Serologic testing for syphilis: missing the point

You may have seen several news sources touting the recent CDC finding that nearly one in five positive reactions with a newer syphilis test are wrong.  These headlines may grab the reader's attention, but the press took the finding out of context and failed to deliver the real message that the CDC was trying to convey.  Worse, the press reports may needlessly confuse and worry those who are being treated for syphilis.

Serological tests for syphilis are grouped into two categories.  Nontreponemal tests such as the VDRL and RPR are based on antibody generated against the lipid cardiolipin.  Presumably cardiolipin is released from damaged tissue in syphilis patients and gets incorporated into the membrane of Treponema pallidum.  The reason that these tests are "nontreponemal" is that antibodies to cardiolipin accompany many other conditions.  On the other hand, treponemal tests use T. pallidum proteins or even the entire spirochete as antigen to detect antibodies against the spirochete.  Although the classic treponemal tests such as the FTA-ABS (fluorescent treponemal antibody-absorption) and TP-PA (Treponema pallidum particle agglutination) are still used, the newer automated EIA (enzyme immunoassay) and CIA (immunochemiluminescence) treponemal tests enable clinical laboratories to rapidly screen a large number of sera.

The traditional approach to syphilis testing is to first screen the patient's serum with a nontreponemal test.  Since nontreponemal tests can give false positive reactions, reactive sera are retested with one of the treponemal tests.  However, the low cost of executing the automated treponemal tests have led some high-volume clinical laboratories to reverse the order of the assays:  they screen with the EIA/CIA treponemal test and confirm positive results with a nontreponemal test.  The CDC report in the Morbidity and Mortality Weekly Report deals with this so-called "reverse sequence" testing.

So where did the "nearly one in five" figure come from?  From 2006 to 2010, five large clinical laboratories screened 140,176 sera specimens with the reverse sequence procedure.  Of the 4,834 reactive with the EIA/CIA treponemal test, 2,743 gave negative results with the nontreponemal RPR test.  When the samples that gave discrepant results were tested further with one of the classic treponemal tests, 866 of the 2,743 samples were negative.  Overall, among the 4,834 samples that were reactive with the newer treponemal test, 866 or 18% were nonreactive with two subsequent tests.  These 866 were assumed to be false positives.

The news media pounced on the 18% figure and declared that hundreds may have been given antibiotics to treat a disease that they didn't have.  But they ignored the fact that doctors don't diagnose syphilis on the basis of a single lab test.  It is standard practice to perform a second test when the first comes back positive and to do even a third one if warranted.  Doctors also take into account the physical exam and the sexual and medical history of the patient before making the decision to treat with antibiotics.

Here's how the CDC responded to the assertion that those among the 18% may have been falsely diagnosed and treated unnecessarily with antibiotics:

There are two problems with this assertion. First, the current report does not document whether or not treatment was provided. Second, in those cases where treatment was provided, it may have been justified based on sexual risk and findings on clinical evaluation. It is also important to note that syphilis is not diagnosed on the basis of a single blood test. Many labs routinely will do additional testing when the first test is positive, without notifying the patient. Doctors diagnose syphilis after considering at least two syphilis tests, the patient's history, the physical exam, and a review of past syphilis test results. The MMMR analysis, while important, does not allow us to conclude that the newer tests led to inaccurate syphilis diagnosis or inappropriate treatment.

So what was the message that the CDC was trying to communicate to readers of the MMMR report?  Their intention was to provide guidance in the management of cases for which the reverse sequence screening is performed instead of the traditional sequence, which is still recommended by the CDC.  Specifically, when conflicting results occur (positive with the treponemal test, negative with the nontreponemal test), a third test should be done with the TP-PA.  (The CDC does not recommend the FTA-ABS because it is less specific and probably less sensitive.)  A positive reaction with the TP-PA indicates past or present syphilis; a negative reaction indicates that syphilis is unlikely.  As always, the clinical observations and medical history of the patient should also be considered in making an informed treatment decision.


Reference

Centers for Disease Control and Prevention (February 11, 2011).  Discordant results from reverse sequence syphilis screening -- five laboratories, United States, 2006-2010.  MMMR. Morbidity and Mortality Weekly Report 60(5):133-137.  link

Another reason not to own a pet rat

A pair of European doctors wrote about the trials of a 37-year-old man who came down with Weil's disease, a severe form of leptospirosis.  The case report appeared in the recent issue of The Lancet Infectious Diseases.

The individual showed up in the emergency room complaining of a sudden fever, muscle pain, and a severe headache.  Notable were the signs of jaundice and conjunctival suffusion (reddening of the eye), which suggested leptospirosis.

Figure from Jansen and Schneider, 2011

Shortly after he was admitted, the patient had to be moved to the intensive care unit because his kidneys were failing.  Blood tests for Leptospira antibodies came back negative, most likely because it was too early in the infection for antibodies to appear.

An infectious diseases specialist noted that the patient suffered from the classic signs of Weil's disease:  kidney failure, jaundice, and an enlarged spleen.  Accordingly, the patient was started on intravenous penicillin, the favored treatment for severe leptospirosis.  Further blood tests conducted during the second week of treatment finally revealed antibodies to Leptospira.  The patient completely healed within several weeks.  The only question that remained is how the patient got infected with Leptospira.

At some point the patient admitted that he owned four pet rats.  Since rats can silently carry Leptospira in their kidneys, his pets were sacrificed to check their kidneys for the presence of the spirochete.  All four kidneys turned out to be culture positive for Leptospira interrogans, serovar Icterohaemorrhagiae or Copenhageni (the typing test couldn't distinguish the two serovars).  The patient's antibodies reacted against the same serovars, suggesting that the source of the infection was tainted urine from his pet rats.

Although cases of leptospirosis acquired from pet rats appear to be rare, anyone who wishes to own a rat should be aware that rats can be carriers of a potentially deadly spirochete.  This is yet another reason why rats should be obtained from a responsible breeder.

For another case, check out this post from the Worms & Germs Blog (make sure you read the comments for the complete story).

Reference

Jansen, A. and Schneider, T. (February 2011).  Weil's disease in a rat owner.  The Lancet Infectious Diseases 11(2):152.  DOI: 10.1016/S1473-3099(10)70106-7

The quest for outer membrane proteins of the stealth pathogen Treponema pallidum: the cliffhanger episode

Syphilis patients are able to generate antibodies against the spirochete Treponema pallidum.  However, if you were to mix sera from these patients with T. pallidum in a test tube, very few of the antibodies in the sera would bind to the spirochetes.  The reason is that the strange outer membrane architecture of T. pallidum makes the spirochete invisible to the antibodies.  Most of the proteins and lipid molecules targeted by the antibodies lie beneath the outer membrane.

The outer membrane of T. pallidum differs considerably from that of a typical Gram-negative bacterium.  The most glaring difference is that T. pallidum lacks lipopolysaccharide (LPS), a favorite target of the immune response.  The outer membrane is also bare of other potential surface antigens except for a very small number of transmembrane outer membrane proteins (Omps) and (possibly) surface lipoproteins.  The poor surface antigencity may help the so-called "stealth pathogen" persist in the body despite a robust immune response to the infection.

The few Omps displayed on the surface of T. pallidum must be doing something really important if the spirochete is willing to risk exposing them to attack by antibodies.  For this reason, scientists have been seeking the identity of these Omps to figure out what they do.  These rare Omps could also be fashioned into a long-desired syphilis vaccine.

Unfortunately, the scarcity of Omps and the delicate nature of the outer membrane of T. pallidum have stymied efforts to identify Omps.  The routine centrifugation and washing steps used to prepare other bacteria for analysis easily damage the outer membrane of T. pallidum, causing the loss of Omps and exposing the abundant periplasmic and inner membrane proteins.  Consequently, probes used to identify exposed proteins may react with the periplasmic and inner membrane proteins, which are normally shielded by the outer membrane.  Without the proper controls, this would lead one to conclude wrongly that a non-Omp that reacts with the probe (such as antibodies raised against the protein of interest) is surface exposed.  In addition, when the outer membrane is purified with the intention to identifying Omps, it is hard to distinguish the tiny amounts of Omps from proteins from other bacterial compartments contaminanting the outer membrane preparation.

Despite these technical challenges, several Omp candidates have been proposed, but those proteins are either mired in controversy (TprK, for example) or await experimental confirmation of their surface exposure.  To date, no protein that has been demonstrated unambiguously to be displayed on the exterior of T. pallidum.

The past decade has seen the development of several computer programs that can be used to predict whether a given gene encodes a transmembrane Omp.  The algorithms differ, but all of these programs attempt to identify amino acid sequences that fold into a β-barrel, which forms the core of transmembrane Omps whose 3D structures are known.  The β-barrel forms when an anti-parallel β-sheet rolls into the shape of a barrel.  Most of the β-barrel is embedded in the outer membrane so that the loops connecting the β-strands stick out from the two surfaces of the membrane.  The loops displayed on the external face of the outer membrane would be accessible to antibodies.  The size and composition of the loops vary among different Omps, but all β-strands tend to have alternating hydrophobic amino acids with their nonpolar side chains acid protruding out from the barrel into the hydrophobic interior of the lipid bilayer.  Each β-strand consists of 9-11 amino acid residues and is tilted up to 45° out of the transmembrane axis.  Different β-barrels have as few as 8 and as many as 22 transmembrane β-strands.

The ribbon representation of OmpA, an 8-stranded transmembrane Omps from E. coli, is shown below as one example.

From Figure 1b of Smith et al., 2007.  The N- and C-terminal β-strands are colored brown and blue, respectively.  The side chains of the "aromatic girdle" are shown.

Below, OmpA is unfurled to show the topology of the protein:

From Figure 1a of Smith et al., 2007.  β-strand amino acid residues are depicted as diamonds, and loop residues are depicted as circles.  Alternating hydrophobic amino acid residues within the  β-strand are colored red (aromatic residues of the girdle) and yellow.

Assuming that the rare transmembrane Omps of T. pallidum share the β-barrel structure, the obvious computational approach to finding these Omps would be to run all of the proteins encoded by the T. pallidum genome through one of these programs.  One problem with these programs is that they will pick up a few proteins that are not truly transmembrane Omps.  To minimize this problem, Justin Radolf's group, as reported in the December 2010 issue of Infection and Immunity, ran the 1038 protein-coding sequences of T. pallidum through seven different Omp-predicting programs.  They found that two proteins were predicted by all seven programs to have the β-barrel structure; another four candidates were identified by six programs.

One of the proteins at the top of the list, identified by all seven programs, was TP0326, a BamA homolog encoded by the genomes of many Gram-negative bacteria.  Experiments with other bacteria have shown that BamA is a member of an outer membrane protein complex that assembles other transmembrane Omps into the outer membrane, so it would make sense for T. pallidum to possess such a protein.  BamA itself is thought to be a transmembrane Omp. 

This wasn't the first time that a syphilis researcher has encountered TP0326.  In a study published 11 years ago, before the function of BamA was known, Caroline Cameron and colleagues demonstrated that antibodies raised against TP0326 (also called "Tp92" in their paper) stimulated macrophages to engulf T. pallidum in a process called opsonophagocytosis.  In addition, TP0326 was somewhat effective as a vaccine in the rabbit model of syphilis:  rabbits that had been immunized with TP0326 experienced milder skin lesions than unimmunized rabbits following inoculation of T. pallidum into the skin.  These observations indirectly supported the localization of TP0326 to the outer membrane since opsonophagocytosis and effective vaccination require a target that is accessible on the surface of the spirochete.  However, this earlier work lacked a more direct test such as the indirect immunofluorescence assay to confirm that TP0326 was exposed on the surface.

The problem with the standard two-step indirect immunofluorescence assay is that it is not sensitive enough to detect the rare Omps of T. pallidum.  Therefore, as described in the Infection and Immunity paper, Radolf's group tinkered with the assay and managed to amplify the output signal by adding a third step to the procedure.  To minimize damage to the outer membrane during the centrifugation and washing steps, the spirochetes were encased in gel microdroplets, which protected the delicate outer membrane while allowing antibodies to permeate to probe the T. pallidum surface.

With the modified immunofluorescence assay, the investigators were able to detect surface proteins with syphilitic antibodies for the first time, although only in a small minority of the spirochetes in the field of view lit up with the red color (see figure below).  Presumably, the other spirochetes failed to react with the antibodies because they didn't quite have enough Omp antigens being expressed on their surface (although a more interesting explanation would be that the nonreactive spirochetes had down-regulated their surface Omps).  Regardless of the true explanation, these results indicated that at least some of the antibodies generated by syphilis patients were directed against surface components of T. pallidum.  When the investigators treated the spirochetes with the detergent Triton X100 to intentionally damage the outer membrane, all of the spirochetes glowed, indicating that most of the antibodies targeted proteins beneath the surface of T. pallidum.  As a negative control, they demonstrated that sera from healthy patients failed to react with intact spirochetes.

To keep track of how many spirochetes were damaged by the procedure, the investigators added antibody raised against the periplasmic flagella along with the patient antibodies.  The flagellar antibodies would bind to the spirochetes only if the integrity of the outer membrane was compromised by handling the spirochetes.  The assay was designed so that bound flagellar antibodies would glow green.

From Figure 4 of Cox et al., 2010.   (A) All spirochetes, whether or not they fluoresced, could be seen with darkfield optics (DF).  Spirochetes that bound to antibodies from syphilis patients (HSS) glowed red.   Spirochetes with a disrupted outer membrane reacted with the flagellar antibody (anti-FlaA) and glowed green.  (B) 5.8% of the spirochetes observed were undamaged and reacted with patient antibodies (glowed red but not green).  Another 5.1% were damaged (glowed red and green).  89.0% of the spirochetes failed react with the patient antibodies.  100% of the spirochetes fluoresced when treated with the detergent Triton X100 before adding the antibodies.

With an improved immunofluorescence assay, the investigators were poised to test the proteins at the top of the list for surface exposure.  As I was nearing the end of the paper, I was expecting the authors to describe their test of the BamA homolog TP0326 for surface exposure.  Surprisingly, they ended the paper without testing any of the proteins near the top of the list.

I can only assume that the authors are planning to submit a separate manuscript in the future describing the successful detection of TP0326 or another protein near the top of the list.  But the problem with ending the paper without demonstrating surface localization of even a single protein is that one can question whether even the 3-step immunofluorescence assay is sensitive enough to detect an Omp exposed on the T. pallidum surface.  They did test two proteins lower down on the list that other labs believe are surface exposed (TprK and a fibronectin-binding lipoprotein)  but neither protein was detected on the outer membrane surface by the modified immunofluorescence assay.  So we are left with an assay that certainly has more sensitivity, but is it sensitive enough?

To be continued...(?)

Featured paper

Cox, D.L., Luthra, A., Dunham-Ems, S., Desrosiers, D.C., Salazar, J.C., Caimano, M.J.., and Radolf, J.D. (December 2010).  Surface immunolabeling and consensus computational framework to identify candidate rare outer membrane proteins of Treponema pallidum.  Infection and Immunity 78(12):5178-5194.  DOI: 10.1128/IAI.00834-10

Other references

Radolf, J.D. (June 1995).  Treponema pallidum and the quest for outer membrane proteins.  Molecular Microbiology 16(6):1067-1073.

Cameron, C.E., Lukehart, S.A., Castro, C., Molini, B., Godornes, C., and Van Voorhis, W.C. (April 2000).  Opsonic potential, protective capacity, and sequence conservation of the Treponema pallidum subspecies pallidum Tp92.  Journal of Infectious Diseases 181(4):1401-1413.  DOI: 10.1086/315399

Cox, D.L., Akins, D.R., Porcella, S.F., Norgard, M.V., and Radolf, J.D. (1995).  Treponema pallidum in gel microdroplets:  a novel strategy for investigation of treponemal molecular architecture.  Molecular Microbiology 15(6):1151-1164.

Image source

Smith, S.G.J, Mahon, V., Lambert, M.A., and Fagan, R.P. (August 2007).  A molecular Swiss army knife:  OmpA structure, function and expression.  FEMS Microbiology Letters 273(1):1-11.  DOI: 10.1111/j.1574-6968.2007.00778.x

Designing a Lyme disease vaccine to attack the tick vector

Conventional vaccines target the surface components or secreted toxins of pathogens.  Erol Fikrig's group at Yale University has been exploring an unconventional approach towards developing a vaccine for Lyme disease, which is caused by a tick-borne pathogen.  Their recent work, published in the November issue of PLoS Pathogens, demonstrated partial success in protecting laboratory mice by immunization with a protein found in the saliva of the Ixodes tick vector.

Ixodes ticks spend several days feeding on blood while attached to the victim's skin.  B. burgdorferi is carried into the victim's skin in the Ixodes tick's saliva starting 3-4 days (on average) after attachment.  Tick saliva contains a blend of biological substances that aid the tick in drinking blood from its victim.  These substances include cement proteins to keep the tick's feeding apparatus tightly bound to the skin, anti-coagulants to keep the blood flowing into the tick, and anti-inflammatory factors that ward off the local inflammatory response.  The activity of these substances also promote transmission of B. burgdorferi from the feeding tick to the victim.  Hence a vaccine that targets a saliva component may protect humans from Lyme disease.

A Lyme vaccine that targets the tick has a few advantages over one that targets the spirochete.  First, a tick-based Lyme disease vaccine is unlikely to interfere with laboratory diagnosis, which currently relies on detection of antibodies against the Lyme Borrelia spirochete.  Second, an effective vaccine that targets the tick may also prevent transmission of other pathogens carried by the Ixodes tick by interfering with tick feeding or with the tick's countermeasures against the host inflammatory response at the feeding site.

In their recent work, the investigators focused their efforts on a salivary protein called tick histamine release factor (tHRF).  Because tHRF levels in the salivary glands of feeding Ixodes ticks were higher when B. burgdorferi was present in the tick, they guessed that tHRF was doing something to help transmit B. burgdorferi from the tick to the victim.  The authors turned out to be correct.  Transmission of B. burgdorferi was impaired when they knocked down the tick's production of tHRF by RNAi.

The investigators went on to test the vaccine potential of tHRF in their mouse model.  They passively immunized mice with antiserum raised against tHRF or actively immunized the rodents with recombinant tHRF.  Actively immunized mice were also given booster injections with tHRF (the paper did not say how many).  Control mice were not immunized.  They then challenged the mice with ticks infected with B. burgdorferi.  One or three weeks later, tissues were removed from the mice, and the bacterial load of B. burgdorferi in skin, heart, and joints was measured by quantitative PCR.

Their data showed that immunization with tHRF was somewhat effective.  Depending on the experiment, B. burgdorferi DNA could not be detected in any of the three tissues in 20-33% of immunized mice, whereas the spirochete's DNA was detected in at least one tissue in all control mice.  Even in immunized mice with detectable B. burgdorferi DNA, the levels were often lower than the average level found in the control mice.  It would have been nice to know how much inflammation was present in the tissues of the immunized mice.  Unfortunately, the histopathology of the tissues was not presented in the paper.

Figure 4, panels E-G from Dai et al., 2010.  Bacterial burden in skin (day 7 after challenge) and joint and heart (day 21) was determined by quantitative PCR with flaB primers.  Horizonal lines represent the mean value ± SEM.  * p < 0.05 and ** p < 0.01.  Results were pooled from 3 independent experiments.

tHRF is not the first Lyme vaccine candidate to target a protein found in tick saliva.  An earlier report from Fikrig's group demonstrated that active and passive immunization with Salp15, another tick salivary protein, was also somewhat effective in protecting mice from colonization with B. burgdorferi.  tHRF was superior to Salp15 in impairing feeding by ticks. Ticks feeding on Salp15-immunized mice were able to complete their blood meal.  In contrast, most of the ticks had a hard time feeding on mice immunized with tHRF and could not complete their blood meal, as assessed by tick weights following detachment from the mice.

Although immunization with tHRF and Salp15 prevented colonization in only some mice, Fikrig's work shows for the first time that it may be possible to design a Lyme disease vaccine that targets the tick vector.  Ultimately, the most effective vaccine may be a mixture that targets multiple components in both the tick and the spirochete.

References

Dai, J., Narasimhan, S., Zhang, L., Liu, L., Wang, P., & Fikrig, E. (2010). Tick histamine release factor is critical for Ixodes scapularis engorgement and transmission of the Lyme disease agent PLoS Pathogens, 6 (11) DOI: 10.1371/journal.ppat.1001205

Dai, J., Wang, P., Adusumilli, S., Booth, C.J., Narasimhan, S., Anguita, J., & Fikrig, E. (2009). Antibodies against a tick protein, Salp15, protect mice from the Lyme disease agent. Cell Host & Microbe, 6 (5), 482-492 DOI: 10.1016/j.chom.2009.10.006

Related post

• A fresh approach towards a Lyme disease vaccine: targeting the tick

How bacteria sort their lipoproteins (Lol!)

Bacterial lipoproteins are proteins with covalently-attached lipid molecules that anchor the protein to the cytoplasmic or outer membrane.  The lipid molecules are attached to the cysteine located at the amino terminus of the lipoprotein.  The lipoprotein's protein component, being hydrophilic (water-loving), sticks out from the membrane. Different bacterial lipoproteins participate in a variety of functions, including transport of molecules, stabilization of the cell wall, signal transduction, motility, and interaction with host molecules.

The Lyme disease spirochete Borrelia burgdorferi is exceptional in that a number of different lipoproteins have been found on its surface.  Most other bacteria lack (or have few) surface-exposed lipoproteins.  To give one example, the figure below shows the arrangement of lipoproteins in the cell envelope of E. coli, home to roughly 90 lipoproteins, none known to be displayed on the surface.  Lipoproteins are depicted as colored ovals with the attached squiggles representing the lipid molecules. To perform their functions properly, some lipoproteins must be anchored to the outer leaflet of the inner membrane (blue ovals) whereas the rest must be anchored to the inner leaflet of the outer membrane (red ovals).  In both cases, the protein component of the lipoprotein protrudes into the periplasm.  The figure also shows the other major category of membrane proteins, the integral membrane proteins, which are embedded in the membrane.  There are also proteins that reside in the periplasm, which are not depicted in the figure.

Figure 1 of Narita and Tokuda (2010)

In this post I will describe how lipoproteins are brought to their correct location in the bacterial envelope.  I will first describe how lipoproteins are sorted in E. coli since that's where most of the earlier work was conducted. Since many other diderms (bacteria having two membranes) have homologs of the proteins used by E. coli to export and sort lipoproteins, E. coli is a good model for studying localization of lipoproteins.  Monoderm bacteria also have lipoproteins, but since they have only one membrane, they don't need to worry about sorting lipoproteins.  (I will save the explanation of how lipoproteins get to the bacterial surface for a future post.)

Most proteins to be exported out of the cytoplasm are marked with an amino-terminal signal peptide ≈20 amino acids in length.  The sequences of the signal peptides (plus five additional amino acid residues) from two E. coli lipoproteins are shown below.  A cytoplasmic membrane protein complex called the Sec translocon transfers proteins harboring the signal peptide to the periplasm, where the signal peptide is lopped off by one of two signal peptidases.  Signal peptidase I cleaves off the signal peptide from nonlipoproteins (such as periplasmic or transmembrane outer membrane proteins), and signal peptidase II slices off the signal peptide from lipoproteins.

All lipoproteins harbor a short sequence called a "lipobox" at the end of the signal peptide (underlined in sequences below).  The lipobox consensus sequence  is -(leu, ala, val)_-4-leu_-3-(ala, ser)_-2-(gly, ala)_-1↓cys₊₁, with the arrow specifying the cleavage site for signal peptidase II and the subscripts denoting positions relative to the cleavage site.

E. coli Braun's lipoprotein (OM) MKATKLVLGAVILGSTLLAG↓CSSNA...

E. coli lpp-28 (IM)           MKLTTHHLRTGAALLLAGILLAG↓CDQSS...

(IM, inner membrane; OM, outer membrane)

The lipobox is recognized by the inner membrane enzyme phosphatidylglycerol:prolipoprotein diacylglyceryl transferase (Lgt).  Before the signal peptide is removed, Lgt attaches diacylglycerol to the sulfhydryl (-SH) of the lipobox cysteine.  After the signal peptide is cleaved off by signal peptidase II, another inner membrane enzyme, apolipoprotein N-acyltransferase (Lnt), attaches a fatty acid molecule to the newly exposed amino (-NH₃) group of the cysteine.  Only exported proteins with lipoboxes become lipidated.  The lipoprotein remains associated with the inner membrane throughout these processing steps.

The machinery responsible for sorting lipoproteins to the outer membrane is the LolCDE protein complex, a type of ABC transporter that sits in the inner membrane.  Lol stands for lipoprotein outer membrane localization.  LolCDE recognizes the lipidated cysteine at the amino terminus of lipoproteins.  LolCDE loads lipoproteins onto the periplasmic protein LolA, which ferries lipoproteins to the LolB receptor, a lipoprotein that protrudes from the periplasmic face of the outer membrane.  After capturing the lipoprotein from LolA, LolB anchors the lipoprotein into the periplasmic layer of the outer membrane.

from figure 3 of Tokuda and Matsuyama (2004)

How does LolCDE know which lipoproteins are supposed to be delivered to the outer membrane and which need to stranded in the inner membrane?  For E. coli and other members of the Enterobacteriaceae family of bacteria, the answer is fairly simple.  Lipoproteins with aspartate at the +2 position (which follows the lipidated cysteine) remain in the inner membrane.

How does the +2 aspartate prevent transfer of lipoproteins to the outer membrane?  It turns out that LolCDE doesn't directly sense the amino acid at the +2 position.  Instead, the abundant membrane phospholipid phosphatidylethanolamine (PE) is thought to interfere with LolCDE recognition of the lipidated cysteine when asparatate is at the +2 position.  When the side chain carboxyl group (-COO^-) of the +2 aspartate interacts electrostatically with the positively-charged head group of PE, the fatty acids of PE become perfectly positioned to form hydrogen bonds with the lipid molecules attached to the cysteine (see figure below).  LolCDE is unable to recognize the amino-terminal cysteine associated with five fatty acid groups (three covalently bound to the cysteine and two from PE).  Thus asparatate, when it follows the cysteine, acts indirectly as a Lol avoidance signal.  The amino acid at the +3 position can also influence the Lol avoidance signal.  For example, negatively-charged amino acids (aspartate and glutamate) at the +3 position strengthen the +2 aspartate Lol avoidance signal by stabilizing the complex between phosphotidylethanolamine and the +2 aspartate (see figure below).

Modified from figure 6 of Tokuda and Matsuyama (2004)

Additional studies with engineered lipoproteins have shown that phenylalanine, tryptophan, tyrosine, lysine, and proline, although rarely found in lipoproteins at the +2 position, can also serve as inner membrane retention signals when asparagine is at the +3 position.  Since none of these are negatively-charged amino acids, the mechanism for avoiding LolCDE must differ from those lipoproteins having aspartate at the +2 position.

The nature of the sorting signal differs for bacteria that are not members of Enterobacteriaceae.  For example, the three amino acids at positions +2 through +4 dictate whether lipoproteins will remain in the inner membrane of Pseudomonas aeruginosa.  For the spirochete B. burgdorferi, a clear rule has yet to emerge from the few studies that have been done.  What can be said is that negatively-charged amino acids (aspartate and glutamate) placed within the first several amino acids following the lipidated cysteine sometimes allows the lipoprotein to remain in the membrane.  Whether the negatively-charged amino acid functions as an inner membrane retention signal depends on which amino acids are surrounding it.  It is not yet possible to simply look at the amino-terminal sequence of B. burgdorferi lipoproteins and confidently predict in which membrane they will be found.

Although the "+2/+3/+4 rule" is useful for predicting whether a newly discovered lipoprotein will be found in the inner or outer membrane, it may not give the complete picture of all of a lipoprotein's features that govern its localization.  The rules for sorting lipoproteins were worked out primarily by examining the localization of engineered fusion proteins consisting of the amino termini of lipoproteins (signal peptide with lipobox plus the first several amino acids following the lipobox cysteine) fused to unrelated reporter proteins such as red fluorescent protein (RFP) from corals.  For example, placing asp at the +2 position of such a fusion protein would cause RFP to be retained in the inner membrane.  Changing the +2 amino acid to serine would cause RFP to be transported to the outer membrane.  However, localization of a full-length lipoprotein may not be altered by simply changing its +2 amino acid from aspartate to another amino acid or vice versa.  This indicates that the rest of the lipoprotein, the part that's removed when reporters are used, also influences the localization of lipoproteins.


References

TOKUDA, H. (2009). Biogenesis of outer membranes in Gram-negative bacteria. Bioscience, Biotechnology, and Biochemistry, 73 (3), 465-473 DOI: 10.1271/bbb.80778

TOKUDA, H. (2004). Sorting of lipoproteins to the outer membrane in E. coli. Biochimica et Biophysica Acta (BBA) - Molecular Cell Research, 1693 (1), 5-13 DOI: 10.1016/j.bbamcr.2004.02.005

Schulze, R., & Zückert, W. (2006). Borrelia burgdorferi lipoproteins are secreted to the outer surface by default. Molecular Microbiology, 59 (5), 1473-1484 DOI: 10.1111/j.1365-2958.2006.05039.x