this locus, as well as the entireppe25-pe19
locus, noting the high sequence identity be-
tweenppe25andppe27genes (86.2%) and
pe18andpe19genes (78.4%) in this locus
(fig. S3, A and C). As shown in Fig. 3A, only
theDppe25-pe19andDpe19mutants showed
resistance to 3bMP1. The growth defect of
theDppe25-pe19mutant was also rescued by
loss of PDIM (fig. S3, B and D). PE18 and
PE19 are highly conserved (89% identity),
and extra-chromosomal expression of either
PE18 or PE19 restored the susceptibility of
theDppe25-pe19mutant to 3bMP1 and re-
scued the growth of theDppe25-pe19mu-
tant on glycerol (fig. S3D and table S3). It
is known thatpe18produces antisense trans-
cripts naturally ( 25 ), indicating that the ex-
pression of PE19 is negatively regulated by
the antisense expression ofpe18. Overex-
pression of the antisense transcript ofpe18in
wild-type H37Rv caused resistance to 3bMP1
(Fig. 3B), which is consistent with this regu-
latory role. To confirm the interaction be-
tween PPE51 and PE19 biochemically, we
coexpressed His-tagged PPE51 and HA-tagged
PE19 inM. tuberculosiscells and performed
a coimmunoprecipitation assay with the
cross-linker dithiobis(succinimidyl propionate)
(DSP). As expected, PE19 coprecipitated with
PPE51, and PE25 showed only slight copre-
cipitation with PPE51, likely because the
N-terminal half of PE25 is nearly 50% iden-
tical to PE19 (Fig. 3C). Because PE/PPE pro-
teins are usually exported as pairs by ESX
systems, we examined the export of PPE51
in theDpe19mutant and found that export
of PPE51 was blocked by the deletion ofpe19
(Fig. 3D).
We next asked whether this porin-like func-
tion was specific for the PE19/PPE51 pair or if
otherPE/PPEpairshadsimilarfunctions.The
genomic locus encoding PE20-PPE33 has been
shown to be up-regulated during Mg2+limita-
tion. This locus clustered with themgtCgene,
which is known to be involved in virulence and
is also up-regulated upon magnesium depriva-
tion ( 26 ). Furthermore, a magnesium riboswitch
(Mbox) was found upstream of thepe20-ppe33
locus ( 27 ), indicating that these proteins play a
role in magnesium homeostasis. We therefore
generatedDpe20andDppe31-ppe33mutants
inM. tuberculosisin both a PDIM-positive and
a PDIM-negative background (fig. S5, A and B,
and fig. S8A). As expected, both PDIM-positive
mutants showed a growth defect in Mg2+-
limiting media, especially at mildly acidic pH,
while the PDIM-negative mutants grew sim-
ilarly to wild type (Fig. 4, A to C, and figs. S6,
S7, and S8, D to F). The phenotype ofDppe31-
ppe33could be complemented by express-
ingppe31-ppe33,ppe31alone, ormspAfrom
M. smegmatis, but not by expressingppe32
orppe33(Fig. 4, A to C, and fig. S6). PE20
and PPE31 were also found to associate by
coprecipitation (fig. S8B). Together, these re-
sults suggest that PE20/PPE31 form a com-
plex, possibly with other proteins, to mediate
Mg2+transport across the outer membrane.PE20 was also found to interact with PPE33
(fig. S8B); however, PPE33 did not affect Mg2+
uptake (Fig. 4, A to C, and fig. S6). Further
studies will be needed to identify the substrate
transported by PE20/PPE33.
Finally, the genomic locippe25-pe19and
pe32-ppe65are known to be strongly up-
regulated by starvation for inorganic phos-
phate ( 28 ). First, we tested the growth of a
Dppe25-pe19mutant in defined media con-
taining various concentrations of phosphate.
TheDppe25-pe19mutants were impaired
slightly for growth at 100mM phosphate, and
at 30mM, growth was severely restricted (fig.
S9). We were unable to obtain a single knock-
out of thepe32-ppe65locus or the double
knockout of thepe32-ppe65locus in the
Dppe25-pe19background that retained PDIM.
Therefore, we knocked down the expression
ofpe32orppe65in theDppe25-pe19mutant
using CRISPR interference. As shown in Fig.
4D and fig. S10, silencing the expression of
pe32-ppe65inDppe25-pe19resulted in a pro-
nounced growth defect in phosphate-limiting
media, whereas the PDIM-negativeDppe25-
pe19/Dpe32-ppe65double knockout mutant
grew similarly to the wild type.
In slow-growingM. marinum, the ESX-5
secretion system is essential for PDIM outer-
membrane permeability, but the substrates
it transports were not identified ( 29 ). The
ESX-5 secretion system maps closely to the
ppe25-pe19gene cluster, and it seems likely
that the observed phenotype was the result ofWanget al.,Science 367 , 1147–1151 (2020) 6 March 2020 4of5
Fig. 4. PE20/PPE31 and
PE32/PPE65 are required
byM. tuberculosisfor
growth during Mg2+and
PO 32 −restriction.Growth of
M. tuberculosisstrains in
modified detergent-free
Sauton’s medium containing
10 mMMg2+at pH 7.0 (A),
pH 6.5 (B), and pH 6.2 (C).
Media were buffered
with 50 mM MOPS [3-(N-
morpholino)propanesulfonic
acid]. After growth in
24-well plates for 15 days
at 37°C, cells were treated
with 0.1% Tween 80 over-
night to prepare homoge-
neous cell suspensions, and
cell growth was quantified
by alamarBlue assay. Data
are from technical triplicates
corresponding to fig. S6
(P< 0.05, P< 0.01,
P< 0.001, #P< 0.0001,
unpairedttest). (D)M. tuberculosisrequiresppe25-pe19andpe32-ppe65for efficient growth during phosphate limitation. Growth ofM. tuberculosisstrains in
modified Sauton’s medium containing 30mM Pi buffered with 50 mM MOPS. A final concentration of 100 ng/ml aTC was used to knock down the expression of
pe32-ppe65in theppe25-pe19mutant. Data are generated from three independent experiments and shown as mean ± SD. RFU, relative fluorescence units.
010,00020,00030,000RFU010,00020,00030,000RFU010,00020,00030,000RFUABCDfadD26::618delA
Δppe31-ppe33
fadD26::618delA
/Δppe31-ppe33
Δppe31-ppe33+ppe31-ppe3 3
Δppe31-ppe33+ppe31
Δppe31-ppe33+ppe32
Δppe31-ppe33+ppe33
Δppe31-ppe33+mspA
fadD26::618delA
/Δppe31-ppe33+fadD26wt***
##
**** #*####
**####nsns# #
#OD6000.00.20.40.60.81.00 5 10 15 20 25
Dayswt
Δppe25-pe19+pe32 sgRNA (-aTC)
Δppe25-pe19+pe32 sgRNA (+aTC)
Δppe25-pe19+ppe65 sgRNA (-aTC)
Δppe25-pe19+ppe65 sgRNA (+aTC)
Δppe25-pe19
fadD26::618delA/Δppe25-pe19
/Δpe32-ppe65pH=7 pH=6.5pH=6.2RESEARCH | REPORT