DISCUSSION
Despite the emerging evidence for
the biological importance of GLP-2 as a trophic hormone for the gut and
bones, very little structural information is available of the GLP-2R. An
increasing number of high-resolution structures of class B1 GPCRs, have
been published (Liang et al. 2018; Qiao et al. 2020; Wu et al. 2020;
Zhang et al. 2018; Zhang et al. 2017; Zhao et al. 2020), yet the
structure of the GLP-2R remains to be determined. However, a handful of
studies focusing on the GLP-2 structure and its interaction with the
GLP-2R have elucidated structural requirements for GLP-2’s interaction
with its receptor. In 2000, DaCambra et. al. performed an
Alanine(Ala)-scan within the DPP-4 resistant
h[Gly2]GLP-2(1-33) and showed reduced receptor
activation (cAMP accumulation), of the rGLP-2R by alterations in the
N-terminus part of the peptide (Dacambra et al. 2000). Here, Ala
replacement of the Histidine1 and the Asparticacid3 of hGLP-2, severely
reduced receptor activation with only modest changes in binding
affinity. These data demonstrate the importance of the GLP-2 N-terminus
for receptor activation, as also illustrated by the partial agonism (and
competitive antagonism) of GLP-2(3-33) (Thulesen et al. 2002) (figure
1). Ala-scan within the C-terminus part of
h[Gly2]GLP-2 severely reduced the binding affinity
demonstrating a central role of the C-terminus part for receptor
binding. In 2011, Venneti et. al. presented the first three-dimensional
solution structure of GLP-2 by nuclear magnetic resonance (NMR) (Venneti
et al. 2011). This structure supported the distinct roles of the N- and
C-terminus part of GLP-2 and revealed a stable alpha-helical
conformation at the central region (between Phe6 and Ile27) and a less
well-defined helical conformation in the C-terminus region. The binding
interface with the extracellular domain (ECD) of the receptor was
predicted to be between Leucine17 and Lysine30, while the N-terminus
part of GLP-2 from Histiine1 to Aspargine16 lacked contact with the
extracellular domains of the GLP-2R. The central roles of the N- and
C-terminus part of GLP-2 in respectively, receptor activation and
receptor binding were supported by Yamazaki et. al. in 2013 (Yamazaki et
al. 2013), showing a decreased intrinsic placental alkaline phosphatase
(PALP) activity (driven by cAMP) for GLP-2(3-33), (6-33) and (11- to
13-33). Most recently, Wisniewski et. al. replaced each residue in the
DPP-4 resistant [Gly2,Nle10]hGLP-2(1-30) analog with its
d-enantiomer in a systematic approach to gain insight into the GLP-2R
recognition revealing a loss of potency at position 5, 8, 9, 12, and 14
in the N-terminus, and similar loss for position 17-20, 25, and 29 in
the C-terminus (Wisniewski et al. 2016). Consistent with this, the
C-terminal of GLP-2 orientates towards a hydrophilic cavity in the NMR
structure (Venneti et al. 2011). Thus, the N-terminal part of GLP-2
plays a central role in receptor activation, while the C-terminus adopts
an alpha-helical conformation that plays a central role of receptor
binding of GLP-2 consistent with the suggested “two-step” activation
model of class B1 GPCRs, a model that is now much more refined (Liang et
al. 2018; Qiao et al. 2020; Wu et al. 2020; Zhang et al. 2018; Zhang et
al. 2017; Zhao et al. 2020).
The M10Y-modification barely changed the functional properties of the
two endogenous hGLP-2 variants, demonstrating, in agreement with the
model discussed above, that the Met10 of GLP-2 neither plays an
important role in ligand binding nor receptor activation. According to
the NMR structure, Met10 is positioned at the beginning of the
alpha-helix and is not part of the binding interface of the GLP-2R
(Venneti et al. 2011). Consistent with this, Wisniewski et. al. replaced
the oxidation and alkylation-prone Met residue at position 10 of hGLP-2
by the isosteric Nor-leucine (Nle) (Wisniewski et al. 2016). Met is
characterized by a sulfur atom in the sidechain, which is highly
sensitive to reactive oxygen species (ROS) that often changes structural
and functional properties of proteins (Black et al. 1991; Kim et al.
2014). ROS-mediated oxidation occurs by the addition of a single oxygen
molecule to the sulfur atom, forming methionine sulfoxide (MetSO) (Kim
et al. 2014), which creates a chiral center around the sulfur atom and
overall results in a stiffer and more polar side chain compared to the
unoxidized Met residue (Black et al. 1991). These changes can have
profound structural and functional consequences (Chao et al. 1997; Hoshi
et al. 2001; Gu et al. 2015; Sugamura et al. 2011). To protect for
oxidative damage of the Met in GLP-2 during the oxidative iodination,
and since Met is dispensable for GLP-2 function (Drucker et al. 2013;
Venneti et al. 2011; Wisniewski et al. 2016; Yamazaki et al. 2013), we
replaced Met10 with a Tyr residue. Thereby we created a target site for
oxidative iodination using [125I] in the full
agonist (GLP-2(1-33)) and in the antagonist and partial agonist
(GLP-2(3-33)). These modifications created the two peptides;
hGLP-2(1-33,M10Y) and hGLP-2(3-33,M10Y). These two M10Y-substituted
peptides acted as their wildtype counterparts, and with these, we were
in a unique position allowing us to investigate both agonist
[I125]-hGLP-2(1-33,M10Y) and antagonist
[I125]-hGLP-2(3-33,M10Y) binding.
The similar affinities (KD) support the main role of the
N-terminus in receptor activation and not in receptor binding (Couvineau
et al. 2011). Moreover, the higher Bmax for the antagonist follows the
general trend for more antagonist binding conformations versus agonist
conformations of GPCRs ( Rosenkilde et al. 1994). Interestingly, for the
first time among class B1 GPCRs, we describe the binding kinetics of a
peptide agonist in comparison with a peptide antagonist and show, that
the on-rate for the antagonist is significantly faster than for the
agonist. Binding kinetics parameters, including kon and
koff, have been highlighted to be more important in
describing a ligand’s in vivo efficacy and the onset of action,
than the classical parameters such as KD and
KI (Velden et al. 2020). The slower on-rate for the
agonist could reflect a more complex binding compared to the antagonist
in line with expected induction of active receptor states (Zhang et al.
2018). When comparing the apparent affinities for the agonists and the
antagonists obtained in competition with two radioligands, we observed
similar affinities irrespective of choice of radioligand. This suggests
that all four ligands (initially) interact similarly with the ECDs of
the hGLP-2R, and that the receptor easily interchanges between
(sequential) conformations induced by the agonist and the antagonist.
The location of the GLP-2R remains controversial in both rodent and
humans. It has been reported that the GLP-2R mRNA transcript and protein
is expressed in SEMFs (El-Jamal et al. 2014; Ørskov et al. 2005). Here
we confirm receptor expression at the protein level in SEMFs in the
intestine and in the pancreatic islet cells of mice by using the
agonistic radioligand [I125]-hGLP-2(1-33,M10Y).
The prevention of [125I]-hGLP-2(1-33,M10Y)
labeling by co-injection with excess amounts of unlabeled
hGLP-2(1-33,M10Y) demonstrates the specificity of
[125I]-hGLP-2(1-33,M10Y) binding. The strong
staining of the pancreatic islet cells by
[125I]-hGLP-2(1-33,M10Y) could result from GLP-2R
expression in the pancreatic islet cells, in agreement with what was
previously shown at the mRNA level (De Heer et al. 2007). Alternatively,
it could result from cross-interaction of GLP-2 with the GLP-1R, or a
combination of the two. The strong binding properties of both
radioligands to the mGLP-1R, and the low potency activation of the
hGLP-1R by GLP-2(1-33) and GLP-2(1-33, M10Y), demonstrate that the
pancreatic staining could be a result of GLP-1R binding. Also,
promiscuity is known within GPCRs, demonstrated by the activation of the
GIPR by GLP-2 (Skov-Jeppesen et al. 2019), the binding and activation of
both the GLP-1R and the GCGR by oxyntomodulin (Holst et al. 2018;
Jorgensen et al. 2007), and the activation of the mGLP-1R by glucagon
(Svendsen et al. 2018). Thus, cross-activation is a common phenomenon
within class B1 GPCRs, which is reflected in the high sequence
similarities observed among the receptors and across species. For rodent
GLP-2R’s, 81% and 79% sequence identities are found for the mGLP-2R
and rGLP-2R, respectively, explaining the high-affinity binding observed
for both radioligands to the rodent GLP-2R’s.
As we observed no binding of either hGLP-2 radioligand to the rGLP-2R,
future autoradiography studies in rats would eliminate the binding of
GLP-2 to the GLP-1R. Another possibility would be to use GLP-1R
knock-out (KO)-mice, or eliminate hGLP-1R binding by modifications of
GLP-2 at the C-terminus, as suggested recently in study where
replacement of position 11 and/or 16 of hGLP-2(1-30) eliminated hGLP-1R
actively, while retaining high hGLP-2R activity (Wisniewski et al.
2016). Thus, it is possible to decrease GLP-2 binding to the GLP-1R
without compromising the GLP-2R binding.