of 18
Article
Peptidoglycan Production by an Insect-Bacterial
Mosaic
Graphical Abstract
Highlights
d
Mealybugs have two bacterial endosymbionts; one symbiont
lives inside the other
d
The mealybug genome has acquired some bacterial
peptidoglycan (PG)-related genes
d
This insect-symbiont mosaic pathway produces a PG layer
at the innermost symbiont
d
Endosymbionts and organelles have evolved similar levels of
biochemical integration
Authors
DeAnna C. Bublitz, Grayson L. Chadwick,
John S. Magyar, ..., Pamela J. Bjorkman,
Victoria J. Orphan, John P. McCutcheon
Correspondence
john.mccutcheon@umontana.edu
In Brief
A functional biosynthetic pathway formed
from a combination of genes encoded by
a bacterial endosymbiont and its insect
host exhibits strong parallels to organelle
evolution.
Bublitz et al., 2019, Cell
179
, 703–712
October 17, 2019
ª
2019 The Author(s). Published by Elsevier Inc.
https://doi.org/10.1016/j.cell.2019.08.054
Article
Peptidoglycan Production
by an Insect-Bacterial Mosaic
DeAnna C. Bublitz,
1
Grayson L. Chadwick,
2
John S. Magyar,
2
Kelsi M. Sandoz,
3
Diane M. Brooks,
1
Ste
́
phane Mesnage,
4
Mark S. Ladinsky,
5
Arkadiy I. Garber,
1
Pamela J. Bjorkman,
5
Victoria J. Orphan,
2
and John P. McCutcheon
1,6,
*
1
Division of Biological Sciences, University of Montana, Missoula, MT 59812, USA
2
Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125, USA
3
Coxiella Pathogenesis Section, Laboratory of Bacteriology, Rocky Mountain Laboratories, National Institute of Allergy and Infectious
Diseases, NIH, Hamilton, MT 59840, USA
4
Department of Molecular Biology and Biotechnology, University of Sheffield, Sheffield S10 2TN, UK
5
Division of Biology and Biological Engineering, California Institute of Technology, Pasadena, CA 91125, USA
6
Lead Contact
*Correspondence:
john.mccutcheon@umontana.edu
https://doi.org/10.1016/j.cell.2019.08.054
SUMMARY
Peptidoglycan (PG) is a defining feature of bacteria,
involved in cell division, shape, and integrity. We
previously reported that several genes related to
PG biosynthesis were horizontally transferred from
bacteria to the nuclear genome of mealybugs. Mealy-
bugs are notable for containing a nested bacteria-
within-bacterium endosymbiotic structure in special-
ized insect cells, where one bacterium,
Moranella
,
lives in the cytoplasm of another bacterium,
Trem-
blaya
. Here we show that horizontally transferred
genes on the mealybug genome work together with
genes retained on the
Moranella
genome to produce
a PG layer exclusively at the
Moranella
cell periphery.
Furthermore, we show that an insect protein en-
coded by a horizontally transferred gene of bacterial
origin is transported into the
Moranella
cytoplasm.
These results provide a striking parallel to the genetic
and biochemical mosaicism found in organelles, and
prove that multiple horizontally transferred genes
can become integrated into a functional pathway
distributed between animal and bacterial endosym-
biont genomes.
INTRODUCTION
Horizontal gene transfer (HGT) occurs when a gene is moved from
the genome of one organism to another outside of the normal pro-
cesses of vertical inheritance. HGT can, in principle, occur be-
tween any two DNA-based life-forms, but most often involves
either movement of genes between microorganisms (
Koonin
et al., 2001; Ochman et al., 2000; Richards et al., 2011
)orfrommi-
croorganisms to larger eukaryotic hosts (
Dunning Hotopp, 2011;
Husnik and McCutcheon, 2018
). The process of HGT in the evolu-
tion of the cellular organelles derived from bacteria—the mito-
chondrion and the plastid—is not disputed and is often referred
to as endosymbiont gene transfer (EGT) when the transferred
genes seem to originate from the ancestral organelle genome
(
Keeling and Palmer, 2008; Martin et al., 2002; Timmis et al.,
2004
). The role that HGT (that is, transfer from sources other
than ancestral organelle genomes) has played in the evolution of
organelles is less clear, but numerous examples of HGTs from
bacteria unrelated to the mitochondrial or plastid ancestor are
foundineukaryoticgenomes (
Gray, 2015; Kuetal.,2015
).No mat-
ter their genome of origin, the proteins that are produced from
these EGTs and HGTs, and that function in organelles are trans-
ported there by specific multiprotein complexes (
Neupert and
Herrmann, 2007; Schleiff and Soll, 2000
). This history of gene
loss on organelle genomes and gene gain on nuclear genomes
has led to complex mosaic biochemical pathways in organelles,
where genes of diverse taxonomic origin reside on different ge-
nomes, and the protein products of these genes are shuttled to
differentpartsofthecellwithoutstrictdeferencetotheirtaxonomic
origins(
Duche
ˆ
neetal.,2005;Gabaldo
́
n,2018;Ko

reny
́
etal.,2013
).
Eukaryotic genome sequencing has led to the discovery of
many potential HGT candidates unrelated to organelle function,
most often originating from bacterial and fungal sources
(
Dunning Hotopp and Estes, 2014; Milner et al., 2019; Moran
and Jarvik, 2010; Scho
̈
nknecht et al., 2013; Slot and Rokas,
2011
). The roles of these HGTs are diverse, but most often
include nutrition or protection from predators, pathogens, and
environmental stress (
Husnik and McCutcheon, 2018
). The func-
tion of some of these HGTs has been verified (
Chou et al., 2015;
Dean et al., 2018; Kominek et al., 2019; Metcalf et al., 2014; Mil-
ner et al., 2019; Moran and Jarvik, 2010; Stairs et al., 2018
), but
these examples all involve single-step biochemical processes or
functions gained through the transfer of multiple genes linked by
residence on the same fragment of transferred DNA. These func-
tionally verified HGT events serve as important milestones in
HGT research, but none approach the complex cellular and
biochemical mosaicism observed in some organelle biochemical
pathways that result from EGT and HGT.
Genomic work on sap-feeding insects and their nutritional
endosymbiotic bacteria has revealed a few cases where the
complexity of bacterial integration into host cells seems to
approach that of organelles (
McCutcheon, 2016; McCutcheon
and Moran, 2011; Moran and Bennett, 2014
). These bacteria
Cell
179
, 703–712, October 17, 2019
ª
2019 The Author(s). Published by Elsevier Inc.
703
This is an open access article under the CC BY-NC-ND license (
http://creativecommons.org/licenses/by-nc-nd/4.0/
).
provide essential nutrients to their hosts and are thus required for
normal host biology and survival (
Akman Gu
̈
ndu
̈
z and Douglas,
2009; Baumann, 2005
). Many of these endosymbionts are also
long-term associates of their hosts, often living exclusively in
special insect cells for tens or hundreds of millions of years
(
Moran et al., 2005
). Like organelles, they are also faithfully
transmitted from one host generation to the next by maternal
transmission (
Koga et al., 2012
). This strict host association
has resulted in extreme levels of gene loss and genome reduc-
tion in some endosymbionts, leading to bacterial genomes that
are similar to organelle genomes in terms of gene number and
genome size (
McCutcheon and Moran, 2011; Moran and Ben-
nett, 2014
). In a final parallel to organelle evolution, some of
the insects harboring endosymbionts with tiny genomes appear
to use both native host genes and genes acquired from bacterial
HGTs to fill gaps in pathways formed by endosymbiont gene
loss (
Husnik et al., 2013; Luan et al., 2015; Nakabachi et al.,
2014; Nikoh and Nakabachi, 2009; Sloan et al., 2014
). However,
none of these putatively mosaic host-symbiont pathways have
been functionally verified.
One of the most complex of these insect-endosymbiont
relationships is found in
Planococcus citri
, an insect in a group
commonly referred to as mealybugs. The
P. citri
symbiosis exists
in an unusual structure (
Figure 1
A), where a gammaproteo-
bacterium (
Candidatus
Moranella endobia; hereafter
Moranella
)
resides in the cytoplasm of a betaproteobacterium (
Candidatus
Tremblaya princeps; hereafter
Tremblaya
), which together exist
in specialized insect cells called bacteriocytes (
von Dohlen
et al., 2001
). The
Tremblaya
genome is extremely small and
maintains only

120 protein coding genes, whereas the
Mora-
nella
genome is comparably larger but still encodes only

400
protein coding genes (
Lo
́
pez-Madrigal et al., 2013; McCutcheon
and von Dohlen, 2011
). Complementary patterns of gene loss
and retention on the
Tremblaya
and
Moranella
genomes suggest
that these two endosymbionts work together to make the essen-
tial nutrients required by their host insect (
McCutcheon and von
Dohlen, 2011
). Some of the genes missing on the endosymbiont
genomes are found on the insect nuclear genome, the result
of numerous HGTs from various bacteria unrelated to either
Tremblaya
or
Moranella
(
Husnik et al., 2013
).
One potential mosaic biochemical pathway in
P. citri
is that for
peptidoglycan (PG) biosynthesis. A PG layer is an almost univer-
sal feature of bacteria and is critical for bacterial cell division,
shape, and integrity (
Errington, 2013; Otten et al., 2018; Typas
et al., 2011
). Because PG is not produced by eukaryotes but is
a nearly ubiquitous feature of bacteria, it is a potent activator
of the eukaryotic innate immune response, and the highly
conserved PG biosynthetic pathway is the target of many antibi-
otics (
Dziarski, 2003; Otten et al., 2018; Wolf and Underhill,
2018
). The presence of numerous HGTs involved in PG synthesis
on the nuclear genome of an insect was therefore surprising.
Here we test whether mealybugs can use these PG-related
Figure 1. A Complete PG Biosynthesis Pathway Is Predicted by Genomics
(A) Schematic representation of a single
P. citri
bacteriocyte (blue), where
Moranella
cells and their two lipid bilayers (green) reside inside of triple-membrane-
bound
Tremblaya
cells (yellow).
(B) Adapted from
Typas et al., 2011
. A pictorial representation of the genes present and expressed on the genomes of
P. citri
(blue represents native eukaryotic
genes, brown represents HGTs of Gammaproteobacteria origin, orange represents HGTs of Bacteroidetes origin, and yellow represents Alphaproteoba
cteria
origin) and
Moranella
(green) that are involved in PG production. The locations of the small colored cell schematics labeled with gene names are based on the
predicted location of the protein product of that gene.
See also
Table S1
.
704
Cell
179
, 703–712, October 17, 2019
genes of diverse taxonomic origins and scattered genomic loca-
tions to produce a spatially and chemically coherent PG layer.
RESULTS
A Complete PG Pathway Is Predicted by Genomics
The synthesis of a complete PG layer in Gram-negative bacteria
can be divided into three classes of reactions based on enzyme
location (
Figure 1
B). The first set of reactions involves the
cytoplasmic synthesis of the PG precursor, a
b
-1,4-linked
N
-acetylglucosamine (GlcNAc)-
N
-acetylmuramic acid (MurNAc)
disaccharide linked to an L-alanine (L-Ala)-D-glutamic acid
(D-Glu)-
meso
-diaminopimelic acid (
m
Dap)-D-alanine (D-Ala)-D-
Ala pentapeptide. The second set of reactions involves inner
membrane-associated enzymes that flip this PG precursor into
the bacterial periplasm, and the third set of reactions involves
the cross-linking of precursor molecules into the growing PG
matrix, as well as PG modifications such as trimming the penta-
peptide to an L-Ala-D-Glu-
m
Dap-D-Ala tetrapeptide. Our orig-
inal annotation of the PG biosynthetic pathway of
P. citri
focused
on the first two sets of reactions because the
Moranella
genome
encoded some of the enzymes required to flip PG into its peri-
plasm, and because the bacterial-to-insect HGTs mostly
involved the first set of reactions (
Husnik et al., 2013
). But our first
annotation was not complete; there were unresolved holes in the
cytoplasmic part of the pathway, and we paid little attention to
the third part of PG synthesis.
In an effort to better understand the functional potential of this
mosaic PG biosynthetic pathway, we performed a detailed
reannotation of the PG-related genes in this system (
Figure 1
B;
Table S1
). Our expanded annotation indicates that a PG layer,
if produced, should be comprised of the canonical crosslinked
b
-1,4-linked GlcNAc-MurNAc disaccharide with an L-Ala-D-
Glu-
m
Dap-D-Ala tetrapeptide (
Figure 1
B;
Table S1
). Because
no genes related to PG biosynthesis are found in the
Tremblaya
genome, we predict that
Tremblaya
should not possess a
PG layer.
PG Constituent Parts Are Present in Whole Insect
Preparations
We used liquid chromatography coupled to tandem mass spec-
trometry (LC-MS/MS;
Bern et al., 2017
) to detect the presence of
PG in samples prepared from

1,500 pooled
P. citri
mealybugs.
Ions matching the expected
m/z
values for GlcNAc and MurNAc
monomers and dimers, some bound to a tetrapeptide chain
(L-Ala-D-Glu-
m
Dap-D-Ala), were identified in the first stage of
MS, and the structure of these molecules was confirmed by
tandem MS (
Figure 2
). To rule out the presence of contaminating
bacteria in mealybugs as a significant source of PG signal, we
performed five replicate 16S rRNA gene amplicon sequencing
runs on age-matched insects collected from the same mealybug
colony used for LC-MS/MS and a sixth sample taken from a
lysate used for MS analysis. On average

0.61% (range of
0.0604%–4.29%) of all amplicon reads were from bacteria other
than
Tremblaya
(76.1% of all reads) and
Moranella
(19.7% of all
reads) (
Table S2
). These results suggest that bacterial contami-
nation is an unlikely source of significant PG signal in our LC-MS/
MS experiments. Although these data allow us to conclude that
PG is present in the
P. citri
-
Tremblaya
-
Moranella
symbiosis, they
provide no information on the spatial localization of this putative
PG layer.
The PG-Specific Molecule D-Ala Is Specifically
Localized at the
Moranella
Periphery
We next attempted to visualize incorporation of D-Ala, a PG-spe-
cific molecule, in this symbiosis. Inspired by recent work using
localization of fluorescently labeled D-Ala as a proxy for PG
biosynthesis (
Kuru et al., 2012; Liechti et al., 2014; van Teeseling
et al., 2015
), we developed a similar approach based on nano-
meter-scale resolution secondary ion mass spectrometry (nano-
SIMS) (
Dekas et al., 2016; Dekas and Orphan, 2011
). We soaked
and injected an
15
N-labeled D-Ala solution onto sprouting
potatoes on which the mealybugs fed (
Figure S1
, related to
Fig-
ure 3
). After a week of feeding on labeled potatoes, mealybugs
were sacrificed, and the bacteriome tissue was removed and
thin-sectioned onto microscope slides. The location of insect
bacteriocytes,
Tremblaya
, and
Moranella
cells was first estab-
lished using fluorescence
in situ
hybridization (FISH) microscopy
(
Figures 3
A and 3F). These FISH-imaged tissue sections were
then subjected to nanoSIMS, where we observed strong, spe-
cific rings of enriched
15
N signal exclusively at or near the periph-
ery of all
Moranella
cells (
Figures 3
B and 3C–3E). In contrast,
mealybugs grown on
15
N-labeled L-Ala showed uniform label
incorporation throughout the bacteriome tissue, as expected
Figure 2. PG Constituent Parts Are Present
in Whole Insect Preparations
(A) MS/MS spectrum showing fragments of the
tetrapeptide disaccharide shown in (B) as anno-
tated by Byonic (
Bern et al., 2017
). The label
HexNAc indicates that GlcNAc cannot be formally
distinguished from the stereochemically identical
molecule
N
-acetyl-galactosamine. Pep+ 1+ rep-
resents the bare tetrapeptide L-Ala-D-Glu-
m
Dap-
D-Ala fragmented from PG glycans, and C
6
H
8
NO
2
and C
7
H
8
NO
2
represent rearrangements of
HexNAc after neutral losses such as H
2
O.
(B) Schematic of the reduced
b
-1-4 linked GlcNAc-
MurNAc
R
disaccharide attached to a tetrapeptide
stem made of L-Ala, D-Glu,
m
DAP, and D-Ala.
Cell
179
, 703–712, October 17, 2019
705
because L-Ala should be incorporated without bias into all
proteins in the symbiosis (
Figures 3
F and 3G). We verified a
Moranella
periphery-specific pattern of D-Ala enrichment using
bioorthogonal, Cu-catalyzed click-chemistry and fluorescence
microscopy (
Figure 3
H;
Kuru et al., 2012; Liechti et al., 2014;
van Teeseling et al., 2015
). The localization of enriched D-Ala
signal at the cell periphery of
Moranella
, but not
Tremblaya
or
the insect tissue, in both nanoSIMS and fluorescent microscopy
is consistent with our genomic prediction that PG production
should be specifically localized to the
Moranella
periphery, but
should not exist in
Tremblaya
(
Figure 1
B). Similar patterns of
D-Ala enrichment are routinely interpreted as strong evidence
for PG biosynthesis in bacteria (
Kuru et al., 2012; Liechti et al.,
2014; van Teeseling et al., 2015
).
A PG-Targeting Antibiotic Specifically Affects the
Moranella
Cell Envelope
Our nanoSIMS and click fluorescence data show that the
P. citri
PG layer is specifically located at the
Moranella
periph-
ery, but both nanoSIMS and light microscopy lack the spatial
resolution to place this PG signal precisely in the periplasm of
Moranella
. To provide additional evidence that the PG layer
we observe is located in the periplasm of
Moranella
,wefed
age-matched mealybugs on diets with and without the anti-
biotic cefsulodin, which specifically targets the periplasmic-
localized penicillin-binding protein, Pbp1B. The gene for
Pbp1B is encoded on the
Moranella
genome (
Figure 1
B;
Table
S1
). Pbp1B functions as a glycosyltransferase and transpepti-
dase, joining new GlcNAc-MurNAc-pentapeptide precursors
to the nascent PG matrix in the periplasm (
Cho et al., 2014;
Typas et al., 2011
). A common phenotype for bacterial cells
grown in hypotonic media in the presence of antibiotics is
membrane blebbing and cell lysis, although the exact re-
sponses vary considerably in different bacteria (
Chung et al.,
2009; Yao et al., 2012
). Given that
Moranella
lives exclusively
in the cytoplasm of another bacterium, and therefore likely in
an isotonic environment, we suspected that any antibiotic-
related phenotype we might observe would be subtle. Using
Figure 3. The PG-Specific Molecule D-Ala Is
Specifically Localized at the
Moranella
Pe-
riphery
(A) FISH imaging of a sectioned bacteriome from a
mealybug treated with
15
N D-Ala.
Tremblaya
cells
are green,
Moranella
cells are yellow-red, and the
insect nucleus is blue. Scale bar, 10
m
m.
(B) Reconstructed montage of a bacteriocyte from
multiple nanoSIMS images shown as a heatmap of
the fractional abundance of
15
N over
14
N[
15
N/
(
15
N+
14
N)] from the same mealybug section de-
picted in (A);
15
N enrichment is observed as yellow
rings around the edges of
Moranella
cells.
(C) Close-up of a single
Moranella
cell highlighted
in the red box in (B) shows enrichment of
15
N D-Ala
along
Moranella’s
periphery.
(D) Three-dimensional representation of
15
N D-Ala
enrichment of the
Moranella
cell shown in (C).
(E) A cross section through a portion of the bac-
teriocyte (black rectangle in B) reveals less
15
N
enrichment in either
Tremblaya
(green) or the in-
sect tissue (blue) as compared with
Moranella
(orange). The row labeled
15
F is the fractional
abundance of
15
N over
14
N[
15
N/(
15
N+
14
N)], the
14
N/
12
C row depicts the ratio of the abundant
natural isotope
14
N over the common isotope
12
C,
and the bottom row shows this section of tissue in
FISH microscopy.
(F) FISH imaging of a sectioned bacteriome from a
control mealybug treated with
15
N L-Ala.
Trem-
blaya
cells are red,
Moranella
cells are green, and
insect nuclei are blue.
(G) The fractional abundance of
15
N L-Ala as de-
tected by nanoSIMS from the portion of (F) outlined
in the pink box. The signal of
15
N L-Ala is nearly
uniform throughout the three organisms repre-
sented in this tissue, as expected for a molecule
that is uniformly incorporated into protein.
(H) Representative confocal image of Cu-cata-
lyzed click-chemistry to a D-Ala variant showing
enrichment at the
Moranella
periphery (green). In-
sect nuclei are stained with DAPI (blue). Image is
comprised of four merged slices from a z stack.
Scale bar, 10
m
m.
706
Cell
179
, 703–712, October 17, 2019
transmission electron microscopy (TEM), we found that the
periplasmic space of
Moranella
was specifically enlarged in
the presence of cefsulodin versus control animals (
Figures
4
A and 4C). Because
Tremblaya
has three membranes, we
also measured both the distance between the inner two and
from the innermost to the outer
most membranes as a control.
No significant changes in membrane spacing were observed
in
Tremblaya
between control and cefsulodin treatment (
Fig-
ures 4
A and 4B), indicating that this membrane spacing
phenotype is specific to
Moranella
and not simply due to
a general disruption of mealybug health in the presence of
antibiotics.
Figure 4. A PG-Targeting Antibiotic Specif-
ically Affects the
Moranella
Cell Envelope
(A) Quantification of the distance between the in-
ner- and outer-most membranes of
Tremblaya
(n =
200) and
Moranella
(n = 400) and the inner two
membranes of
Tremblaya
(n = 200) from control
and cefsulodin-treated insects; mean
±
SEM with
a jitter plot of all data points. All data points were
collected from random sections from two inde-
pendent biological replicates. There is a significant
difference only in the periplasmic space of
Moranella
, 11.5 versus 26.1 nm (Student’s t test)
with an effect size of 2.3. The effect sizes for
Tremblaya
’s inner-only and inner-to-outer mea-
surements were 0.0016 and 0.034, respectively.
(B and C) Representative TEM images of control
(B) and 100
m
g/mL cefsulodin-treated (C) insects
with
Tremblaya
(blue arrow) and
Moranella
(red
arrow) membranes visible.
See also
Table S3
.
A PG-Related HGT of
Alphaproteobacterial Origin Is
Localized to the
Moranella
Cytoplasm
Because most of the genes involved in the
cytoplasmic portion of PG synthesis exist
asHGTsontheinsectgenome(
Figure1
B),
it is formally possible that the GlcNAc-
MurNAc-pentapeptide PG precursor is
produced in the insect cytoplasm and
then transported into
Moranella
.Wesus-
pected, however, that it was more likely
that these HGTs were first translated in
the insect tissue and then transported as
proteins into the
Moranella
cytoplasm to
produce the PG precursor molecule. This
suspicion is primarily based on previous
results from another related insect, where
the protein product of an HGT on the pea
aphid genome was shown to be trans-
ported into the cytoplasm of its bacterial
endosymbiont,
Buchnera aphidicola
(
Na-
kabachi et al., 2014
). We note that if the
PG layer we report here is located in the
Moranella
periplasm, and if it is con-
structed in the
Moranella
cytoplasm as we predict in
Figure 1
B,
the proteins that result from HGT to the insect genome and that
function in the
Moranella
cytoplasm must cross five lipid bilayers
to get there:
Tremblaya
has three lipid bilayers, and
Moranella
has
two (
von Dohlen et al., 2001
)(
Figure 1
A).
In order to differentiate between these two scenarios for PG
precursor production, we produced a polyclonal antibody to a
predicted protein encoded by a bacteria-to-insect HGT and
observed its localization to test whether it was found in the
Moranella
cytoplasm. The antibody was generated against a
peptide fragment of MurF, a ligase that normally functions in
the bacterial cytoplasm to join D-Ala-D-Ala to the MurNAc-linked
Cell
179
, 703–712, October 17, 2019
707
tripeptide to produce the pentapeptide (
Otten et al., 2018; Typas
et al., 2011
). In the
P. citri
symbiosis, the
murF
gene exists only as
an HGT that was transferred from an alphaproteobacterium to
the host insect nuclear genome (
Figure 1
B;
Husnik et al.,
2013
). Immunohistochemistry on paraffin-embedded insect tis-
sue showed that MurF is found uniformly throughout the cyto-
plasm of
Moranella
and in portions of the insect tissue, but is
completely missing from the
Tremblaya
cytoplasm (
Figure 5
).
The restriction of signal to the insect and
Moranella
cytoplasm
is consistent with the hypothesis that the insect makes MurF
and transports it into
Moranella
as a protein but does not appre-
ciably accumulate in
Tremblaya
. This localization pattern sug-
gests that genes resulting from HGTs on the
P. citri
nuclear
genome can be made into proteins by host machinery and can
be transported across the three lipid bilayers of
Tremblaya
and
the two lipid bilayers of
Moranella
to end up in the
Moranella
cytoplasm. We note it is formally possible that the
murF
mRNA, not protein, is transported into
Moranella
and then trans-
lated by
Moranella
ribosomes. The data in
Figure 5
also suggest
that PG is made entirely in
Moranella
and that the alternative
scenario, where PG precursors are built in the insect cytoplasm
and transported into
Moranella
, is unlikely.
Collectively, our data from LC-MS/MS experiments (
Figure 2
),
nanoSIMS (
Figures 3
B–3E and 3G), fluorescence microscopy
(
Figure 3
H), antibiotic treatments (
Figure 4
), and immunohisto-
chemistry (
Figure 5
) allow us to conclude that the genetic mosaic
depicted in
Figure 1
B is functional and produces a PG layer that
is likely built in the
Moranella
cytoplasm and located in the
Moranella
periplasm.
DISCUSSION
A PG-based cell wall is an ancient and defining feature of bac-
teria. Not surprisingly, the highly conserved set of genes that
encode PG biosynthesis is normally exclusively found on bac-
terial genomes. Until now, there were two known exceptions
to this pattern, both related to cyanobacterial-derived endo-
symbionts of photosynthetic eukaryotes. The first example of
PG-related HGT comes from the ‘‘chromatophore’’ of the rhi-
zarian protist
Paulinella chromatophora
. EM suggests that the
Paulinella
chromatophore has a PG layer (
Kies, 1974
), which
is encoded primarily on the chr
omatophore genome with the
exception of one bacterial HGT to the host protist genome
(
Nowack et al., 2016
). The second example comes from the
group of photosynthetic eukaryotes whose ancestor formed
the original endosymbiosis with the cyanobacterium that
became the chloroplast. This group, called the Archaeplastida,
includes land plants, red algae, green algae, and glaucophyte
algae (
Lane and Archibald, 2008; McFadden, 2001
). Many arch-
aeplastidal nuclear genomes encode some PG-related EGT
and HGTs (
van Baren et al., 2016; Sato and Takano, 2017
),
but these genes do not always seem to work together to form
a functional PG layer at the chloroplast periphery. A chloro-
plast-localized PG layer has been verified using fluorescently
labeled D-Ala in the moss
Physcomitrella patens
(
Hirano
et al., 2016
), and possible chloroplast PG layers have been
observed by EM in glaucophytes (
Schenk, 1970
). But in the
land plant
Arabidopsis thaliana
, which retains some PG-related
genes on its nuclear genome, although fewer than in the moss
P. patens
, no PG layer exists at the chloroplast periphery and at
least one PG-related enzyme has been coopted for a different
function (
Garcia et al., 2008
). These results serve as a
cautionary note about inferrin
g function from the presence of
HGTs alone: gene presence is not a reliable predictor of biolog-
ical function (
Doolittle, 2013
).
Exactly how the
Moranella
PG layer is built, its function, and its
precise cellular location remain unknown. One important remain-
ing question is the source of D-Ala and D-Glu in
Moranella
’s PG.
Our original annotation left the activities of Alr (alanine racemase)
and MurI (glutamate racemase) unaccounted for: they did not
exist as HGTs on the insect genome, and they were not present
on the
Moranella
genome (
Husnik et al., 2013
). Our new annota-
tion confirms that homologs of these genes are missing in the
P. citri
symbiosis. Interestingly, GlyA and MetC have been shown
to moonlight as alanine racemases in
Chlamydia trachomatis
and
Escherichia coli
, respectively (
De Benedetti et al., 2014; Kang
et al., 2011; Otten et al., 2018
), and we find eukaryotic homologs
for these genes on the mealybug genome (
Table S1
). Similarly,
DapF has been shown to moonlight as a glutamate racemase
in
C. trachomatis
(
Liechti et al., 2018
), and this gene exists as
an HGT of alphaproteobacterial origin on the
P. citri
genome
(
Husnik et al., 2013
). These data suggest that the loss of
alr
in
Moranella
may be compensated by a moonlighting insect
enzyme, and that the loss of
dapF
on the
Moranella
genome
may be compensated for by a moonlighting alphaproteobacte-
rial
dapF
HGT (
Table S1
). But it is also possible that the source
of D-Ala and D-Glu is not from these putatively moonlighting en-
zymes at all, but rather from either the plant sap diet of the insect
or from D-amino acids in
P. citri
produced from normal insect
biochemistry. Although not studied extensively, D-amino acids
have been found in both plants (
Robinson, 1976
) and insects (
Au-
clair and Patton, 1950; Corrigan and Srinivasan, 1966; Corrigan,
Figure 5. MurF, a PG-Related HGT of Alphaproteobacterial Origin, Is
Localized to the
Moranella
Cytoplasm
Representative confocal image of a sectioned bacteriome stained with an anti-
MurF antibody (red). Insect nuclei are stained with Hoechst (blue). Signal is
detected inside of the
Moranella
cells and insect tissue, but not
Tremblaya.
Scale bar, 10
m
m.
708
Cell
179
, 703–712, October 17, 2019
1969
), although to our knowledge the levels of these compounds
have not been measured in
P. citri
. The source(s) of D-Ala and
D-Glu in
Moranella
could therefore be from the diet of the insect,
the insect’s native amino acid biochemistry, moonlighting en-
zymes of various origins, or from some combination of all of
these sources.
Our MurF immunohistochemistry localization data show that
the protein products of HGTs on the insect genome can be spe-
cifically targeted to the
Moranella
cytoplasm (
Figure 5
). Import of
enzymes (or mRNA) to the
Moranella
cytoplasm for precursor
production rather than producing the GlcNAc-MurNAc-penta-
peptide precursor in the insect cytoplasm may limit the risk to
the insect host of triggering a PG-based immune response.
Other insects with long-term endosymbionts devote resources
to scavenging PG fragments in order to prevent continuous im-
mune activation (
Maire et al., 2019
). By apparently sequestering
PG production to inside of
Moranella
,
P. citri
may avoid the need
for such contingency pathways, at least until
Moranella
cells are
recycled near the end of the mealybug’s life (
Kono et al., 2008
).
It is notable that most of the proteins made from PG-related
HGTs on the insect genome are predicted to function in the cyto-
plasmic part of PG synthesis, whereas the PG-related genes
retained by
Moranella
all code for inner membrane- or peri-
plasm-associated proteins (
Figure 1
B). These patterns of gene
loss, gene retention, and HGT suggest that genes encoding pro-
teins that function in the
Moranella
cytoplasm are more likely to
be successfully transferred to the host insect nucleus. It is
tempting to speculate that these HGT and protein localization
patterns reflect the (currently unknown) mechanism used by the
symbiosis to traffic proteins or RNAs made from the host genome
to the correct subcellular compartment. It may be that this traf-
ficking mechanism can transport soluble proteins and RNA, but
is unable to transfer membrane-associated proteins. Whatever
the trafficking mechanism, our data show that these proteins of
foreign genomic origin, now encoded on the insect genome,
work together with proteins still encoded on the
Moranella
genome to produce a genuine PG layer at the
Moranella
periphery. Future work will be required to definitively locate the
Moranella
PG layer because our current experiments lack the res-
olution to place this PG layer precisely in the
Moranella
periplasm.
Although the exact role that this PG layer plays in the
P. citri
symbiosis remains unknown, work from plastids suggests that
host takeover of endosymbiont PG production can be an
important step in the regulation of endosymbiont cell division
and potentially further integration with the host organism (
de
Vries and Gould, 2018
). In moss, knocking out a PG-related
HGT on the nuclear genome results in enlarged chloroplasts
(
Machida et al., 2006
), and treatment with various PG-target-
ing antibiotics results in fewer and larger chloroplasts per
host cell (
Katayama et al., 2003
). Together these data
suggest that the movement of PG-related genes from organ-
elle genome to the host is a way for hosts to regulate organelle
division (
de Vries and Gould, 2018; Katayama et al., 2003;
Machida et al., 2006
). In
P. citri
mealybugs,
Tremblaya
was ac-
quired before
Moranella
(
Hardy et al., 2008; Thao et al., 2002
),
and so the host insect must have found a way of controlling
Tremblaya
as the sole endosymbiont prior to the acquisition
of
Moranella
. Because the patterns of HGT and protein target-
ing we observe here are strongly convergent with moss chlo-
roplasts (
Garcia et al., 2008; Katayama et al., 2003; Machida
et al., 2006
), it is tempting to speculate that the function is
also convergent; that is, PG-related HGTs have been retained
on the insect genome as a way of controlling the cell division
of a bacterium that lives inside of another bacterium inside of
insect cells.
The frequency and importance of bacteria-to-eukaryote
HGT are a matter of debate (
Husnik and McCutcheon, 2018;
Martin, 2017; Martin and Herrmann, 1998
). Although in our
view numerous studies using genomic and transcriptomic
data strongly support the idea that bacteria-to-eukaryote HGT
is common in some groups of eukaryotes and is likely to be bio-
logically significant (
Husnik and McCutcheon, 2018
), functional
validation of most of these HGTs is lacking. In some cases,
HGTs involving single-step (or single operon) biochemical func-
tions have been experimentally validated by
in vitro
protein
expression, enzymatic assays, and/or
in situ
RNAi (
Chou et al.,
2015; Dean et al., 2018; Kominek et al., 2019; Metcalf et al.,
2014; Milner et al., 2019; Moran and Jarvik, 2010; Stairs et al.,
2018
). Although these examples serve as important milestones
in HGT research, none approaches the genetic complexity we
describe here, which is more akin to the host-organelle genetic
mosaic used by eukaryotes to build proto-heme (
Ko

reny
́
et al.,
2013; Obornı
́k
and Green, 2005
). Our data show that multi-
gene, multi-genome, and multi-cellular compartment conglom-
erations are not unique to organelles (
Booth and Doolittle,
2015a; McCutcheon and Keeling, 2014
; but see
Lane and Martin,
2015
and
Booth and Doolittle, 2015b
), and that cell biological,
genetic, and biochemical mosaics can become functional in ex-
amples outside of the mitochondrion and plastid.
STAR
+
METHODS
Detailed methods are provided in the online version of this paper
and include the following:
d
KEY RESOURCES TABLE
d
LEAD CONTACT AND MATERIALS AVAILABILITY
d
EXPERIMENTAL MODEL AND SUBJECT DETAILS
B
Mealybugs
d
METHOD DETAILS
B
Gene annotation
B
Mealybug feeding and dissection
B
LC-MS/MS
B
16S rRNA sequencing
B
FISH-nanoSIMS
B
Cu-click chemistry
B
TEM and dual-axis tomography
B
Immunohistochemistry
d
QUANTIFICATION AND STATISTICAL ANALYSIS
B
Membrane measurements
d
DATA AND CODE AVAILABILITY
SUPPLEMENTAL INFORMATION
Supplemental Information can be found online at
https://doi.org/10.1016/j.
cell.2019.08.054
.
Cell
179
, 703–712, October 17, 2019
709
ACKNOWLEDGMENTS
We thank Bil Clemons for helpful discussions about PG; Denghui David Xing of
the University of Montana Genomics Core for sequencing expertise; Carol
Garland, Matthew Hunt, and the Caltech Kavli Nanoscience Institute for aid
in maintaining the TF-30 electron microscope; and the Gordon and Betty
Moore and Beckman Foundations for gifts to Caltech to support electron
microscopy. PG Mass spectrometry analyses were performed by the biOMICS
Facility of the Faculty of Science Mass Spectrometry Centre at the University
of Sheffield. We thank Adelina E. Acosta-Martin and Ankur Patel for their help
with peptidoglycan analyses. This work was supported by the Gordon and
Betty Moore Foundation (GBMF5602), the National Aeronautics and Space
Administration Astrobiology Institute (NNA15BB04A), the National Science
Foundation (IOS-1553529), and the Biotechnology and Biological Sciences
Research Council (BB/N000951/1 and 2058718).
AUTHOR CONTRIBUTIONS
D.C.B.: conceptualization, investigation, analysis, methodology, validation,
visualization, and writing; G.L.C. and J.S.M.: investigation, methodology, anal-
ysis, validation, visualization, and writing; K.M.S.: conceptualization, analysis,
methodology, investigation, resources, and writing; S.M.: analysis, methodol-
ogy, resources, software, and writing; D.M.B.: investigation, methodology,
and visualization; M.S.L.: investigation, methodology, resources, validation,
and visualization; A.I.G.: data curation, analysis, investigation, software, and
visualization; P.J.B.: methodology, resources, and administration; V.J.O.:
conceptualization, funding acquisition, resources, and administration;
J.P.M.: conceptualization, funding acquisition, administration, resources,
visualization, and writing.
DECLARATION OF INTERESTS
The authors declare no competing interests.
Received: June 18, 2019
Revised: August 6, 2019
Accepted: August 28, 2019
Published: October 3, 2019
REFERENCES
Akman Gu
̈
ndu
̈
z, E., and Douglas, A.E. (2009). Symbiotic bacteria enable insect
to use a nutritionally inadequate diet. Proc. Biol. Sci.
276
, 987–991
.
Altschul, S.F., Gish, W., Miller, W., Myers, E.W., and Lipman, D.J. (1990). Basic
local alignment search tool. J. Mol. Biol.
215
, 403–410
.
Auclair, J.L., and Patton, R.L. (1950). On the occurrence of d-Alanine in the
haemolymph of the milkweed bug,
oncopeltus fasciatus
. Rev. Can. Biol.
9
, 3–8
.
Baumann, P. (2005). Biology bacteriocyte-associated endosymbionts of plant
sap-sucking insects. Annu. Rev. Microbiol.
59
, 155–189
.
Bern, M., Beniston, R., and Mesnage, S. (2017). Towards an automated anal-
ysis of bacterial peptidoglycan structure. Anal. Bioanal. Chem.
409
, 551–560
.
Bolger, A.M., Lohse, M., and Usadel, B. (2014). Trimmomatic: a flexible
trimmer for Illumina sequence data. Bioinformatics
30
, 2114–2120
.
Booth, A., and Doolittle, W.F. (2015a). Eukaryogenesis, how special really?
Proc. Natl. Acad. Sci. USA
112
, 10278–10285
.
Booth, A., and Doolittle, W.F. (2015b). Reply to Lane and Martin: Being and
becoming eukaryotes. Proc. Natl. Acad. Sci. USA
112
, E4824
.
Cho, H., Uehara, T., and Bernhardt, T.G. (2014). Beta-lactam antibiotics
induce a lethal malfunctioning of the bacterial cell wall synthesis machinery.
Cell
159
, 1300–1311
.
Chou, S., Daugherty, M.D., Peterson, S.B., Biboy, J., Yang, Y., Jutras, B.L.,
Fritz-Laylin, L.K., Ferrin, M.A., Harding, B.N., Jacobs-Wagner, C., et al.
(2015). Transferred interbacterial antagonism genes augment eukaryotic
innate immune function. Nature
518
, 98–101
.
Chung, H.S., Yao, Z., Goehring, N.W., Kishony, R., Beckwith, J., and Kahne, D.
(2009). Rapid
b
-lactam-induced lysis requires successful assembly of the cell
division machinery. Proc. Natl. Acad. Sci. USA
106
, 21872–21877
.
Cimino, M., Alamo, L., and Salazar, L. (2006). Permeabilization of the myco-
bacterial envelope for protein cytolocalization studies by immunofluorescence
microscopy. BMC Microbiol.
6
,35
.
Corrigan, J.J. (1969). D-amino acids in animals. Science
164
, 142–149
.
Corrigan, J.J., and Srinivasan, N.G. (1966). The occurrence of certain D-amino
acids in insects. Biochemistry
5
, 1185–1190
.
De Benedetti, S., Bu
̈
hl, H., Gaballah, A., Klo
̈
ckner, A., Otten, C., Schneider, T.,
Sahl, H.-G., and Henrichfreise, B. (2014). Characterization of serine hydroxy-
methyltransferase GlyA as a potential source of D-alanine in
Chlamydia pneu-
moniae
. Front. Cell. Infect. Microbiol.
4
,19
.
de Vries, J., and Gould, S.B. (2018). The monoplastidic bottleneck in algae and
plant evolution. J. Cell Sci.
131
, jcs203414
.
Dean, P., Sendra, K.M., Williams, T.A., Watson, A.K., Major, P., Nakjang, S.,
Kozhevnikova, E., Goldberg, A.V., Kunji, E.R.S., Hirt, R.P., and Embley, T.M.
(2018). Transporter gene acquisition and innovation in the evolution of Micro-
sporidia intracellular parasites. Nat. Commun.
9
, 1709
.
Dekas, A.E., and Orphan, V.J. (2011). Identification of diazotrophic microor-
ganisms in marine sediment via fluorescence
in situ
hybridization coupled to
nanoscale secondary ion mass spectrometry (FISH-NanoSIMS). Methods
Enzymol.
486
, 281–305
.
Dekas, A.E., Connon, S.A., Chadwick, G.L., Trembath-Reichert, E., and
Orphan, V.J. (2016). Activity and interactions of methane seep microorganisms
assessed by parallel transcription and FISH-NanoSIMS analyses. ISME J.
10
,
678–692
.
Doolittle, W.F. (2013). Is junk DNA bunk? A critique of ENCODE. Proc. Natl.
Acad. Sci. USA
110
, 5294–5300
.
Duche
ˆ
ne, A.-M., Giritch, A., Hoffmann, B., Cognat, V., Lancelin, D., Peeters,
N.M., Zaepfel, M., Mare
́
chal-Drouard, L., and Small, I.D. (2005). Dual targeting
is the rule for organellar aminoacyl-tRNA synthetases in
Arabidopsis thaliana
.
Proc. Natl. Acad. Sci. USA
102
, 16484–16489
.
Dunning Hotopp, J.C. (2011). Horizontal gene transfer between bacteria and
animals. Trends Genet.
27
, 157–163
.
Dunning Hotopp, J.C., and Estes, A.M. (2014). Biology wars: the eukaryotes
strike back. Cell Host Microbe
16
, 701–703
.
Dziarski, R. (2003). Recognition of bacterial peptidoglycan by the innate im-
mune system. Cell. Mol. Life Sci.
60
, 1793–1804
.
Errington, J. (2013). L-form bacteria, cell walls and the origins of life. Open Biol.
3
, 120143
.
Fuchs, B.M., Pernthaler, J., and Amann, R. (2007). Cell identification by fluo-
rescence in situ hybridization. In Methods for General and Molecular Microbi-
ology, Third Edition, C.A. Reddy, T.J. Beveridge, J.A. Breznak, G.A. Marzluf,
T.M. Schmidt, and L.R. Snyder, eds. (ASM Press), pp. 886–896
.
Gabaldo
́
n, T. (2018). Relative timing of mitochondrial endosymbiosis and the
‘‘pre-mitochondrial symbioses’’ hypothesis. IUBMB Life
70
, 1188–1196
.
Garcia, M., Myouga, F., Takechi, K., Sato, H., Nabeshima, K., Nagata, N., Ta-
kio, S., Shinozaki, K., and Takano, H. (2008). An Arabidopsis homolog of the
bacterial peptidoglycan synthesis enzyme MurE has an essential role in chlo-
roplast development. Plant J.
53
, 924–934
.
Gray, M.W. (2015). Mosaic nature of the mitochondrial proteome: Implications
for the origin and evolution of mitochondria. Proc. Natl. Acad. Sci. USA
112
,
10133–10138
.
Hardy, N.B., Gullan, P.J., and Hodgson, C.J. (2008). A subfamily-level classi-
fication of mealybugs (Hemiptera: Pseudococcidae) based on integrated mo-
lecular and morphological data. Syst. Entomol.
33
, 51–71
.
Hirano, T., Tanidokoro, K., Shimizu, Y., Kawarabayasi, Y., Ohshima, T., Sato,
M., Tadano, S., Ishikawa, H., Takio, S., Takechi, K., and Takano, H. (2016).
Moss chloroplasts are surrounded by a peptidoglycan wall containing
D-amino acids. Plant Cell
28
, 1521–1532
.
710
Cell
179
, 703–712, October 17, 2019