A widespread proteinaceous sulfur storage compartment in bacteria

Robert Benisch

, Michael P. Andreas

, and Tobias W. Giessen

Program in Chemical Biology, University of Michigan, Ann Arbor, MI 48109, USA

Department of Biological Chemistry, University of Michigan Medical School, Ann Arbor, MI 48109, USA

*correspondence: tgiessen@umich.edu

Abstract

Intracellular compartmentalization is essential for all cells and enables the regulation and

optimization of metabolism

. One of the main functions of subcellular compartments is

the storage of nutrients

2-4

. As bacteria do generally not possess membrane-bound

organelles, they often have to rely on functionally analogous protein-based

compartments

2,5-7

. Encapsulin nanocompartments are one of the most prevalent protein-

based compartmentalization strategies found in prokaryotes

5,8

. Here we show that

desulfurase encapsulins represent a novel sulfur storage compartment in bacteria able to

sequester large amounts of crystalline elemental sulfur. We determined the 1.78 Å cryo-

EM structure of a 24 nm desulfurase-loaded encapsulin highlighting the molecular details

of the protein shell and desulfurase encapsulation. We found that elemental sulfur crystals

can be formed inside encapsulin shells in a desulfurase-dependent manner with L-

cysteine acting as the sulfur donor. Intracellular sulfur accumulation can be influenced by

the concentration and type of sulfur source in growth media. The selectively permeable

protein shell allows the long-term intracellular storage of redox-labile elemental sulfur by

excluding cellular reducing agents from its interior. We found that encapsulation

substantially improves desulfurase activity and stability while also preventing substrate

inhibition. These findings represent the first example of a dedicated and widespread

storage system for the essential element sulfur in bacteria and provide the basis for

understanding how this novel protein-based storage compartment is integrated within

bacterial metabolism.

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted June 22, 2023. ; https://doi.org/10.1101/2023.06.21.545984doi: bioRxiv preprint

Introduction

Subcellular compartmentalization is essential for all cells and enables the

regulation and optimization of metabolism

6,1

. This is not only true for large and complex

eukaryotic cells, but also for prokaryotes. In recent years, significant progress has been

made to highlight that bacterial cells are highly organized entities often relying on protein-

based strategies to coordinate and compartmentalize complex metabolic

functions

6,9,10,11,12,13

. One of these strategies are protein organelles and compartments

which represent nano-sized functional analogues of eukaryotic membrane organelles and

utilize semipermeable protein shells to sequester specific enzymes and processes. For

example, bacterial microcompartments (BMCs) sequester combinations of enzymes in

self-assembling protein shells and are involved in the anabolic fixation of carbon

14,15

and

catabolic processes like carbon and nitrogen source utilization

7,16

. Besides serving as

nanoscale reaction chambers, another important use of protein compartments is the

storage of nutrients

1,6,10,13

. The most widely distributed protein-based storage system is

ferritin, an 8-12 nm protein cage used by eukaryotic and prokaryotic cells to store iron

Many cells contain further systems for storing nutrients such as polyphosphate-

polyhydroxyalkanoate-

, and sulfur-storage granules or globules whose detailed

functions, compositions, and formation are still being debated

13,19

. In general, storage

compartments enable organisms to accumulate and retain high-value compounds for

later use when encountering changing, nutrient-limited, or stress conditions

17,18,20

A further and only recently discovered class of prokaryotic protein compartments

involved in storage and other functions are encapsulin nanocompartments

(encapsulins)

5,21

. Encapsulins consist of self-assembling protein shells sequestering

dedicated cargo enzymes and are among the most widespread protein compartments in

prokaryotes

5,8,22

. Cargo encapsulation is mediated by targeting sequences present at the

N- or C-terminus of all cargo proteins

5,21, 23,24

. Encapsulin shells possess icosahedral

symmetry with triangulation numbers of T=1 (60 subunits, 24 nm), T=3 (180 subunits,

32 nm), or T=4 (240 subunits, 42 nm) and an evolutionary connection with viral capsids

has been proposed

5,21,25,26,27

. Encapsulins are classified into four families based on

sequence similarity and operon organization, with Family 1 encapsulins having been

shown to be involved in iron storage, detoxification, and stress resistance

8,22,26-29

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted June 22, 2023. ; https://doi.org/10.1101/2023.06.21.545984doi: bioRxiv preprint

Bioinformatic analyses have further identified a novel widespread Family 2 encapsulin

system putatively involved in redox or sulfur metabolism

. A recent study in

Synechococcus elongatus confirmed that this Family 2A system is induced under sulfur

starvation conditions and encodes a cysteine desulfurase (CD) cargo protein sequestered

inside an encapsulin shell

. CDs are pyridoxal-5’-phosphate (PLP)-dependent enzymes

that catalyze the desulfurization of L-cysteine, yielding L -alanine and an enzyme-bound

persulfide intermediate

30,31

. It was found that desulfurase cargo loading is facilitated by

an N-terminal cargo-loading domain (CLD) and that desulfurase activity is increased upon

encapsulation

8,23

. So far, the molecular logic of CD encapsulation as well as the biological

function of this class of encapsulins are unknown.

Here, we present structural and biochemical data on a cysteine desulfurase

encapsulin system found in Acinetobacter baumannii 118362, a member of the

Acinetobacter calcoaceticus/baumannii complex. Using cryo-electron microscopy (cryo-

EM), we determine the 1.78 Å structure of the Family 2A encapsulin shell and report

evidence for a novel cargo-loading mechanism. We find that encapsulation increases CD

stability and notably enables high catalytic activity in the absence of a sulfur acceptor. CD

activity can lead to the mineralization of large amounts of crystalline elemental sulfur

inside the encapsulin shell which is protected from the reducing environment of the

cytosol. Together, our data suggest that desulfurase encapsulin systems represent a

novel and widespread intracellular sulfur storage system in bacteria.

Results

Cysteine desulfurase encapsulin operons are widespread in bacteria

All encapsulin shell proteins possess the HK97 phage-like fold, a widespread viral

capsid protein fold found in bacteriophages of the order Caudovirales and select

eukaryotic viruses including members of the Herpesviridae

. Recent structural and

phylogenetic analyses suggest an evolutionary relationship between encapsulins and

viruses

5,8

. It has been proposed that encapsulins originate from defective prophages

whose capsid protein has been co-opted by the prokaryotic cellular host to now serve its

own metabolic needs and increase its fitness. Recent sequence similarity and gene

neighborhood analyses allowed the classification of encapsulin systems into four

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted June 22, 2023. ; https://doi.org/10.1101/2023.06.21.545984doi: bioRxiv preprint

families

. So far, mostly Family 1 encapsulins have been characterized with only one

example of a Family 2 system – a CD encapsulin from S. elongatus

– having been

studied. Based on the absence or presence of a putative cNMP-binding insertion domain

in the encapsulin shell protein, Family 2 encapsulins can be further classified into Family

2A and 2B, respectively

5,8

Fig. 1: Distribution and diversity of cysteine desulfurase (CD) encapsulins. (a)

Phylogenetic tree of 1,462 Family 2A CD encapsulins highlighting their distribution in

bacterial phyla and operon type diversity. The outer ring color indicates bacterial phyla

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted June 22, 2023. ; https://doi.org/10.1101/2023.06.21.545984doi: bioRxiv preprint

distribution (see b) and the gray scale inner ring highlights operon type distribution (see

c). (b) Bacterial phyla encoding CD encapsulin operons and number of identified CD

encapsulin operons per phylum (left and middle). Distribution of the four identified operon

types (see c) in bacterial phyla (right). (c) The four identified operon organizations for CD

encapsulin systems and their prevalence. R: rhodanese, AT: L-serine O-

acetyltransferase, Enc: encapsulin shell protein, CD: cysteine desulfurase. (d) SDS-

PAGE analysis (left) and negative stain TEM micrograph (right) of the heterologously

expressed and purified Acinetobacter baumannii 118362 CD encapsulin shell. (e) SDS-

PAGE analysis (left) and negative stain TEM micrograph (right) of purified CD-loaded

Acinetobacter baumannii 118362 encapsulin resulting from the heterologous expression

of a four-gene operon. M: molecular weight marker.

Phylogenetic analysis of all identified Family 2A CD encapsulins revealed that

these systems are prevalent and widespread in bacteria, with 1,462 CD encapsulin

operons identified across 11 bacterial phyla (Fig. 1a,b and Supplementary Data 1). Most

CD encapsulins are present in Proteobacteria, Actinobacteria, Bacteroidetes, and

Cyanobacteria and many of these systems can be found in important model organisms

and pathogens including Mycobacterium leprae, Mycobacterium avium, Burkholderia

cepacia, Klebsiella pneumoniae, and Acinetobacter baumannii.

Closer analysis of CD encapsulin gene clusters identified four common operon

organizations (Fig. 1c). All operons code for an encapsulin shell protein and a CD. Often,

two additional co-regulated operon components can be present. These are annotated as

a rhodanese (R) and L-serine O-acetyltransferase (AT). Rhodaneses are a diverse class

of proteins with various functions, one of them being to serve as sulfur acceptor proteins

33-

. Sulfur acceptors generally directly interact with CDs to facilitate the transfer of the CD-

bound sulfur atom – intermittently stored as a persulfide intermediate – to a conserved

cysteine residue in the acceptor protein

. Sulfur acceptors can then distribute sulfur to

various downstream processes like iron-sulfur cluster assembly or thiocofactor

biosynthesis

30,31,35

. L-serine O-acetyltransferases are key enzymes in the biosynthesis of

L-cysteine, converting L-serine into O-acetyl-L-serine, the direct precursor of L-cysteine

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted June 22, 2023. ; https://doi.org/10.1101/2023.06.21.545984doi: bioRxiv preprint

Thus, gene annotation suggests that CD encapsulin operons are involved in sulfur

metabolism.

Heterologous expression of an Acinetobacter desulfurase encapsulin operon

identifies CD as the sole cargo protein

Here, we focus on a CD encapsulin operon found in a member of the Acinetobacter

calcoaceticus/baumannii complex (Acinetobacter baumannii 118362) encoding all four

operon components discussed above – rhodanese (J517_0525), L-serine O-

acetyltransferase (J517_0526), encapsulin shell protein (J517_0527), and CD

(J517_0528). Heterologous expression of the encapsulin shell gene or the complete

native four-gene operon in E. coli BL21 (DE3), followed by purification via polyethylene

glycol precipitation as well as size exclusion and ion exchange chromatography, yielded

readily assembled encapsulin nanocompartments as confirmed by negative stain

transmission electron microscopy (TEM) (Fig. 1d,e and Extended Data Fig. 1a).

Consistent with previous reports, protein shells appeared spherical with a diameter of ca.

24 nm, suggesting a T=1 shell assembly

5,23

. However, only when the four-gene operon

was expressed could a second major co-purifying band on SDS-PAGE be observed (Fig.

1e). Based on molecular weight and mass spectrometric analysis, this band was identified

as CD (71 kDa), indicating that CD represents the sole cargo protein of this Family 2A

encapsulin system. The purified sample was additionally subjected to native PAGE

analysis resulting in a major 1 MDa band and no lower molecular weight bands, further

confirming that CD is likely encapsulated inside the encapsulin shell (Extended Data Fig.

1b).

Single particle cryo-EM analysis of the desulfurase-loaded encapsulin

To gain molecular level insights into the structure and cargo-loading mechanism

of the CD-loaded encapsulin, single particle cryo-EM analysis was carried out (Extended

Data Fig. 2). The encapsulin shell was determined to 1.78 Å via icosahedral (I) refinement

(Extended Data Fig. 2a-c). This represents the highest resolution encapsulin shell

structure reported to date and allowed for accurate atomic model building (Extended Data

Fig. 2d,e). As suggested by negative stain TEM, the encapsulin shell was found to be

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted June 22, 2023. ; https://doi.org/10.1101/2023.06.21.545984doi: bioRxiv preprint

24 nm in diameter and to consist of 60 subunits, showing T=1 icosahedral symmetry (Fig.

2a). In contrast to Family 1 encapsulins, Family 2A encapsulins have been reported to

possess turret-like morphology at their 5-fold vertices, similar to many HK97-fold

Caudovirales capsids

5,23,32

. This is also the case for this Family 2A Acinetobacter

encapsulin where an extended C-terminus and an extra 12° backwards tilt of the shell

protein at the 5-fold symmetry axis results in turret-like vertices (Extended Data Fig. 3a,b).

The asymmetric unit contains a single shell protein subunit which exhibits the canonical

HK97 phage-like fold consisting of an A-domain (axial domain), P-domain (peripheral

domain), and E-loop (extended loop) (Fig. 2b)

. Whereas Family 1 encapsulins possess

an N-terminal helix located on the interior of the assembled shell

, this Family 2A

encapsulin contains an N-arm extension reminiscent of HK97-fold viruses

23,32

. The N-arm

interacts with neighboring subunits to form a chainmail-like topology (Extended Data Fig.

3c), often observed in HK97-fold viral capsids

, and two N-arms outline and mostly close

the pore found at the 2-fold symmetry axis (Extended Data Fig. 3d). The shell contains

differently sized pores at the 5-, 3-, and 2-fold axes of symmetry with likely only the 5-fold

pore being large enough (6 Å) for small molecule transmission to the compartment lumen

(Fig. 2c-e and Extended Data Fig. 3e). The exterior and narrowest point of the 5-fold pore

are mostly uncharged and non-polar, in contrast to the Family 2A S. elongatus 5-fold

Fig. 2: cryo-EM analysis of the encapsulin shell. (a) cryo-EM density resulting from

icosahedral (I) symmetry refinement of the CD-loaded encapsulin. The shell is 24 nm in

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted June 22, 2023. ; https://doi.org/10.1101/2023.06.21.545984doi: bioRxiv preprint

diameter and exhibits T=1 icosahedral symmetry. Cargo densities are not visible in the I

refinement. Shell density was colored radially from the center of the shell. One subunit is

highlighted in rainbow coloring from red (N-terminus) to blue (C-terminus). (b) A single

HK97-fold encapsulin subunit is shown in rainbow coloring highlighting the canonical A

(axial)-domain, P (peripheral)-domain, E (extended)-loop, and N (N-terminal)-arm. (c)

View from the shell exterior down the 5-fold symmetry axis highlighting the 5-fold pore.

Electrostatic coloring is shown. The narrowest point of the 5-fold pore is 6 Å wide. (d)

View from the shell exterior down the 3-fold symmetry axis highlighting the 3-fold pore.

Electrostatic coloring is shown. The narrowest point of the 3-fold pore is 4 Å wide. (e)

View from the shell exterior down the 2-fold symmetry axis highlighting the 2-fold pore.

Electrostatic coloring is shown. The narrowest point of the 2-fold pore is 2 Å wide.

pore, reported to be positively charged

. Pore size likely limits the range of molecules

able to enter and exit the compartment, as has been proposed for other encapsulin

systems

5,42

. The 6 Å 5-fold pore, however, is likely large enough to allow the substrate

(L-cysteine) and product (L-alanine) of the encapsulated CD to pass through, whereas the

2- and 3-fold pores are likely too restrictive.

Whereas icosahedral (I) refinement yielded the highest quality shell density, no

signal for internalized CD cargo could be observed. However, C1 refinement resulted in

a 2.18 Å volume where clear internal densities could be visualized (Fig. 3a). These

represent the encapsulated CD cargo with 12 low-resolution densities located below each

of the 12 pentameric vertices of the T=1 icosahedral shell. As has been reported for most

other encapsulins

5,27,43,44

, the observed CD cargo densities are substantially lower

resolution (~15 Å) than the shell, likely due to flexible tethering to the shell interior, with

only a limited number of CLD residues tightly interacting with the luminal surface (see

below). The flexible linker sequences are not visible in the cryo-EM reconstruction

resulting in the cargo densities appearing disconnected from the shell. CD belongs to the

class II SufS/CsdA-like desulfurases with almost all characterized members of this class

forming stable homodimers

31,45

. This is consistent with the observed size of the internal

cargo densities and SEC analysis of the unencapsulated Acinetobacter CD (Extended

Data Fig. 4). Thus, the maximal number of CD cargo proteins per encapsulin shell is likely

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted June 22, 2023. ; https://doi.org/10.1101/2023.06.21.545984doi: bioRxiv preprint

24, or 12 CD dimers which is in good agreement with cargo-loading estimates based on

SDS-PAGE gel densitometry analysis (Fig. 1e). Shell density subtraction followed by 2D

classification of shell-subtracted particles further confirmed the presence of cargo,

yielding 2D classes with clearly visible internal densities representing CD (Fig. 3b).

Fig. 3: cryo-EM analysis of the CD cargo inside the encapsulin shell. (a) Interior view

of the cryo-EM density resulting from an asymmetric (C1) refinement of the CD-loaded

encapsulin showing internal CD cargo densities (yellow). The shown volume is a

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted June 22, 2023. ; https://doi.org/10.1101/2023.06.21.545984doi: bioRxiv preprint

composite of the 2.18 Å shell density and a gaussian blurred (sdev 2, ChimeraX

46,47

) map

to highlight internal lower resolution CD cargo densities. Shell density was colored radially

from the center of the shell. (b) Shell-subtracted 2D classes of CD-loaded encapsulin

highlighting discrete internal CD densities. (c) Composite volume of the C1 high resolution

shell and a gaussian blurred map to highlight extra shell-associated densities (orange)

around the 3- and 5-fold symmetry axes likely belonging to the N-terminal cargo-loading

domain of the CD cargo. (d) A single encapsulin shell protein subunit (electrostatic

coloring) is shown with (top) or without (bottom) the closely associated non-shell density

(orange) likely representing the CD CLD to highlight their interaction at the shell protein

A-domain (5-fold) and P-domain (3-fold). The interaction surfaces are primarily uncharged

and non-polar (bottom, outlined in orange-black).

One unusual feature of CD is the presence of a ca. 225 residue long unannotated

and disordered N-terminal domain, in addition to the catalytic C-terminal desulfurase

domain. This domain is rich in proline, glycine, and serine and is not well conserved

among putative Family 2A CD cargo proteins with only 5 relatively short motifs found to

be partially conserved (Extended Data Fig. 5a)

8,23

. Previously, it was shown that a similar

domain in a Family 2A system from S. elongatus acts as the CLD responsible for

mediating cargo loading into the encapsulin shell and potentially interacts with the interior

shell surface close to the 3-fold pores

. In our cryo-EM analysis, we observed additional

non-shell densities along the interior surface of the encapsulin lumen, localized around

the 3-fold pores and A-domains (Fig. 3c). These densities likely represent parts of the N-

terminal CLD, however, resolution is too low for model building or sequence assignment.

The observed CLD densities are not connected and localized at mostly non-polar or

hydrophobic surface patches (Fig. 3d and Extended Data Fig. 5b). This suggests a model

of CLD-shell interaction where different parts of the N-terminal domain specifically interact

with conserved parts of the shell interior in a discontiguous way, connected by flexible

linker sequences, not visible in our cryo-EM density. The 5 conserved motifs found in the

CLD (Extended Data Fig. 5a) may represent the residues interacting with the shell while

the long stretches of less conserved sequence between them could serve as flexible

linkers. The identified surface patches at the 3-fold pores and A-domains are mostly

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted June 22, 2023. ; https://doi.org/10.1101/2023.06.21.545984doi: bioRxiv preprint