrifugation for 10 min at 4,000 x g at 4°C,suspended in broth, and kept on ice until used (0.5 to 1.5 h). Insome experiments, bacteria were heat inactivated at 100°C for

VOL. 62, 1994

on July 21, 2019 by guesthttp://iai.asm

.org/D

ownloaded from

4850 JANSON ET AL.

kanrj

0° -°G°\ +0° sGOiG0 5tI

-o. hpd

L4

hpdprobe

kanr

pUC4K

pHIC348

Ie pHIK3

hpd

FIG. 1. Insertion and deletion mutagenesis of pHIC348 for theconstruction of a protein D-deficient mutant. The restriction endonu-clease sites used in the construction of the mutant and Southern blotanalysis of the resultant H. influenzae mutant are indicated. VectorDNA sequences are represented by thicker lines. The kanamycinresistance gene is indicated by a gray box. The direction of the hpdopen reading frame (open box) is indicated by an arrow.

10 min prior to the operation. The middle ears of maleSprague-Dawley rats were injected with approximately 50 ,1 ofbacterial suspension exactly as previously described (27). Op-erations and inspections were carried out under a conventionalotomicroscope by two team members blinded to the identity ofeach animal and each inoculum. The animals were observedfor signs of infection on days 4 and 8 after the operation.

RESULTS

Construction of protein D-negative mutant. To construct amutant deficient in the expression of protein D, we began withplasmid pHIC348 (17), which contains two HaeIII sites withinthe protein D-encoding gene. The internal 534-bp HaeIII sitewas replaced with a kanamycin cassette isolated on a 1.3-kbpHincII fragment from pUC4K (Fig. 1). Protein D-negativetransformants were identified by a colony immunoassay withMAb 19B4. Plasmid DNA was prepared from a selectedprotein D-negative transformant. Restriction endonucleaseanalysis of this plasmid, designated pHIK3, showed that thekanamycin cassette from pUC4K had been inserted into theexpected site of the gene encoding protein D. In addition,protein D was not detectable in Western blots of whole-celllysates of this E. coli transformant when probed with proteinD-directed MAb 19B4 (data not shown).pHIK3 was linearized with BamHI and transformed into

NTHi 772. Kanamycin-resistant transformants lacking reactiv-ity to MAb 19B4 were identified by a colony immunoassay.Polyclonal antiserum directed against protein D was used inWestern blot analysis to exclude the possibility that the lack ofreactivity to MAb 19B4 was due to a loss of the epitope towhich it binds. Triton X-114 extracts of one selected Mab19B4-unreactive transformant (772Ahpdl) and the wild-typeparent strain (772) were subjected to SDS-PAGE analysis andthen Coomassie blue staining or Western blot analysis (Fig. 2).The solubilized protein D from H. influenzae 772 partitioned tothe detergent-phase fraction, consistent with previous results(16). A considerable amount of protein D was also found inTriton X-114-insoluble material. The MAb 19B4-unreactivetransformant did not express any immunologically detectablelevels of protein D when probed with the affinity-purifiedantiserum raised against purified protein D (Fig. 2). Nodifferences in protein profiles except for protein D expression

772 772AhpdlA D A D

67.1 _ ..-.....

43.0

-30

20.1 F L

14.4 I

1

772 772AhpdlI AD AD

2FIG. 2. Triton X-1 14 extraction and phase partitioning of proteins

from H. influenzae 772 and H. influenzae 772Ahpdl. Solubilizedproteins were acetone precipitated prior to SDS-PAGE through 10.6%T-3.3% C polyacrylamide gels. Triton X-114-insoluble residue andprecipitated proteins were suspended in sample buffer and boiled for5 min under reducing conditions prior to loading. (1) Coomassiebrilliant blue staining; (2) Western blot with affinity-purified polyclonalrabbit anti-protein D antibodies. Lanes I, Triton X-114-insolubleproteins; lanes A, aqueous-phase fractions; lanes D, detergent-phasefractions. Molecular sizes (in kilodaltons) are given on the left.

were observed for extracts from the mutant and the parentalwild-type strain. Southern blot analysis with the protein D geneprobe described above demonstrated an increase in the size ofthe 15-kbp EcoRI fragment containing the 5' part of hpd (18)to approximately 16 kbp in strain 772Ahpdl. In contrast, the1.9-kbp PstI fragment containing the gene encoding protein Ddemonstrated a decrease in size to 0.7 kbp (Fig. 3A). Toconfirm that the differences in sizes for strain 772Ahpdl weredue to the insertion of the kanamycin cassette and deletion ofthe 534-bp HaeIII fragment, a probe corresponding to the

1 234 1234

23.1-,,9.46.6- !'4.4-

2.3-2.0

A BFIG. 3. Southern blot analysis of chromosomal DNA from H.

influenzae 772 and H. influenzae 772Ahpdl. Restriction fragments ofchromosomal DNA hybridizing with a digoxigenin-labelled hpd-specific probe (A) or a 32P-labelled kanamycin linker-specific probe(B). Lanes 1 and 2, EcoRI-digested DNA from H. influenzae 772 andH. influenzae 772Ahpdl, respectively. Lanes 3 and 4, PstI-digestedDNA from H. influenzae 772 and H. influenzae 772Ahpdl, respectively.Molecular sizes (in kilobase pairs) are indicated on the left. The sizesof fragments larger than 8 kbp are approximations as a result of thetechnique's low resolution of large DNA fragments.

P .7

INFECT. IMMUN.

on July 21, 2019 by guesthttp://iai.asm

.org/D

ownloaded from

PROTEIN D AFFECTS VIRULENCE IN A RAT OTITIS MODEL 4851

M 1 2 3 42.0 [

1.0

0.5

00(0

0

0.1

0 2 4 6 8

Time (h)

FIG. 4. Comparison of the growth rates of the wild-type parentstrain, 772 (open circles), and the protein D mutant (closed squares) insBHI. Cultures of 100 ml in 1-liter flasks were grown at 37°C with200-rpm agitation. Samples were taken every 30 min, and the opticaldensity at 600 nm (OD6.) was determined. Data are from onerepresentative experiment of the four performed.

kanamycin linker of pUC4K was prepared. This probe hybrid-ized to a 16-kbp EcoRI fragment and a 1.3-kbp PstI fragmentofDNA from the mutated strain (Fig. 3B). The latter fragmentcorresponds to the size of the kanamycin cassette since itcontains PstI sites flanking the kanamycin resistance gene (Fig.1). DNA from H. influenzae 772 did not hybridize with thekanamycin probe. From these data, it is concluded that strain772Ahpdl is an isogenic mutant of NTHi strain 772 that nolonger expresses detectable levels of protein D.

Phenotypic characterization of the protein D-deficient mu-tant. The protein D-negative mutant was compared withwild-type strain 772 in various experiments before it was testedin an animal model. The inability to express protein D wouldlead to a deficiency in catalyzing the reaction from glycero-phosphodiesters to glycerol-3-phosphate. The lack of proteinD would also result in the inability to bind to human IgDmyeloma 4490. To test these hypotheses, we performed en-zyme activity and IgD binding experiments. The glycerophos-podiester phosphodiesterase activity in a sonic extract of strain772Ahpdl was less than 4% of the activity in the sonicatesupernatant of strain 772. In binding experiments with radio-labelled human IgD myeloma 4490, the protein D-deficientmutant bound 8% of the added IgD in comparison with 59%by the parental strain, in agreement with earlier results (17).The growth characteristics of the NTHi mutant lacking

protein D were compared with those of the wild-type parentstrain, 772. The experiment was repeated three times. Nosignificant difference in cell density or generation time wasfound between the two strains (Fig. 4).

Characterization of the LPS of the mutant strain in thisstudy involved both antigenic and physicochemical analyses ofLPS molecules. The protein D-negative mutant and the paren-tal strain showed identical binding patterns to the six MAbsrecognizing different epitopes in the oligosaccharide portion ofH. influenzae LPS in a whole-cell ELISA. SDS-PAGE analysisand silver staining of purified LPS were used to study LPSexpression from these two strains. No difference in the amountof LPS expressed or in electrophoretic mobility was foundbetween LPS from these NTHi strains (Fig. 5).

FIG. 5. Silver stain after Tricine-SDS-PAGE analysis of LPS. LaneM, molecular size standard, with sizes given on the left; lane 1, H.influenzae 772; lane 2, H. influenzae 772Ahpdl; lane 3, mouse-passagedH. influenzae 772; lane 4, mouse-passaged H. influenzae 772Ahpdl.Samples were boiled for 5 min prior to being loaded. The position ofthe band reacting with LPS-specific MAbs is indicated by an arrow.

Virulence in an experimental rat otitis model. To investigatethe role of protein D in the pathogenesis of the upperrespiratory tract, we used an otitis media model that wasoriginally designed for experimental pneumococcal otitis me-dia (15). This model was recently used for studying the courseof experimental otitis media caused by H. influenzae (27). H.influenzae 772 and 772Ahpdl, grown to mid-logarithmic phase,were used to inoculate the middle ears of the animals. Equalnumbers of animals were inoculated with each strain at each ofthe three operation occasions. The animals were examined forsigns of infection by otomicroscopy on days 4 and 8 after theoperation. The diagnosis of acute otitis media required directvisualization of opaque fluid behind the tympanic membrane,and a clear effusion was diagnosed as a reaction. Figure 6reports the diagnoses of animals that were injected withvarious amounts of H. influenzae 772 or H. influenzae772Ahpdl. The minimal concentration of NTHi 772 requiredto cause otitis media among all operated rats was 107 CFU/ml(Fig. 6A), whereas a 100-times-higher concentration of theprotein D-deficient isogenic mutant 772Ahpdl was required tocause otitis media among eight of nine animals challenged(Fig. 6B). Heat-killed bacteria, given in high doses, did notinduce otitis to the same degree as viable bacteria (Fig. 6).

DISCUSSION

NTHi has become well established as an important humanpathogen in the past several years. Clear distinctions betweenNTHi strains and encapsulated strains of serotype b concern-ing patient populations, types of infections caused, differencesin surface antigens presented to the host, and genetical differ-ences have been established (31). The pathogenesis of NTHidiffers from that of Hib strains in that NTHi strains oftenestablish their infection locally in the nasopharynx, from whichthere is a contiguous spread to other parts of the upperrespiratory tract. Encapsulated strains are often more invasiveand enter the bloodstream or cerebrospinal fluid. The capsuleis believed to be the most important virulence factor of Hib.NTHi has developed adhesins in order for it to persist on themucosa, but the mechanism behind its ability to cause diseaseis still unknown. LPS of H. influenzae demonstrates biologic

VOL. 62, 1994

on July 21, 2019 by guesthttp://iai.asm

.org/D

ownloaded from

4852 JANSON ET AL.

A

8*.~ 7-

E 6-

o 5

1x104 1x105 1x106 1X107 1X,08 dxeoadead

Inoculum size (cfu/ml)

B

9

8

7

1X106 1x107 1x108 1x109 1x1011 1x109

dead

Inoculum size (cfu/mI)

FIG. 6. Rat otitis model. Rats were challenged with approximately

50 pJw of H. influenzae 772 (A) or H. influenzae 772Ahpdl (B) at the

given concentrations. Animals were divided into different categories,

depending on observations performed on days 4 and 8 after the

operation. The most severe diagnosis of each animal is shown. A white

bar represents no otitis, a gray bar represents a reaction with clear

effusion, and a black bar represents otitis with purulent effusion. The

columns at the far right (1 x 109, dead) represent bacteria that were

heat killed (100°C for 10 min) before injection into the middle ears of

rats.

activities characteristic of endotoxins (11) and is an important

virulence factor, but its role in the pathogenesis of NT7Hi is not

fully understood.

We sought to investigate what role the surface-exposed

lipoprotein protein D has in the pathogenesis of NtcHi in themiddle ear. Protein D binds to immunoglobulin D, an antibodythat is present only in trace amounts in serum (9). A substan-

tial local synthesis of IgD in the nasopharynx and middle ear

cavity has been observed (39, 40); therefore, it is tempting to

speculate that the IgD binding of protein D is important forthe virulence of NTHi in the upper respiratory tract. It was

recently shown that protein D does not bind the majority ofhuman IgD myelomas tested (37). However, it was not shown

that unreactive IgD myelomas are typical of the IgD present in

the upper respiratory tract. There is a possibility that protein D

only binds to IgD of a certain configuration or in association

with some factor found only in the upper respiratory tract. We

are currently analyzing differences in the structures of IgDs

and the role that structure has in the binding of IgD to proteinD.

In this study, we have shown that the expression of proteinD renders an NTHi more than 100 times more virulent in anexperimental rat otitis model than an isogenic protein D-

deficient mutant. It is not possible to draw any conclusionsabout the mechanism behind this difference in virulence fromour data, since we have no evidence that protein D interactswith rat IgD. The difference in virulence is probably not due todifferences in LPS expression or some cell wall structure, sinceheat-killed bacteria at high concentrations were less patho-genic to rats than viable bacteria (Fig. 6). The bacteria used forthe inoculation of the middle ears of rats were passagedthrough mice and obtained from the animals' blood prior toinjection into rats since it was observed in another study thatprolonged passage in vitro made bacteria less virulent (26).The LPS expression of Hib can vary, depending upon the sitein the body from which bacteria are isolated and/or cultureconditions (20). The expression of pili also varies in a similarmanner. Hib strains isolated from mucosal sites are oftenpiliated, whereas bacteria from systemic sites do not expresspili (25). We were unable to detect any variation in electro-phoretic mobility or antigenicity of LPS between animal-passaged and non-animal-passaged strains, and we used bac-teria isolated from the blood of mice in our studies. It is,therefore, less likely that the alteration in virulence observed inour experiment was due to changes in the phenotypic expres-sion of LPS or differences in the degree of piliation. Theinsertion of an antibiotic resistance gene into the gene encod-ing protein D might have polar effects on possible virulencegenes situated immediately downstream of hpd. This risk is,however, relatively small since we have recently identified the5' end of an open reading frame approximately 220 bpdownstream of hpd (1) which is homologous to the glyceroldiffusion facilitator protein (GlpF) of E. coli (46). The down-stream open reading frame is preceded by a putative promoter;therefore, it probably belongs to some operon other than hpd.

It was recently discovered that protein D is homologous tothe periplasmic glycerophosphodiester phosphodiesterase(GlpQ) of E. coli (30, 43). The lack of an enzyme involved inthe metabolism of an important carbon source might lead toimpaired growth of a mutant lacking this enzyme. Our proteinD-negative mutant of H. influenzae 772 grew as well as thewild-type strain in a complex broth medium. However, theenzyme might be more essential to the organism under thegrowth conditions of the host; consequently, it cannot beexcluded that the observed difference in virulence is a result ofgrowth differences in vivo. E. coli possesses two enzymes withglycerophosphodiester phosphodiesterase activity, cytoplasmicUgpQ (6) and periplasmic GlpQ (22). We were unable todetect any glycerophosphodiester phosphodiesterase activity insonicate extracts of NTHi 772Ahpdl. This result does not,however, show that only one enzyme with this activity ispresent in H. influenzae 772, since no UgpQ activity has beendetected in cellular extracts of E. coli glpQ mutants (6). Itseems that UgpQ can only hydrolyze diesters that are trans-ported by the ugp-encoded transport system of E. coli (6). Animportant difference between protein D and GlpQ of E. coliK-12 is that protein D is a surface-exposed lipoprotein (16)while GlpQ is a soluble, periplasmic protein (23). We have noexplanation for why protein D is localized on the surface of H.influenzae, but if it is involved in the virulence of NTHi, itssurface localization would be the most natural. There is apossibility that protein D might have a broader substratespecificity than GlpQ of E. coli (22) and act as a phospholipase.We are, therefore, currently analyzing the substrate specificityof protein D.

ACKNOWLEDGMENTSWe thank Anders Grubb for sharing his limited amount of IgD with

us and Gertrud Hansson for skillful technical assistance.

INFECT. IMMUN.

on July 21, 2019 by guesthttp://iai.asm

.org/D

ownloaded from

PROTEIN D AFFECTS VIRULENCE IN A RAT OTITIS MODEL 4853

This work was supported by grants from the Swedish MedicalResearch Council(projects 03163 and 05196), the Medical Faculty ofLund, the Alfred Osterlund Foundation, the Crafoordska Foundation,the G. and J. Kock Foundation, and the Malmo Medical ServiceFoundation.

REFERENCES1. Akkoyunlu, M., H. Janson, and A. Forsgren. Unpublished results.2. Akkoyunlu, M., M. Ruan, and A. Forsgren. 1991. Distribution of

protein D, an immunoglobulin D-binding protein, in Haemophilusstrains. Infect. Immun. 59:1231-1238.

3. Anderson, P. W., M. E. Pichichero, and E. M. Connor. 1985.Enhanced nasopharyngeal colonization of rats by piliated Hae-mophilus influenzae type b. Infect. Immun. 48:565-568.

4. Barenkamp, S. J., R. S. Munson, Jr., and D. M. Granoff. 1982.Outer membrane protein and biotype analysis of pathogenicnontypable Haemophilus influenzae. Infect. Immun. 36:535-540.

5. Barnhart, B. J., and R. M. Herriot. 1963. Penetration of deoxyri-bonucleic acid into Hemophilus influenzae. Biochim. Biophys. Acta76:25-39.

6. Brzoska, P., and W. Boos. 1988. Characteristics of a ugp-encodedand phoB-dependent glycerophosphoryl diester phosphodiester-ase which is physically dependent on the Ugp transport system ofEscherichia coli. J. Bacteriol. 170:4125-4135.

7. Bublitz, C., and 0. Wieland. 1962. Glycerokinase. Methods Enzy-mol. 5:354-361.

8. Chanyangam, M., A. L. Smith, S. L. Moseley, M. Kuehn, and P.Jenny. 1991. Contribution of a 28-kilodalton membrane protein tothe virulence of Haemophilus influenzae. Infect. Immun. 59:600-608.

9. Dunnette, S. L., G. J. Gleich, R. D. Miller, andR A. Kyle. 1977.Measurement of IgD by a double antibody radioimmunoassay:demonstration of an apparent trimodal distribution of IgD levelsin normal human sera. J. Immunol. 119:1727-1731.

10. Emanuel, J. R 1991. Simple and efficient system for synthesis ofnon-radioactive nucleic acid hybridization probes using PCR.Nucleic Acids Res. 19:2790.

11. Flesher, A. R, andR A. Insel. 1978. Characterization of lipopoly-saccharide of Haemophilus influenzae. J. Infect. Dis. 138:719-730.

12. Fomsgaard, A., G. H. Shand, M. A. Freudenberg, C. Galanos,R S.Conrad, G. Kronborg, and N. H0iby. 1993. Antibodies fromchronically infected cystic fibrosis patients react with lipopylosac-charides extracted by new micromethods from all serotypes ofPseudomonas aeruginosa. APMIS 101:101-112.

13. Forsgren, A., and A. 0. Grubb. 1979. Many bacterial species bindhuman IgD. J. Immunol. 122:1468-1472.

14. Goodgal, S. H., andR M. Herriot. 1961. Studies of transformationof Hemophilus influenzae. I. Competence. J. Gen. Physiol. 44:1201-1227.

15. Hermansson, A., P. Emgard, K. Prellner, and S. Hellstrom. 1988.A rat model for pneumococcal otitis media. Am. J. Otolaryngol.9:97-101.

16. Janson, H., L.-O. Heden, and A. Forsgren. 1992. Protein D, theimmunoglobulin D-binding protein of Haemophilus influenzae, is alipoprotein. Infect. Immun. 60:1336-1342.

17. Janson, H., L.-O. Heden, A. Grubb, M. Ruan, and A. Forsgren.1991. Protein D, an immunoglobulin D-binding protein of Hae-mophilus influenzae: cloning, nucleotide sequence, and expressionin Escherichia coli. Infect. Immun. 59:119-125.

18. Janson, H., M. Ruan, and A. Forsgren. 1993. Limited diversity ofthe protein D gene (hpd) among encapsulated and nonencapsu-lated Haemophilus influenzae strains. Infect. Immun. 61:4546-4552.

19. Kar, S., S. C.-M. To, and C. C. Brinton, Jr. 1990. Cloning andexpression in Escherichia coli of LKP pilus genes from a nontype-able Haemophilus influenzae strain. Infect. Immun. 58:903-908.

20. Kimura, A., and E. J. Hansen. 1986. Antigenic and phenotypicvariations of Haemophilus influenzae type b lipopolysaccharide andtheir relationship to virulence. Infect. Immun. 51:69-79.

21. Kimura, A., C. C. Patrick, E. E. Miller, L. D. Cope, G. H.McCracken, Jr., and E. J. Hansen. 1987. Haemophilus influenzaetype b lipooligosaccharide: stability of expression and associationwith virulence. Infect. Immun. 55:1979-1986.

22. Larson, T. J., M. Ehrmann, and W. Boos. 1983. Periplasmicglycerophosphodiester phosphodiesterase of Escherichia coli, anew enzyme of the glp regulon. J. Biol. Chem. 258:5428-5432.

23. Larson, T. J., and A. T. van Loo-Bhattacharya. 1988. Purificationand characterization of glpQ-encoded glycerophosphodiesterphosphodiesterase from Escherichia coli K-12. Arch. Biochem.Biophys. 260:577-584.

24. Maniatis, T., E. F. Fritsch, and J. Sambrook 1982. Molecularcloning: a laboratory manual. Cold Spring Harbor Laboratory,Cold Spring Harbor, N.Y.

25. Mason, E.O., Jr., S. L. Kaplan, B. L. Wiedermann, E. P. Norrod,and W. A. Stenback 1985. Frequency and properties of naturallyoccuring adherent piliated strains of Haemophilus influenzae typeb. Infect. Immun. 49:98-103.

26. Melhus, A. Unpublished results.27. Melhus, A, A. Hermansson, and K Prellner. 1994. Nontypeable

and encapsulated Haemophilus influenzae yield different clinicalcourses of experimental otitis media. Acta Oto-laryngol. 114:289-294.

28. Morrison, D. A. 1979. Transformation and preservation of com-petent bacterial cells by freezing. Methods Enzymol. 68:326-331.

29. Moxon, E. R, and K. A. Vaughn. 1981. The type b capsularpolysaccharide as a virulence determinant of Haemophilus influ-enzae: studies using clinical isolates and laboratory transformants.J. Infect. Dis. 143:517-524.

30. Munson, R S., Jr., and K. Sasaki. 1993. Protein D, a putativeimmunoglobulin D-binding protein produced by Haemophilusinfluenzae, is glycerophosphodiester phosphodiesterase. J. Bacte-riol. 175:4569-4571.

31. Murphy, T. F., and M. A. Apicella. 1987. Nontypable Haemophi-lus influenzae: a review of clinical aspects, surface antigens, andthe human immune response to infection. Rev. Infect. Dis.9:1-15.

32. Murphy, T. F., K. C. Dudas, J. M. Mylotte, and M. A. Apicella.1983. A subtyping system for nontypable Haemophilus influenzaebased on outer membrane proteins. J. Infect. Dis. 147:838-846.

33. Murphy, T. F., M. B. Nelson, K. C. Dudas, J. M. Mylotte, andM. A. Apicella. 1985. Identification of a specific epitope ofHaemophilus influenzae on a 16,600-dalton outer membrane pro-tein. J. Infect. Dis. 152:1300-1307.

34. Musher, D. M., K.R Kubitschek, J. Crennan, and R. E. Baughn.1983. Pneumonia and acute febrile tracheobronchitis due toHaemophilus influenzae. Ann. Intern. Med. 99:444-450.

35. Plaut, A. G. 1978. Microbial IgA proteases. N. Engl. J. Med.298:1459-1463.

36. Ruan, M., M. Akkoyunlu, A. Grubb, and A. Forsgren. 1990.Protein D of Haemophilus influenzae. A novel bacterial surfaceprotein with affinity for human IgD. J. Immunol. 145:3379-3384.

37. Sasaki, K., and R S. Munson, Jr. 1993. Protein D of Haemophilusinfluenzae is not a universal immunoglobulin D-binding protein.Infect. Immun. 61:3026-3031.

38. Schigger, H., and G. von Jagow. 1987. Tricine-sodium dodecylsulfate-polyacrylamide gel electrophoresis for the separation ofproteins in the range from 1 to 100 kDa. Anal. Biochem. 166:368-379.

39. S0rensen, C. H. 1984. IgD and secretory immunoglobulins insecretions from the upper respiratory tract of children withsecretory otitis media, p. 150-152. In D. J. Lim and B. C. Decker(ed.), Recent advances in otitis media with effusion. B. C. Decker,Toronto.

40. S0rensen, C. H. 1987. Secretory immunoglobulins in the nasophar-ynx: implications in otitis media, p. 259-278. In J. Bernstein and P.Ogra (ed.), Immunology of the ear. Raven Press, New York.

41. St. Geme, J. W., Ill, S. Falkow, and S. J. Barenkamp. 1993.High-molecular-weight proteins of nontypable Haemophilus influ-enzae mediate attachment to human epithelial cells. Proc. Natl.Acad. Sci. USA 90:2875-2879.

42. Swancutt, M. A., B. S. Riley, J. D. Radolf, and M. V. Norgard.1989. Molecular characterization of the pathogen-specific, 34-kilodalton membrane immunogen of Treponema pallidum. Infect.Immun. 57:3314-3323.

43. Tommassen, J., K. Eiglmeier, S. T. Cole, P. Overduin, T. J.Larson, and W. Boos. 1991. Characterization of two genes, glpQ

VOL. 62, 1994

on July 21, 2019 by guesthttp://iai.asm

.org/D

ownloaded from

4854 JANSON ET AL.

and ugpQ, encoding glycerophosphoryl diester phosphodiesterasesof Escherichia coli. Mol. Gen. Genet. 226:321-327.

44. Wald, E. R., G. J. Milmoe, A. Bowen, J. Ledesma-Medina, N.Salamon, and C. D. Bluestone. 1981. Acute maxillary sinusitis inchildren. N. Engl. J. Med. 304:749-754.

45. Weber, A., K Hams, S. Lohrke, L. Forney, and A. L. Smith. 1991.Inability to express fimbriae results in impaired ability of Hae-mophilus influenzae b to colonize the nasopharynx. Infect. Immun.59:4724-4728.

46. Weissenborn, D. L., N. Wittekind, and T. J. Larson. 1992. Struc-ture and regulation of the gipFK operon encoding glycerol diffu-sion facilitator and glycerol kinase of Escherichia coli K-12. J. Biol.Chem. 267:6122-6131.

47. Weller, P. F., A. L. Smith, P. Anderson, and D. H. Smith. 1977.The role of encapsulation and host age in the clearance ofHaemophilus influenzae bacteremia. J. Infect. Dis. 135:34-41.

48. Westphal, O., and K. Jann. 1965. Bacterial lipopolysaccharides:extraction with phenolwater and further applications of the pro-cedure. Methods Carbohydr. Chem. 5:83-91.

49. Yanisch-Perron, C., J. Vieira, and J. Messing. 1985. ImprovedM13 phage cloning vectors and host strains: nucleotide sequencesof the M13mpl8 and pUC19 vectors. Gene 33:103-119.

50. Zwahlen, A., L. G. Rubin, and E. R. Moxon. 1986. Contribution oflipopolysaccharide to pathogenicity of Haemophilus infiuenzae:comparative virulence of genetically related strains in rats. Microb.Pathog. 1:465-473.

INFECT. IMMUN.

on July 21, 2019 by guesthttp://iai.asm

.org/D

ownloaded from