*WingMesh***: A Matlab-Based Application for Finite Element Modeling of Insect Wings**

**Shahab Eshghi 1,\* , Vahid Nooraeefar <sup>2</sup> , Abolfazl Darvizeh <sup>2</sup> , Stanislav N. Gorb <sup>1</sup> and Hamed Rajabi <sup>1</sup>**


Received: 20 July 2020; Accepted: 18 August 2020; Published: 18 August 2020

**Simple Summary:** Manual modeling of complicated insect wings presents considerable practical challenges. To overcome these challenges, therefore, we developed *WingMesh*. This is an application for simple yet precise automatic modeling of insect wings. Using a series of examples, we showed the performance of our application in practice. We expect *WingMesh* to be particularly useful in comparative studies, especially where the modeling of a large number of insect wings is required within a short time.

**Abstract:** The finite element (FE) method is one of the most widely used numerical techniques for the simulation of the mechanical behavior of engineering and biological objects. Although very efficient, the use of the FE method relies on the development of accurate models of the objects under consideration. The development of detailed FE models of often complex-shaped objects, however, can be a time-consuming and error-prone procedure in practice. Hence, many researchers aim to reach a compromise between the simplicity and accuracy of their developed models. In this study, we adapted *Distmesh2D*, a popular meshing tool, to develop a powerful application for the modeling of geometrically complex objects, such as insect wings. The use of the burning algorithm (BA) in digital image processing (DIP) enabled our method to automatically detect an arbitrary domain and its subdomains in a given image. This algorithm, in combination with the mesh generator *Distmesh2D*, was used to develop detailed FE models of both planar and out-of-plane (i.e., three-dimensionally corrugated) domains containing discontinuities and consisting of numerous subdomains. To easily implement the method, we developed an application using the Matlab App Designer. This application, called *WingMesh*, was particularly designed and applied for rapid numerical modeling of complicated insect wings but is also applicable for modeling purposes in the earth, engineering, mathematical, and physical sciences.

**Keywords:** biological structures; computer vision; mesh generation; simulation; digital image processing

#### **1. Introduction**

The finite element (FE) method is a numerical technique which is generally used to simulate a physical phenomenon in the virtual world by solving complex boundary value problems [1,2]. FE software packages were developed to simplify often complicated simulation processes. They are especially very common in engineering applications [3–5] and are becoming increasingly popular in the investigation of the mechanical behavior of biological structures, such as complex human and animal body parts [6–13].

Although providing a user with a high degree of flexibility, all available FE packages have a common need: an accurate model. A model is a domain which is subdivided into smaller polygonal or polyhedral meshes, so-called elements [14]. A modeling process, however, may present many challenges and can be rather time-consuming, especially when dealing with complex geometries [15–17], which is usually the case in biology. The skills of the software user can also strongly influence the process and the final result. These often lead to oversimplified models and, therefore, can affect the accuracy of simulation results. How can this problem be overcome?

In 2004, Persson and Strang aimed to address this problem by developing a simple meshing technique called *Distmesh2D* [18]. As intended by its developers, the method, which was implemented in Matlab code, provided an effective tool to mesh a given domain automatically. The simplicity of the method and the high quality of the produced mesh are the key advantages of the proposed method. However, it also has a major drawback: finding the distance to boundaries by the use of the mathematical equation *f*(*x*, *y*) = 0 or by values of a discrete set of points, as explained by the authors, is a time-consuming and error-prone procedure for complex geometries. Due to the use of a mathematical scheme to define the distance function (see Section 2.5), *Distmesh2D* also has limitations when meshing a domain containing discontinuities.

Here, we aimed to address these challenges and improve the performance of *Distmesh2D* but still maintain its simplicity. To this end, we used computer vision to automatically detect the boundary of a domain in a given image. We combined it with the mesh generator *Distmesh2D* to develop an application for the rapid modeling of geometrically complex domains that consist of several subdomains. The applicability of the method is not limited to in-plane domains, but it can also mesh out-of-plane (i.e., corrugated) objects. We specifically designed and used our method to develop models of insect wings. The proposed application, called *WingMesh*, draws extensively on Persson and Strang's account in an attempt to offer a simple, but more practical, meshing tool.

#### **2. Materials and Methods**

The modeling method presented in this study, called *WingMesh*, consists of several algorithms that interact with each other. The method requires an input image to identify the boundaries of a given domain as well as subdomains within that. The code *Distmesh2D* is then employed to mesh the identified domain. Other algorithms were added to the main algorithms to model out-of-plane domains and create a \**.inp* file, which is the Abaqus input file format [19]. The key parts of the developed Matlab code and the full code of the method, together with a graphical user interface (GUI) (see Section 3), are accessible in the Supplementary Materials (Codes S1 and S3, Method S1).

#### *2.1. Burning Algorithm for Detection of the Boundary of a Given Domain*

The burning algorithm (BA) was used to extract the boundary of a domain in a given image [20,21]. The BA needs a digital black and white image as the input, in which black pixels, with the pixel value of 0, represent the border of the domain and white pixels, with the pixel value of 1, represent regions that are situated inside and outside of the domain. The BA uses the matrix of the input image to detect the boundary of the domain in that image. This process starts with choosing a pixel within the domain by the user (Pixel 1 in Figure 1a) and continues by detecting white pixels around the selected pixel. To this end, the algorithm checks the colors of pixels located in the four orthogonal directions of the selected white pixels (Pixels 2 in Figure 1b). The coordinates of the found black pixels are stored in a matrix, and the colors of detected white and black pixels are changed to 0.8 (light grey) and 0.1 (dark grey), respectively, in order to avoid their reselection in the next iteration. This process continues by searching for white and black pixels around only white pixels detected in the previous iteration (Figure 1c–m). This process continues until all white pixels inside the domain are detected (Figure 1m). Code S2 and Video S1 in the Supplementary Materials are the source code of BA and a simple illustration of how it works, respectively.

*Insects* **2020**, *11*, x FOR PEER REVIEW 3 of 12

**Figure 1.** Detection of the border of an arbitrary domain using the Burning Algorithm. (**a**) A white pixel inside the domain is selected. (**b**) The BA searches for white and black pixels around the selected pixel in four orthogonal directions. (**c**) The BA searches for white and black pixels around the detected white pixels in the previous iteration. (**d**–**m**) This process continues until there is no white pixel inside the domain (**m**). **Figure 1.** Detection of the border of an arbitrary domain using the Burning Algorithm. (**a**) A white pixel inside the domain is selected. (**b**) The BA searches for white and black pixels around the selected pixel in four orthogonal directions. (**c**) The BA searches for white and black pixels around the detected white pixels in the previous iteration. (**d**–**m**) This process continues until there is no white pixel inside the domain (**m**).

#### *2.2. Detection of Subdomains within a Given Domain 2.2. Detection of Subdomains within a Given Domain*

subdomain.

The function BA can also detect subdomains within a given domain. When the main domain contains any subdomain, the application first finds the white pixels outside the domain. This process eventually results in the detection of the boundary of the domain. The function BA can also detect subdomains within a given domain. When the main domain contains any subdomain, the application first finds the white pixels outside the domain. This process eventually results in the detection of the boundary of the domain.

To find the boundary of each subdomain, the user should select a pixel inside that subdomain in the input image. By this, the BA finds the pixels located on the boundary of the subdomain using To find the boundary of each subdomain, the user should select a pixel inside that subdomain in the input image. By this, the BA finds the pixels located on the boundary of the subdomain using

the same method explained earlier. This process continues as long as the user selects a pixel in a new subdomain.

#### *2.3. Detection of Discontinuities in a Given Domain*

The function BA can detect any discontinuity, such as holes, cracks, etc., in a given domain. For this purpose, if any discontinuity exists, the user should select a pixel in each discontinuity in the input image (Video S2). After this, using the same method as described before, BA finds the boundary of each selected discontinuity. By this, the application detects discontinuities and excludes them from the main domain. To this end, after meshing the structure, the application finds all elements inside the region of discontinuity and excludes them from the model.

#### *2.4. Development of a Corrugated Model*

In a recent study, we developed a method for modeling out-of-plane (i.e., corrugated) domains [22]. Here, we modified this technique to make it more efficient and easier to implement. This technique requires an additional input image with the same frame size as the main input image. The other image should include information regarding the corrugation of the out-of-plane domain. The information should include the location of the maximum and minimum heights, indicated by the black and white colors, respectively. The value of pixels in the secondary image, therefore, serves as a measure of the height of that pixel: pixel values 0 and 1 indicate the maximum and minim heights, respectively. If there is more than one maximum or minimum height in a domain, any local extremum can be marked in grey color. The intensity of the grey color in each local extremum indicates the relative height of that extremum compared to the absolute extremum.

The recursive Equation (1) is used to smooth the corrugations to avoid any abrupt change in the height of a model at the location of an extremum.

$$v(r.c) = mean \left(\sum\_{i=-1}^{1} \sum\_{j=-1}^{1} v(r+i.c+j)\right) \tag{1}$$

where *r* and *c* represent the number of the row and column of a pixel in the image, respectively. *v* is the value of that pixel. *i* and *j* are the index of the row and column of the pixels in the image. This equation recursively updates the color intensity of the pixels in the secondary image and, thereby, the height of those pixels in the model. The number of iterations, which is set by the user, controls the sharpness of corrugations in the developed model. The values of pixels in the secondary input image, which are between 0 and 1, represent the relative heights of corrugations.

Figure 2a shows a corrugated object. The image of the object from the top view is shown in Figure 2b. Figure 2c shows the secondary input image, which has the same frame size as the image shown in Figure 2b. The black line in the middle of the image represents the position of the only available height maximum, and the white color corresponds to the regions with the minimum height. When using these two images, the application develops a model similar to that shown in Figure 2d. Figure 2j shows the gradual changes in the corrugation of the model by the use of Equation (1) after 20, 100, 150, 200, and 300 iterations (Figure 2e–i).

**Figure 2.** Modeling of an out-of-plane domain. (**a**) An out-of-plane domain. (**b**) A top view image of the domain. The image is used by the BA to detect the domain. (**c**) The secondary image contains a black line that represents the maximum height. The regions with zero height are colored in white. (**d**) A developed model based on the input images. (**e**–**i**) Smoothing the height of the meshed model using **Figure 2.** Modeling of an out-of-plane domain. (**a**) An out-of-plane domain. (**b**) A top view image of the domain. The image is used by the BA to detect the domain. (**c**) The secondary image contains a black line that represents the maximum height. The regions with zero height are colored in white. (**d**) A developed model based on the input images. (**e**–**i**) Smoothing the height of the meshed model using the iterative algorithm. (**j**) Changes in the corrugation pattern in different iterations.

#### the iterative algorithm. (**j**) Changes in the corrugation pattern in different iterations. *2.5. Mesh Generation*

Matlab code:

*2.5. Mesh Generation Distmesh2D* is a mesh generator in Matlab which employs a distance function, *d*(*x*,*y*), to describe the geometry of a domain [18]. The Delaunay algorithm is used in *Distmesh2D* to generate triangular meshes. The first line of the code *Distmesh2D*, the calling syntax, represents inputs and outputs of the *Distmesh2D* is a mesh generator in Matlab which employs a distance function, *d*(*x*,*y*), to describe the geometry of a domain [18]. The Delaunay algorithm is used in *Distmesh2D* to generate triangular meshes. The first line of the code *Distmesh2D*, the calling syntax, represents inputs and outputs of the Matlab code:

$$\text{function } \{p, t\} = dist \text{mesh} \land \text{2d}(\text{fd}\_{\mathcal{J}} \text{ft}, \text{h0}, \text{bbox}, \text{pfix})$$

*function* [*p*,*t*]*= distmesh2d*(*fd*,*fh*,*h0*,*bbox*,*pfix*) where the input arguments are as follows:


*Distmesh2D* produces the following outputs: *Distmesh2D* produces the following outputs:


Here, coordinates of the nodes on the boundary of a given domain, which are obtained by the BA, are used to define the distance functions *fd* and *fh* for the mesh generator *Distmesh2D*. In addition to the distance functions *fd* and *fh*, *Distmesh2D* has three other inputs: *h0*, *bbox*, and *pfix*. *h0*, the distance between initial nodes, can be set to 1, because the minimum distance between two pixels is 1. *bbox* is equal to the frame size of the imported image (size of the input matrix). The pixels located on the boundaries of the subdomains, extracted by the BA, are defined as fixed points, *pfix*.

*Distmesh2D* can generate both structured and unstructured elements. However, in this study, we set it to create only unstructured elements, because this type of element fits better with our aim for developing models of geometrically complex structures.

#### *2.6. Outputs*

*WingMesh* generates a \**.inp* file (i.e., an Abaqus input file) which contains information regarding the coordinates of the nodal points, their connections, type of elements, sections of the domain, and the material properties of sections. Detailed information about \**.inp* files can be found in the Supplementary Materials (Method S1).

#### **3. Graphical User Interface**

*WingMesh* was coded in Matlab 2019a, and Matlab App Designer was employed to develop a GUI. This GUI makes the method easy to implement and eliminates the need to know a programming language. The GUI is available in Code S3, and its description is available in Method S1.

#### **4. Examples**

• Example 1: An in-plane domain

Figure 3a shows a single in-plane domain with straight-line borders and sharp corners. Figure 3b shows the output model. The \**.inp* output file developed by the method is available in File S1.

• Example 2: An in-plane domain consisting of two subdomains

Figure 3c illustrates the same domain shown in Figure 3a, which is subdivided into two subdomains. As shown in Figure 3d, *WingMesh* was able to detect the border between the subdomains. The subdomains have meshed separately as two sections of a single model. The generated \**.inp* file is available in File S2.

• Example 3: An in-plane domain with subdomains and a discontinuity

We added a circular hole within one of the two subdomains of the domain given in the previous example (Figure 3e). After meshing all subdomains, including the discontinuity in the main domain, the elements generated in the discontinuity were removed before the final model was developed (Figure 3f). The \**.inp* file is available in File S3.

• Example 4: An irregular-shaped in-plane domain with several discontinuities

Figure 3g illustrates an irregular-shaped domain with curved borders, which contains four discontinuities. Previously, it was impossible to model such an irregular domain with complex-shaped discontinuities using the mesh generator *Distmesh2D*. However, the use of the DIP technique enables *WingMesh* to mesh such geometries. Figure 3h shows the meshed model developed based on the given domain. The \**.inp* file is available in File S4.

*Insects* **2020**, *11*, x FOR PEER REVIEW 7 of 12

**Figure 3.** Modeling of in-plane domains. (**a**) Image of a simple in-plane domain. (**b**) The meshed model developed from the image in (**a**). (**c**) Image of an in-plane domain consisting of two subdomains. (**d**) The meshed model developed from the image in (**c**). (**e**) Image of an in-plane domain with two subdomains and a discontinuity. (**f**) The meshed model developed from the image in (**e**). (**g**) Image of an irregular-shaped domain with several discontinuities. (**h**) The meshed model developed from the image in (**g**). (**i**) Image of a complex-shaped in-plane domain with several subdomains. (**j**) The meshed model developed from the image in (**i**). **Figure 3.** Modeling of in-plane domains. (**a**) Image of a simple in-plane domain. (**b**) The meshed model developed from the image in (**a**). (**c**) Image of an in-plane domain consisting of two subdomains. (**d**) The meshed model developed from the image in (**c**). (**e**) Image of an in-plane domain with two subdomains and a discontinuity. (**f**) The meshed model developed from the image in (**e**). (**g**) Image of an irregular-shaped domain with several discontinuities. (**h**) The meshed model developed from the image in (**g**). (**i**) Image of a complex-shaped in-plane domain with several subdomains. (**j**) The meshed model developed from the image in (**i**).

 Example *2:* An in-plane domain consisting of two subdomains • Example 5: A complex-shaped in-plane domain with several subdomains

Figure 3c illustrates the same domain shown in Figure 3a, which is subdivided into two subdomains. As shown in Figure 3d, *WingMesh* was able to detect the border between the subdomains. The subdomains have meshed separately as two sections of a single model. The generated \**.inp* file is available in File S2. Figure 3i shows the world map with the irregular shaped continents. The meshed model, which is apparently in good agreement with the given image, is presented in Figure 3j. The \**.inp* file is available in File S5.

• Example 6: An asymmetric out-of-plane domain with one height maximum and one height minimum

In this example and the next three cases, we used the domain shown in Figure 3a to develop out-of-plane models with different corrugation patterns. Here, we used the image in Figure 4a as the secondary input image to provide information on the corrugation spots. The grey color in this image indicates regions with zero height. The black and white lines indicate a height maximum and a height minimum, respectively. Figure 4b,c shows the perspective and side views of the meshed model. The \**.inp* file is available in File S6.

**Figure 4.** Modeling of out-of-plane domains. The use of different secondary images in combination with the same input image, as shown in Figure 3a, results in the development of models with different corrugated patterns. (**a**,**d**,**g**,**j**) Secondary images contain information on corrugation spots. (**b**,**e**,**h**,**k**) Perspective views of the meshed models created based on the image shown in Figure 3a and secondary images shown in Figure 4a,d,g,j. (**c**,**f**,**i**,**l**) Side views of meshed models.

• Example 7: An out-of-plane domain with two height maxima

Figure 4d shows an image with the black and dark grey lines, which represent two height extrema. Using this as a secondary image results in the development of the meshed model shown in Figure 4e,f. The \**.inp* file is available in File S7.

• Example 8: An out-of-plane domain with two height maxima and a height minimum

In Figure 4g, we added a tilted grey line to those in the secondary image shown in Figure 4a. The grey line is expected to change the corrugation pattern of the meshed model in Figure 4b by adding a region with a height maximum. Figure 4h,i shows the perspective and side views of the model developed using the secondary image in Figure 4g. The \**.inp* file is available in File S8.

• Example 9: An out-of-plane domain with circumferentially oriented height extrema model developed using the secondary image in Figure 4g. The \**.inp* file is available in File S8.

Example 8: An out-of-plane domain with two height maxima and a height minimum

secondary images shown in Figure 4a,d,g,j. (**c**,**f**,**i**,**l**) Side views of meshed models.

Example 7: An out-of-plane domain with two height maxima

Figure 4e,f. The \**.inp* file is available in File S7.

In this example, we aimed to test the precision of our method by developing a more complex corrugated domain. In this domain, the corrugation spot is circumferentially oriented, compared with the other domains that had longitudinal corrugations. Figure 4k,l shows the model developed by using Figure 4j. The \**.inp* file is available in File S9. Example 9: An out-of-plane domain with circumferentially oriented height extrema In this example, we aimed to test the precision of our method by developing a more complex corrugated domain. In this domain, the corrugation spot is circumferentially oriented, compared with the other domains that had longitudinal corrugations. Figure 4k,l shows the model developed by using Figure 4j. The \**.inp* file is available in File S9.

adding a region with a height maximum. Figure 4h,i shows the perspective and side views of the

*Insects* **2020**, *11*, x FOR PEER REVIEW 9 of 12

corrugated patterns. (**a**,**d**,**g**,**j**) Secondary images contain information on corrugation spots. (**b**,**e**,**h**,**k**) Perspective views of the meshed models created based on the image shown in Figure 3a and

Figure 4d shows an image with the black and dark grey lines, which represent two height extrema. Using this as a secondary image results in the development of the meshed model shown in

In Figure 4g, we added a tilted grey line to those in the secondary image shown in Figure 4a.

• Example 10: A beetle wing Example 10: A beetle wing

Figure 5a shows the hind wing of a beetle, *Allomyrina dichotoma* (Coleoptera: Scarabaeidae). Figure 5b shows the secondary image that was used for generating the corrugation on the model. The black lines show the location of elevated longitudinal veins in comparison with the membranes. The use of Figure 5b as a secondary input image results in the development of a model that is shown in Figure 5c,d from both dorsal and ventral sides. Figure 5a shows the hind wing of a beetle, *Allomyrina dichotoma* (Coleoptera: Scarabaeidae). Figure 5b shows the secondary image that was used for generating the corrugation on the model. The black lines show the location of elevated longitudinal veins in comparison with the membranes. The use of Figure 5b as a secondary input image results in the development of a model that is shown in Figure 5c,d from both dorsal and ventral sides.

**Figure 5.** Modeling of the hind wing of the beetle *Allomyrina dichotoma* (Coleoptera: Scarabaeidae). (**a**) Black and white image of the wing. (**b**) The secondary image for generating corrugations showing the location of the elevated veins. (**c**) Dorsal view of the generated model. (**d**) Ventral view of the generated model. **Figure 5.** Modeling of the hind wing of the beetle *Allomyrina dichotoma* (Coleoptera: Scarabaeidae). (**a**) Black and white image of the wing. (**b**) The secondary image for generating corrugations showing the location of the elevated veins. (**c**) Dorsal view of the generated model. (**d**) Ventral view of the generated model.

#### **5. Advantages of** *WingMesh* **5. Advantages of** *WingMesh*

*WingMesh* offers several advantages over existing manual modeling techniques using commercial software packages, such as CATIA, SolidWorks, Abaqus etc.: *WingMesh* offers several advantages over existing manual modeling techniques using commercial software packages, such as CATIA, SolidWorks, Abaqus etc.:


*WingMesh* has improved the applicability of *Distmesh2D*, as listed below:


#### **6. Applications**

The application presented in this study can be used for modeling a wide range of objects in both science and engineering, where a planar FE model is required. For example, models developed by our application could be used to understand the mechanical behavior of biological structures, such as insect wings, plant leaves, etc. (see [25] for more examples). In engineering, it can be employed for FE modeling of plate and shell structures used in aircraft, space crafts, ships, pressure vessels, etc. (see [26,27] for more examples). Our method could also be used in geology and geo-mechanics for the prediction of the mechanical response of complex inhomogeneous rock and concrete structures.

Although *WingMesh* is a promising first step towards the automatic modeling of insect wings, there still remain some other structural features that can be included in a wing model. A few examples of such features are nodus and vein joints, which play key roles in wing deformations both during flight [9,28,29] and at rest (i.e., wing folding [30–32]). Hence, as developers of *WingMesh*, we are currently working to develop the next generation of our program, which is able to create wing models with more structural details.

More information on *WingMesh* is available on our website: https://wingquest.org/wingmesh/.

**Supplementary Materials:** The following are available online at https://doi.org/10.6084/m9.figshare.12355163: Code S1: The source code of *WingMesh*. Code S2: The source code of the BA. Code S3: The GUI of *WingMesh*. Method S1: The description of the *WingMesh* code. Video S1: Finding the points inside a domain using the BA. Video S2: A tutorial about the use of *WingMesh*. File S1: The \**.inp* file of Example 1. File S2: The \**.inp* file of Example 2. File S3: The \**.inp* file of example 3. File S4: The \**.inp* file of example 4. File S5: The \**.inp* file of example 5. File S6: The \**.inp* file of example 6. File S7: The \**.inp* file of example 7. File S8: The \**.inp* file of example 8. File S9: The \**.inp* file of example 9. File S9: The \**.inp* file of example 9. File S10: The \**.inp* file of example 10.

**Author Contributions:** Conceptualization: S.E., V.N., A.D., S.N.G., H.R.; Data Curation: S.E.; Funding Acquisition: S.E., S.N.G., H.R.; Investigation: H.R., S.E.; Methodology: S.E., V.N., H.R.; Project administration: S.N.G., A.D., H.R.; Resources: S.N.G., A.D.; Software: S.E., V.N.; Supervision: S.N.G., H.R., A.D.; Validation: S.E., H.R., V.N.; Visualization: S.E.; Writing—original draft preparation: H.R., S.E.; Writing—review and editing: S.E., V.N., S.N.G., A.D., H.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the German Academic Exchange Service (DAAD) to S.E., grant number 57440921.

**Acknowledgments:** The authors are grateful to Peyman Mayeli (Monash University, Australia) for his valuable comments at the beginning of this study. We also want to thank Zeynab Ranjbar (Ahrar Institute of Technology and Higher Education, Iran) for her assistance regarding the preparation of images.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Nectar Feeding by a Honey Bee's Hairy Tongue: Morphology, Dynamics, and Energy-Saving Strategies**

**Hao Wang <sup>1</sup> , Zhigang Wu <sup>1</sup> , Jieliang Zhao <sup>2</sup> and Jianing Wu 1,\***


**Simple Summary:** This paper reviews the interdisciplinary research on nectar feeding behaviour of honey bees ranging from morphology, dynamics, and energy-saving strategies, which collects a range of knowledge of feeding physiology of honey bees and may inspire the design paradigms of next-generation multifunctional microfluidic transporters.

**Abstract:** Most flower-visiting insects have evolved highly specialized morphological structures to facilitate nectar feeding. As a typical pollinator, the honey bee has specialized mouth parts comprised of a pair of galeae, a pair of labial palpi, and a glossa, to feed on the nectar by the feeding modes of lapping or sucking. To extensively elucidate the mechanism of a bee's feeding, we should combine the investigations from glossa morphology, feeding behaviour, and mathematical models. This paper reviews the interdisciplinary research on nectar feeding behaviour of honey bees ranging from morphology, dynamics, and energy-saving strategies, which may not only reveal the mechanism of nectar feeding by honey bees but inspire engineered facilities for microfluidic transport.

**Keywords:** honey bee; mouth parts anatomy; nectar feeding behaviour; dynamics; energy-saving strategies

#### **1. Introduction**

The majority of flower-visiting insects, including bees, wasps [1,2], flies [3], butterflies [4], moths [5], and some beetles [6,7], obtain nutrition from floral nectar and pollen from flowering plants [8]. The honey bee (*Apis mellifera ligustica*) is a typical pollinator in the world [9]. The specialized proboscis is of great importance for a honey bee to load nectar rapidly and efficiently. The mouth parts of a honey bee are comprised of a pair of galeae, a pair of labial palpi, and a glossa [6]. The honey bee performs two feeding modes, namely lapping and suction [10]. While lapping, the honey bee drives its segmented tongue (glossa) coated in dense hairs back and forth to load nectar. When the honey bee dips nectar, the glossa protracts with the glossal hairs adhered to the glossa body. Then the glossa reaches to the maximum extension with the glossal hairs deployed. Next the brush-like glossa is filled with nectar and retracts to the mouth parts to load nectar. While sucking, the glossa extends out of the proboscis tube, directly sucking with the glossa keeping still [10].

Honey bees can feed on a range of viscous fluids at high efficiencies [8]. This behaviour is challenging because of the physical property of nectar, suggesting the nectar viscosity increases steeply with respect to the concentration, through which the glossa should have to resist high viscous drag [11–14]. In addition, if the glossa dips faster, the energetic intake rate will augment; however, the energy consumption caused by viscous drag will increase, so honey bees should have to meet the contradictive demands of both high energetic intake rate and low energetic loss while feeding on nectar. Investigations of the honey bee's feeding behaviour and related mechanical principles may reflect the health status of

**Citation:** Wang, H.; Wu, Z.; Zhao, J.; Wu, J. Nectar Feeding by a Honey Bee's Hairy Tongue: Morphology, Dynamics, and Energy-Saving Strategies. *Insects* **2021**, *12*, 762. https://doi.org/10.3390/ insects12090762

Academic Editor: Johan Billen

Received: 22 June 2021 Accepted: 18 August 2021 Published: 24 August 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

bees and adaptations to environmental constraints. More extensively, a healthier bee may consume less energy while feeding on nectar, who might be able to optimize the nectar harvest due to the mechanism of drag reduction. In addition, combined biological and mathematical analysis on feeding behaviour of bees may even elucidate the co-evolution between flowering plants and nectarivorous insects. In this review, we will introduce some interdisciplinary problems associated with honey bee's feeding behaviour. We will start with the anatomy of the mouth parts, followed by feeding modes, mechanism of hair erection, and energy saving strategies and conclude with potential engineering applications. The rest of the paper is structured as follows. Section 2 introduces the anatomical structure of the honey bee's mouth parts. Section 3 illustrates feeding fashion of a honey bee glossa from the perspectives of glossa kinematics and drag reduction mechanism, and Section 4 introduces the energy saving strategy by the glossa's dynamic surface. Functional compensation by regulating dipping frequency is shown in Section 5. Section 6 includes conclusions.

#### **2. Honey Bee Mouth Parts Morphology**

For the western bee (*Apis mellifera* L.), the mouth parts are comprised of a pair of galeae, a pair of labial palpi, and a hairy glossa, namely, the glossa in length of 2 mm (Figure 1a–c) [6,12]. Bushy glossal hairs in length of 100 µm, with diameter of 1~3 µm, are attached to the surface of the kidney-shaped sheath, which appear annulated on the glossa (Figure 1c–d). A thin membrane is attached to the edges of the sheath, which is also next to the corresponding sides of the muscular rod inside the glossa [15]. Notably the glossa has ~120 segments and is structured in a compliant manner. In the centre of glossa, a humour-filled cavity is formed by the sheath, muscular rod, and thin membranes. The honey bees are described to have two feeding modes, namely lapping and suction [10]. For the lapping mode, the glossa moves forward and backward with glossal hairs erecting rhythmically to load the nectar (Figure 1f). For the suction mode, the glossa stays still through the proboscis tube, and the nectar is sucked up by the cibarial pump, generating flows across the glossa surface [10]. μ μ

**Figure 1.** The honey bee's mouth parts. (**a**) A honey bee feeding on nectar on a flower. (**b**) The head and mouth parts of a honey bee. The mouth parts, highlighted in a red box, are comprised of a pair of galeae, a pair of labial palpi, and a glossa (*Apis mellifera* L.). (**c**) Scanning electron microscopic images of a bee glossa. (**d**) The glossa with bushy hairs. (**e**) The glossa observed under a microscope. (**f**) Lapping and sucking modes of a honey bee. The galeae and labial palpi form the probocid tube, then the glossa makes reciprocating movements through the tube to lap nectar [16].

#### **3. Feeding Behaviour of a Honey Bee Glossa**

#### *3.1. Section-Wise Wettability of the Glossa*

Wettability is the ability of surface to be wetted by liquid, which is determined by the balance of surface energy in the interface between air, liquid, and solid materials [17]. A honey bee propels its glossa to lap the viscous nectar, so the wettability of a honey bee glossa may be related to the nectar trapping capability [18], and the contact angle (wetting angle) is a measure of the wettability of a solid by a liquid. Generally, if the water contact angle is smaller than 90◦ , the solid surface is considered hydrophilic and inversely if the contact angle is bigger than 90◦ , the solid surface is considered hydrophobic. As a result, it is necessary to test the wettability of a honey bee glossa before we examine its feeding capability [19]. For bee flowers, the average nectar concentration in nature is 36% [20], so 25%, 35%, and 45% sucrose solutions were prepared for lab tests. Under a microscope, the contact angles measured in different glossal regions are shown in Figure 2. The results indicate that the contact angles turn smaller when using the sucrose solution with higher concentration, which insinuates that the surface exhibits stronger hydrophilicity to the thicker nectar. More extensively, the ranking of section-wise hydrophilicity suggests that the dorsal side is much easier to be wetted than the ventral side (Figure 2). We calculated the *p*-value as 0.03 between the two data sets of contact angles on the ventral surface and the dorsal surface, respectively, which suggests a significant difference between these two data sets, denoting that the dorsal surface is much easier to be wetted than the ventral surface for the reason of that the ranking of hydrophilicity, namely, D > A, E > B, and F > C. In order to better understand the comparison groups, we indicate them by stars shown in Figure 2. Moreover, the glossa tip is more hydrophilic than the middle region of the glossa, and the proximal part is the hardest to be wetted by the nectar. The section-wise wettability of the glossa may be caused by the chemical and geometrical differences on these hairy segments; the glossa surface is more hydrophilic to the higher-concentration nectar, which might be beneficial for nectar trapping, especially for the dynamic glossa surfaces. The combined chemical and geometrical differences on these hairy segments may contribute to a high flexibility in adaptation to varying environments, for instance, a broad range of liquid viscosities found in floral sources. In the next subsection, we will introduce the feeding pattern of honey bees via high-speed filming under the conditions of foraging on nectar with varying viscosities. .

**Figure 2.** Section-wise wettability of a honey bee's glossa. Six regions of the glossa immersed in nectar, marked with A~F, and A~C, and D~F, represent the ventral part and dorsal part respectively. Contact angles of different regions on the glossa surface of 25%, 35%, and 45% sucrose solution. The glossa surface is more hydrophilic to the higher-concentration nectar, which is elucidated from the decreasing of contact angles with the increased nectar concentration are shown in the histogram [21]. Asterisks indicates the comparison groups.

#### *3.2. Facultative Feeding Modes in Honey Bees*

The feeding pattern of a honey bee was previously defined as "lapping", which refers to reciprocating movements of its glossa entraining nectar by the glossal hairs. However, bees would also directly suck nectar with glossa keeping protracting and staying still (Figure 3a). Wei et al. [10] demonstrate that bees have facultative feeding behaviour. Bees prefer to suck the nectar with low sugar concentration and they tend to lap the nectar with more sugar content (Figure 3b). Further lab tests showed that honey bees can switch between these two feeding patterns to choose a more efficient ingesting mode (Figure 3c). The capillary-based lapping mechanism that allows honey bees to achieve high energy intake rates when feeding on highly concentrated nectar [22], while sucking directly with glossa protracting and staying still facilitates feeding on less viscous nectar (Figure 3d,e), besides the energy intake rate *E* ′ was calculated by *E* ′ = *ρscQ*/100, where *ρ* denotes the density of the nectar, *s* the sugar concentration, *c* the energy content per unit mass of sugar, and *Q* the nectar intake rate. Experiments validated that the key stimulus of choosing the ingesting technique is the viscosity of the nectar, rather than sugar content, according to the result that most bees feed on nectar with 10% sugar concentration, but with viscosity equivalent to 50% concentration (by adding Tylose) exhibited lapping pattern. This facultative drinking mode that is behaviourally adjusted to fluid viscosity has potentially enhanced the adaptability of honey bees to a wider range of nectar resources [23–25].

#### *3.3. Kinematics of the Glossa and Glossal Hairs*

For the lapping mode, the glossa extends out of the proboscid tube structured by the labial palpi and galeae, with glossal hairs attaching on the glossa body. Then the glossa moves back into the proboscis tube and glossal hairs flatten to offload the nectar. Here is a kinematic asymmetry, in which glossa protracts faster in a spear-like shape and retracts more slowly in a brush-like configuration. This asymmetry functions as a strategy to save energy, especially reducing the energetic consumption induced by viscous fluidic drag [26]. By observing kinematics of the glossa and glossal hairs by high-speed filming, Zhao [16] further found a specific asynchronization between glossa movements and glossal hair erection. A physical model is proposed to describe the feeding process considering the trade-off between nectar-intake volume and energy consumption. This asynchronization may be caused by the material properties of the elastic rod and the compliance of the segmented structures, especially the zig-zag shaped intersegmental membranes of the glossa [12], which is validated to be effective in maximizing the nectar-intake amount by theoretically figuring out the optimal moment when the glossal hairs begin to erect. This asynchronization suggests that the honey bee glossa can perform a scheduled coordination between glossa movements and hair erection, which could serve as valuable models for developing miniature pumps that are both extendable and have dynamic surfaces.

To uncover the anatomical mechanism of the coordination of glossa extension and hair erection, Zhu [27] compared hair erection and segment elongation and discovered a high consistency of their kinematics during the drinking process (Figure 4). In a dipping cycle, when the average erection angle of glossal hairs increases from 20 deg. to 38 deg., the average length of one glossal segment increases from 22.9 ± 1.6 µm to 24.7 ± 2.2 µm. The concordance equation was applied for evaluating correlation between these variables. The concordance measure is equal to 0.99 in the in vivo observation experiments, which shows that the average elongation of a glossal segment is closely correlated to the average erection angle of hairs.

′ ′ *ρ ρ*

**Figure 3.** Switchable feeding pattern in a honey bee. (**a**) High-speed images of a honey bee sucking the artificial nectar. (**b**) Occurrence rates of the two feeding modes in honey bees, when feeding on sucrose solutions with various concentrations [10]. (**c**) Occurrence rates of switching between feeding modes when offered extreme nectar concentrations, and the dotted lines represent binary feeding mechanisms in various nectar concentrations. Each encircled number represents a different individual. (**d**) Nectar intake rates of suction and lapping under different concentrations nectar, dashed line denotes the equivalent point of feeding efficiency and the corresponding sugar concentration, the dotted line denotes an equal nectar intake rate of the feeding modes under a specific nectar concentration. (**e**) Energy intake rates of suction and lapping under different nectar concentrations. Blue dashed line depicts the optimal concentration for suction mode, and green dashed line denotes that the optimal concentration for lapping mode is around 50% or above.

**Figure 4.** Morphological changes in glossal surfaces during dipping nectar and surface configurations through stretching the honey bees' glossae. (**a**) Dipping pattern of a honey bee tongue. (**b**) Asynchronization between tongue displacement and average hair erection angle. Both the in vivo and postmortem observations reveal that shortening and lengthening of the glossal segments is highly coordinated with the erection of glossal hairs, which aids in developing deformable gaps between rows of glossal hairs during nectar trapping [16].

**Figure 5.** Coordinated movements of hair erection and segment elongation of the glossa. (**a**) Natural behaviour of nectar drinking. (**b**) Highly-coordinated movements of the glossal segments and hair erection through stretching the glossa under a microscope [27].

#### *3.4. Coordinated Movements of the Abdomen While Dipping Nectar*

As a lapper, honey bee uses a mop-like glossa to trap nectar from flowering plants. By filming the feeding honey bees, a significant increase in abdominal pumping frequency was observed when honey bees drink the sucrose solution [16]. Zhao [16] combined high-speed filming, X-ray phase contrast imaging, and mathematical models to investigate the effect of abdominal pumping in liquid feeding of honey bee. A honey bee performs abdominal pumping during feeding, which is in concordance with reciprocating movements of the glossa (Figure 6). The modelling framework demonstrates that the abdominal pumping powers the honey bee's feeding efficiency and saves foraging time. The combined experimental and theoretical investigations extend the knowledge about the function of abdominal movements, which is considered only for adjusting flight attitude or crawling through honeycombs [28]. This behaviour is functionally analogous to power suction feeding in some fish that uses most power of axial swimming muscles not only by the cranial muscles [29]. The multifunctional use of muscular actuations fulfils the switchable requirements of these animals and makes the organs structurally compacted and efficient.

**Figure 6.** Dependence of abdominal and glossal movements (three samples are shown here). Both the glossa and abdomen protracted and retracted periodically, the pink arrow denotes the direction of protraction and retraction of the glossa and abdomen, thereby showing an approximate sinusoidal principle. The left vertical axis shows the volume of abdomen when lapping nectar against time, and the right vertical axis shows the real-time length of the glossa [15].

7

1

*ω* ≤ ≤

cos sin

μ

*ρ ρ*

1 1 0

*μ μ*

#### **4. Energy Saving Strategy**

For the viscous dipping mode, a honey bee should have to meet the combined requirement of both high energetic intake rate and low energetic dissipation caused by viscous drag. A honey bee may have to make millions of reciprocating movements during its whole life, so an energy-saving mechanism may be required to reduce the energy consumption and lower the possibility of wear caused by the viscous drag. This section includes some interdisciplinary work that covers morphology, high-speed imaging, and lubrication models, to uncover the energy saving strategy while feeding on nectar [30].

#### *4.1. Modelling for Energy Saving*

Figure 7 shows the actual glossa kinematics and hypothetical cases with various kinematic apportionment. A 7-order Fourier function that fits the actual glossa velocity in a dipping cycle (*R* <sup>2</sup> = 0.9853) is shown as

$$u(t) = K\_1 f(t) = K\_1 \left( a\_0 + \sum\_{i=1}^7 (a\_i \cos(\omega t) + b\_i \sin(\omega t)) \right) \tag{1}$$

where *f*(*t*) is the fitting equation; *a*0, *ω*, *a*<sup>i</sup> and *b*<sup>i</sup> (1 ≤ *i* ≤ 7) are the parameters calculated by Matlab to obtain the best fit for the scatter plot; and *K*<sup>1</sup> (137 µm/cm) is a coefficient that links the sizes in high-speed photographs into those for an actual honey bee. It can be demonstrated that the honey bee can reduce its energy expenditure using the derived protraction kinematics. The power required for resisting viscous drag can be estimated as *P*v~*µLu*<sup>2</sup> , where *µ* is the nectar viscosity and *L* is the glossa length during protraction, and the power to drive glossa can be estimated as *P*t~*mu*'*u*~*ρ*t*a* 2*u* 3 , where *a* and *ρ*<sup>t</sup> are the radius of the glossa and the density of the glossa, respectively. Since the ratio *P*t/*P*<sup>v</sup> ≪ 1, the effect of *P*<sup>t</sup> can be neglected [26]. The viscous drag can be written as *F*<sup>v</sup> ∝ *µ*·(2*πa*)·*x*(*t*)·*u*(*t*)= *K*2*µa*·*x*(*t*)·*u*(*t*), where *K*<sup>2</sup> is a proportionality coefficient. Combining Equation (1) and the formula *x*(*t*) = R *<sup>t</sup>* 0 *f*(*t*)d*t*, the power needed to overcome viscous drag reads.

$$P\_{\mathbf{V}}(t) = F\_{\mathbf{V}} \cdot \boldsymbol{\omega}(t) = \mu a \mathbf{K}\_2 \mathbf{K}\_1^3 f(t)^2 \int\_0^t f(t) \, \mathbf{d}t \tag{2}$$

One can evaluate the benefits of keeping the glossal hairs still during tongue protraction from Equation (1). If the honey bee erects the hairs, the outer diameter of the glossa radius will augment from *a* to (*a* + *h*cos*θ*), which will lead to a significant increase in *P*v(*t*). Scanning electron microscope imaging indicates *a* ≈ 50 µm and *h* ≈ 170 µm, and since *θ* ≈ 45◦ , we then arrive at (*a* + *h*cos*θ*)/*a* = 3.4, which means that hair erection increases the resistance by more than three times. Therefore, the honey bee is equipped with a specific glossal hair erection pattern for energy saving, where the flatten hairs can reduce viscous drag during protraction, whereas the hairs erect to trap more nectar in a single cycle during retraction. When a glossa makes reciprocating movements through viscous fluid, the viscous drag will exert on the hairy glossa surface. The nectar is a specific solution, the physical property of which is analogous to the sucrose solution, and its viscosity rises steeply with respect to concentration. From the perspective of Fluid Mechanics, the viscous drag dissipates energy so we should have to consider the energy dissipation linked to viscous drag. Some previous tests were made to validate the fact that more viscous nectar causes higher rate of wear, which indicates the viscous drag can accelerate the structural deterioration on the seemingly fragile glossal hairs.

ሺ ሻ ሺ ሻ *μ* ሺ ሻ න ሺ ሻ

≪

∝

**Figure 7.** Energy saving by specific glossa kinematics. The scattered points show three independent protraction velocities measured from the high-speed video and the light blue area indicates the error band of the velocities. The bold blue curve represents the Fourier kinematics *u*<sup>1</sup> , which fit the scatter plot well. The bold red curve shows the constant-acceleration-and-deceleration (CAaD) kinematics *u*<sup>2</sup> . The dotted blue line *P*<sup>1</sup> and the dotted red line *P*<sup>2</sup> indicate the protraction power under the fitted kinematics and constant-acceleration-and-deceleration kinematics, respectively [26].

#### *4.2. Effects of Galea Ridges on Drag Reduction*

*μ* ሺ π ሻ ሺ ሻ ሺ ሻ *μ*

*θ* ≈ μ ≈ μ *θ* ≈ *θ* Biological surfaces with unique microstructures in nature may perform specific functions, such as impact absorption and drag reduction in dung beetles or sharks, respectively [31]. The honey bee, *Apis mellifera* L., dips viscous nectar at a high rate which is about 5 Hz by the glossa, which causes non-negligible fluidic drag that results in structural and functional deterioration. By postmortem examination, Li [32] found the ridges are parallel distributed on the inner wall of the galeae and validated its effects on drag reduction. Li then compared the structural discrepancy between workers and drones and proposed some implications about the caste-related behaviour [32].

Scanning electron microscopy (SEM) images indicate that the honey bee galea has internally transverse ridges uniformly distributed (Figure 8). Theoretical analysis show that the ridges on the galeae of honey bee's mouth parts of workers can reduce the friction coefficient by 86%. Li [32] then examined the dimensional diversities of the uniformly-distributed micro-ridges on inner walls of galeae among workers and drones of *Apis mellifera* L. The hydrodynamic model was used to calculate the friction coefficient in the mouth parts, further testing whether the sexually-dimensional variations of the micro-ridges could influence the effect on drag reduction. Theoretical estimations of the friction coefficient with respect to the dipping frequency show that the inner micro-ridges can significantly reduce friction during the feeding process of a honey bee. Li then compared effects of drag reduction regulated by the sexually-selected micro-ridges and demonstrated that the hydrodynamic coefficients of workers and drones are 0.011 ± 0.007 and 0.045 ± 0.010 respectively, which indicates that workers exhibit better capability of drag reduction in their mouth parts than that of drones. This discrepancy may have some more indications in caste-related work of honey bees. The main physiological requirement of drones is to find an airborne queen to mate and accordingly, so drones exhibit strong adaptations to

forceful flying, and drones possess elaborate mating organs and powerful sense organs, such as big eyes and long antennae with many receptors for visual and olfactory orientations toward airborne queens [33]. Thus, although drones have bigger bodies, their mandibles are shorter, and their stomachs for honey storage are slimmer than those of workers [34]. Compared to drones, workers should have to fulfil a variety of tasks [35]. Workers tidy the hive, care the brood, nourish the larvae, drones, and the queen, and work for nest homeostasis [36,37]. Given these various duties, workers are equipped with well-developed hypopharyngeal and possess longer mouth parts than drones. Notably, adult drones are nourished by worker-prepared food, and their feeding ability is weaker than that of workers [34]. This experimental and theoretical combined research elucidated that the sexually-selected micro-ridges, developed inside workers and drones of honey bees' mouth parts, are structurally adapted to meet the demands of caste-related laborers of honey bees.

ξ **Figure 8.** The friction coefficient against the heights of the micro-ridges on the inner wall of the galeae of workers and drones. The blue square and red dot represent the friction coefficient against the heights of the microridges on the inner wall of the galeae of workers and drones, respectively. Here *a* and *b* denote the dimensions of the microridges of different castes of honey bees, in which *a* is the length of the galea and *b* is the ridges, and ξ denotes the average dimensions of the workers and drones. The dotted lines illustrate the measured mean height of the microridges on the galeae of the workers was 3.98 µm, whereas that of the drones was 3.15 µm [32].

#### μ μ **5. Functional Compensation by Regulating Dipping Frequency**

μ <sup>−</sup> Because of the highly-intensive viscous drag exerting on the glossa during nectar feeding, the glossal setae tend to wear out in the high-viscosity nectar. However, bees at varying day ages can maintain the nectar intake rate at 0.39 ± 0.03 µg·s −1 (35% nectar). Shi found that the average glossal setae length decreases with respect to age from 17 to 25 days, and it degrades even faster when fed with higher-viscosity nectar. Lab tests indicated that the older honey bees with short setae dip nectar more quickly. Moreover, a correlation between dipping frequency *f* and the average glossal setae length *h*, is found as *h* = −15.435*f* + 212.04. Based on the glossa anatomy, a fluid transport model is proposed to calculate the nectar intake rate. Theoretical analysis showed that a honey bee with

−

shorter setae can compensate the nectar intake rate by increasing the dipping frequency. Considering the wear of the setae and dipping compensation, Shi arrived at the results that the total energy intake rate is about 106 times the power required to overcome viscous drag; the energy dissipation caused by viscous drag is negligible [26]. Therefore, the effect of augmentation of viscous drag caused by the increase of the dipping frequency on the energy intake rate of bees is almost negligible. Natural selection tends to feed quickly and efficiently, as honey bees are threatened by predators and economic necessities [38]. Therefore, honey bees must meet the contradictive demands of keeping the visit time short and the optimal nectar mass intake rate. Although the natural wear of glossal setae will affect the nectar intake rate, by adjusting the dipping frequency, both requirements can be satisfied, which is in accordance with the results from lab tests of wearing bee tongues in the 35% and 45% sucrose solutions, respectively (Figure 9).

ሶ **Figure 9.** The honey bee augments dipping frequency to compensate for glossa hair deterioration. The relationship between theoretical nectar mass intake rate . *M* and setae wear agrees with the experimental data captured from lab tests for dipping both the 35% and 45% sucrose solutions [39].

#### **6. Conclusions**

Investigations of feeding techniques by a honey bee's glossa are interdisciplinary work that covers morphology, behavioural dynamics, and energy-saving strategies. Future work might be extended to the following aspects. (1) The nectar property may drive the feeding property more complicated, in which, for instance, the nectar viscosity increases steeply with respect to the nectar concentration, and it is also influenced by the temperature [13]. The flowers may have an internal microclimate which is up to 4 ◦C higher than the external temperature, which not only provides more heat to sustain the thoracic temperature of honey bees especially in winter time but makes the nectar a bit thinner, which is much easier to be digested because of the lower nectar viscosity [13]. The combined experimental and theoretical methodology is required to uncover this. (2) The dipping behaviour may be an indicator to reflect the health state of the honey bees. Air pollution, pesticide abuse, and climate change may strongly influence the honey production rate and even survival rate of the bee colony [40,41]. The dipping frequency is closely related to the energetic intake rate, so we may use the dipping frequency as a measure to evaluate the health status of

the bee colony. (3) The bee cannot only feed on nectar in different floral structures but can lick dry sugar during droughts. The functional flexibility in feeding remains unexplored. The bee glossa is comprised of segmented structure which can perform a million times of reciprocating movements. How the bee glossa meets the contradictive demands of high deformability and stiffness is still unknown. Combining various experiments and theoretical frameworks, more extensive research will be conducted, not only to reveal the behavioural characteristics of honey bees but for inspiring the next-generation facilities like micropumps and other viscous fluidic transport facilities.

**Author Contributions:** Conceptualization, J.W. and Z.W.; resources, J.Z. and H.W.; data curation, H.W.; writing—original draft preparation, H.W. and J.W; writing—review and editing, H.W., and J.W.; supervision, Z.W. and J.W.; project administration, J.W.; funding acquisition, J.W. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Natural Science Foundation of China (grant no. 51905556), the Grant for Popularization of Scientific and Technological Innovation of Guangdong Province (grant no.2020A1414040007), and the research grant of Sun Yat-Sen University for Bairen Plan (grant no.76200-18841223).

**Institutional Review Board Statement:** Not applicable.

**Data Availability Statement:** Data is contained within the article or supplementary material. The data presented in this study are available in "Nectar feeding by a honey bee's hairy tongue: morphology, dynamics, and energy-saving strategies".

**Acknowledgments:** We appreciate Wei Zhang and Yu Sun from Sun Yat-Sen University for the preparation of figures.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


### *Article* **How Does the Intricate Mouthpart Apparatus Coordinate for Feeding in the Hemimetabolous Insect Pest** *Erthesina fullo***?**

#### **Yan Wang and Wu Dai \***

Key Laboratory of Plant Protection Resources and Pest Management of the Ministry of Education, College of Plant Protection, Northwest A&F University, Yangling 712100, China; wangyan105422@163.com

**\*** Correspondence: daiwu@nwsuaf.edu.cn; Tel.: +89-29-8708-2098

Received: 9 July 2020; Accepted: 2 August 2020; Published: 4 August 2020

**Simple Summary:** To better understand the feeding mechanism of *Erthesina fullo*, the fine structure of the mouthparts is examined with scanning electron microscopy, and feeding performance are observed directly under laboratory conditions for the first time. The adult feeding process involves several steps, including exploring and puncturing of the host plant epidermis, a probing phase, an engorgement phase, and removal of the mouthparts from the host tissue. Proceeding from labium towards the mandibular stylets, the movement pattern becomes increasingly stereotypical, including the sensilla on the tip of the labium probing, the labium making an elbow-like bend between the first and second segment, the base of the stylet fascicle housing in the groove of the labrum, the mandibular stylets penetrating the site and maxillary stylets feeding. The morphology of mouthparts is similar to those of other Heteroptera. The four-segmented labium has eleven types of sensilla. The mandibular stylet tips have two nodules preapically on the convex external surface. The structure and function of the mouthparts are adapted for the phytophagous feeding habit in this species. This study increases the available detailed morphological and behavioral data for Hemiptera and will potentially contribute to improving our understanding of this pest's feeding behavior and sensory mechanisms.

**Abstract:** The yellow marmorated stink bug, *Erthesina fullo* (Thunberg, 1783), is a major pest of certain tree fruits in Northeast Asia. To better understand the feeding mechanism of *E. fullo*, the fine structure of the mouthparts, including the distribution and abundance of sensilla, are examined with scanning electron microscopy (SEM), and their functions are observed directly under laboratory conditions. The feeding performance is described in detail and illustrated for the first time. The adult feeding process involves several steps, including exploring and puncturing of the host plant epidermis, a probing phase, an engorgement phase, and removal of the mouthparts from the host tissue. Proceeding from labium towards the mandibular stylets, the movement pattern becomes increasingly stereotypical, including the sensilla on the tip of the labium probing, the labium making an elbow-like bend between the first and second segment, the base of the stylet fascicle housing in the groove of the labrum, the mandibular stylets penetrating the site and maxillary stylets feeding. In terms of morphology, the mouthparts are similar to those of other Heteroptera, consisting of a triangular pyramidal labrum, a tube-like and segmented labium with a deep groove on the anterior side, and a stylet fascicle consisting of two mandibular and two maxillary stylets. The four-segmented labium has five types of sensilla basiconica, three types of sensilla trichodea, two types of sensilla campaniformia and 1 type of sensilla coeloconica. Among them, sensilla trichodea one and sensilla basiconica one are most abundant. The tripartite apex of the labium is covered with abundant sensilla trichodea three and a few sensilla basiconica 5. The mandibular stylet tips have two nodules preapically on the dorsal margin of the convex external surface, which may help in penetrating plant tissue and anchoring the mouthparts. The externally smooth maxillary stylets interlock to form a larger food

canal and a smaller salivary canal. The structure and function of the mouthparts are adapted for the phytophagous feeding habit in this species. Similarities and differences between the mouthparts of *E. fullo* and those of other Heteroptera are discussed.

**Keywords:** *Erthesina fullo*; mouthparts; sensillum; ultramorphology; feeding performance

#### **1. Introduction**

Hemiptera is the largest and most diverse non-holometabolous insect order, containing over 75,000 species. They are characterized by specialized piercing-sucking mouthparts, in which the modified mandibles and maxillae form two pairs of stylets sheathed within a modified labium [1,2]. These mouthparts facilitate feeding on fluids of various animal and plant hosts and have sensory organs used in both host location and feeding. The Hemiptera have been classified into four major taxa (suborders: Auchenorrhyncha, Sternorrhyncha, Coleorrhyncha and Heteroptera). Abundant data are available on some aspects of hemipteran mouthpart morphology based on light and scanning electron microscopy, for a few species of Auchenorrhyncha, e.g., Fulgoroidea [3,4], Cicadellidae [5–9], Aphidoidea [10–13], Coccoidea [14] and Aleyrodidae [15,16] of Sternorrhyncha and the Heteroptera [17–20]. These provide insights into feeding mechanisms and contribute to assessment of phylogenetic relationships [7,8,11,13,19,21–25]. However, so far, mouthpart morphology of some major groups remains little studied.

As a biologically successful group of organisms, the heteropterans (true bugs) are prolific and diverse and have acquired a variety of feeding habits. Some heteropterans suck surface fluids (e.g., nectar), some pierce tissues to suck sap or blood, and others obtain nourishment from dried seeds. Numerous modifications of mouthpart structures reflect the diversity of food sources and feeding habits of this group. Cobben [17] studied the heteropteran feeding stylets of 57 families and 145 species, but provided little information on the labium and the types and distributions of sensilla present on this stucture. In the carnivorous heteropterans, different feeding mechanisms are reflected in differences in the labial tip sensilla [26] and the movement and penetration of the stylets during feeding [27].

The strategies used by various phytophagous Heteroptera to feed on a variety of plant structures may include stylet-sheath feeding, lacerate and flush feeding, macerate and flush feeding, and osmotic pump feeding [17,28]. In general, feeding damage from heteropterans can be classified into five categories: Localized wilting and necrosis, abscission of fruiting forms, morphological deformation of fruits and seeds, modified vegetative growth, and tissue malformation [29–31]. All plant-sucking heteropterans are potential vectors of plant disease, and the lesions left behind at the feeding site can facilitate secondary infections by plant pathogens [32].

Pentatomidae (stink bugs) is one of the largest families within the Heteroptera. Stink bugs feed by inserting their stylets into the food source to suck up nutrients and may transmit plant pathogens, resulting in plant wilt and, in many cases, abortion of fruits and seeds. Compared with more specialized Hemiptera, Pentatomidae use diverse feeding strategies that allow them to feed from a wide range of plant structures including vegetative structures, such as stems and leaves, and reproductive plant structures such as seeds, nuts, pods and fruits [33]. Stink bug feeding can damage crops in different ways dependent upon the plant structure(s) attacked, e.g., vegetative or reproductive. Previous studies of mouthparts in Pentatomidae have mostly focused upon differences in certain aspects of the mouthparts among stinkbug species, such as the types and distribution of labial sensilla [34–36] and the internal structure of mandibular and maxillary stylets [37]. Information on the distribution of sensilla on the mouthparts and the relationships between mouthpart structure and function in feeding, useful in the classification of stink bugs, is not yet available.

The yellow marmorated stink bug *Erthesina fullo* (Thunberg, 1783) is one of the most widely distributed phytophagous pests in East Asia. It causes severe loss to many horticultural crops, such as

apples, cherries and pears [38–40], and disturbs humans by invading houses in large numbers when overwintering. Both nymphs and adults of *E. fullo* primarily suck the sap from the trunk, leaves, immature stems and fruits of plants. Previous research on this species has mainly been limited to basic biology, behavior and integrated control [39,41,42]. Although, feeding damage from *E. fullo* has been characterized in agronomic crops, tree fruits and vegetables, little is known about the fine structure of the mouthparts and the feeding mechanism of *E. fullo,* and in particular, how the sensilla are distributed on the mouthparts and function in locating host plants.

Here, we used scanning electron microscopy to investigate the mouthpart morphology and distribution of sensilla of *E. fullo*. We also observed feeding behavior. The outcome of this study increases the available detailed morphological and behavioral data for Hemiptera and will potentially contribute to improving our understanding of this pest's feeding behavior and sensory mechanisms. This study provides more data for future comparative morphological studies in Pentatomidae.

#### **2. Material**

#### *2.1. Insect Collecting*

Adults of *E. fullo* used for SEM in this study were obtained from the campus of Northwest A&F University in Yangling, Shaanxi Province, China (34◦16′ N, 108◦07′ E, elev. 563 m) in August 2016, and were preserved in 70% ethanol and stored at 4 ◦C. For observing feeding behavior, additional adults of *E. fullo* were collected at the same locality in September 2019.

#### *2.2. Samples for SEM*

Ten females and twelve male specimens of *E. fullo* were fixed in 70% ethanol. The labium and the stylet, dissected using fine dissecting needles under 40× magnification (Nikon SMZ 1500, stereomicroscope, Tokyo, Japan), were prepared as study samples. The samples were cleaned in an ultrasonic bath (250 W) (KQ118, Kunshan, China) for 10 to 15 s in 70% ethanol three times, then dehydrated in serial baths of 80%, 90% and 100% ethanol each for 15 min. Samples then underwent dehydration in a mixture of 100% ethanol and 100% tert-Butanol at the ratios 3:1, 1:1, and 1:3 (by volume) for 15 min at each concentration followed by a final replacement treatment in 100% tert-Butanol for 30 min. Specimens were then freeze-dried with liquid CO2, mounted on aluminum stubs with double-sided copper sticky tape and sputtered with gold/palladium (40/60) in a LADD SC-502 (Vermont, USA) high resolution sputter coater. The samples were subsequently examined with a Hitachi S-3400N SEM (Hitachi, Tokyo, Japan) operated at 15 kV [8] or Nova Nano SEM-450 (FEI, Hillsboro, OR, USA) at 5–10 kV.

#### *2.3. Feeding Behavior on Di*ff*erent Types of Substrates*

To observe the feeding behavior of the insects on fresh fruit and twigs, some fruit of orange, pear and grape as well as twigs of pear and grape were offered to twenty male and female individuals of *E. fullo* in an optical quality colorless glass enclosure 100 mm in diameter and 135 mm tall. The insects were observed intermittently throughout the feeding period for one week. Sequential images of adult feeding performance were taken using a mobile phone (vivo Y18L) with an 8-megapixel camera when conditions were suitable. The images were saved directly to a computer for later analysis.

#### *2.4. Image Processing and Terminology*

Photographs and SEMs of mouthparts were observed and measured after being imported into Adobe Photoshop CC 2019 (Adobe Systems, San Jose, CA, USA). Measurements are given as means ± standard error of the mean. Schematic diagrams were drawn with Microsoft Office Word 2007 and processed with Photoshop CC 2019 (Adobe Systems, San Jose, CA, USA). For classification of sensilla, the systems of Altner and Prillinger [43] were used in addition to the more specialized nomenclature from other studies [44].

#### *2.5. Data Analysis*

The lengths of the mouthparts were compared between sexes using a Student t-test. Statistical analyses were executed using SPSS 19.0 (SPSS, Chicago, IL, USA).

#### **3. Results**

#### *3.1. General Morphology and Structure of Mouthparts*

The mouthparts of *E. fullo* are similar to those of other heteropterans, arising from the anteroventral part of the head capsule and composed of a long labrum, a tube-like labium and a stylet bundle comprising two maxillary stylets (Mx) and two mandibular stylets (Md). The four-segmented labium has a long internal labial groove (Lg) that surrounds the stylet fascile (Sf) and is covered with different types of sensilla symmetrically distributed on the surface of both sides of the groove or on the distal surface (Figure 1A–C). The two inner maxillary stylets interlock to form the food and salivary canals; they are partially surrounded by two serrate-edged mandibular stylets. The stylet fascicle is housed inside the labial groove and proximally covered by the small cone-shaped labrum. No obvious differences were noted between the mouthpart structures of females and males except for the length of the labium (t(9) = 9.473, *p* = 0.000) (Table 1).

**Figure 1.** Scanning electron micrographs of the head of *E fullo*. (**A**) Ventral view; (**B**) Lateral view; (**C**) Dorsal view showing four-segmented labium (I–IV); Lg, labial groove; Lm, labrum; Lb, labium.

#### 3.1.1. Labrum

**μ μ** The cone-shaped labrum (Lm) is attached to the anterior margin of the anteclypeus and protrudes forward beyond 2/5 length of the second segment (Figures 1A and 2A). It is closely adpressed over the first labial segment and partly embedded in the labial groove. The surface of the labrum is plicated and densely covered with regular transverse wrinkles (Figure 2A). The ventral side of the labrum bears a pair of sensilla basiconica 1 (Sb1). Sensilla trichodea 1 (St1), sensilla coeloconica (Sco) and cuticular pores (Cpo) are arranged irregularly on the ventral region of the labrum (Figure 2A–H).


**Table 1.** Measurements of labrum and labium (mean ± SE) obtained from scanning electron microscopy. N = sample size. Lm, labrum; Lb, labium; Lb1, first segment of labium; Lb2, second segment of labium; Lb3, third segment of labium; Lb4, fourth segment of labium.

**Figure 2.** SEM of labrum of *E. fullo*. (**A**) Ventral view; (**B**) Enlarged view of box in (**A**), showing sensillum basiconicum 1 (Sb1); (**C**) Enlarged view of box in (**A**), showing sensilla trichodea 1 (St1), sensilla coeloconica (Sco) and cuticular pores (Cpo); (**D**) Sensillum basiconicum 1 (Sb1); (**E**) Enlarged view of box in (**D**); (**F**) Sensillum trichodeum 1 (St1); (**G**) Sensillum coeloconicum (Sco); (**H**) Enlarged view of cuticular pore (Cpo).

#### 3.1.2. Labium

The labium, suspended from the front of the head, is tubular in shape and subdivided into four-segments externally (Figure 1A–C). The anterior surface of the labium is bisected by a deep longitudinal groove, which encases the mandibular and maxillary stylets. All segments of the labium are covered with different types of sensilla mainly positioned symmetrically on each side of the labial groove (Lg) and distally, with fewer sensilla on the posterior and lateral surfaces.

The four segments vary in size (Table 1) and morphology (Figure 1A–C). Overall the labium is broad and of uniform width through most of its length with the distal segment widening near the tip.

The proximal labial segment (Lb1), the shortest and widest of the four segments, is broad at the base, gradually narrows at the middle, and then slightly widens to the apex in posterior view (Figures 2A and 3A,B). The distal part of the dorsum is contracted inward, is crescent-shaped and no sensilla are observed on this surface (Figure 3A). Four types of sensilla (sensilla basiconica 1, sensilla trichodea 1, sensilla coeloconica, sensilla campaniformia 1) and cuticular pores (Cpo) are arranged on the ventral surface (Figure 3B–F).

**Figure 3.** SEM of first labial segment of *E. fullo*. (**A**) Ventral view; (**B**) Lateral view; (**C**) Enlarged view of surface of the first segment, showing sensilla campaniformia 1 (Sca1), sensilla coeloconica (Sco) and cuticular pores (Cpo); (**D**) Enlarged view of outlined box in (**B**), showing sensilla trichodea 1 (St1), sensilla coeloconica (Sco) and sensilla basiconica 1 (Sb1); (**E**) Sensillum campaniformium 1 (Sca1); (**F**) Enlargement of outlined box in (**D**), showing sensillum coeloconicum (Sco) and sensillum basiconicum 1 (Sb1).

The second segment (Lb2) is longer than the first segment (Table 1). Viewed from the ventral and dorsal sides, the base and ends are wider, while the middle part is narrower (Figure 4A,C). However, from the lateral view, the middle part is expanded, and both ends are narrowed (Figure 4B). Six types of sensilla were found on this segment, including three types of sensilla basiconica (Sb1, Sb2, Sb3), one type of sensilla campaniformia (Sca1), and one type of sensilla trichodea (St1) and sensilla coeloconica (Sco) (Figure 4D–H). Also, there are some cuticular pores (Cpo) arranged on the surface of second segment (Figure 4E).

The third segment (Lb3) is a little longer than the second, and of uniform width on both sides (Table 1, Figure 5A–C). Generally, there is a groove on the dorsal surface of the last 3/5 (Figure 5C). A wrinkled area is present on the dorsal surface at the internode between the second and third labial

segment. Three types of sensilla are distributed on this part, including sensilla basiconica 1 (Sb1), sensilla trichodea (St1) and sensilla coeloconica (Sco) (Figure 5D–G).

**Figure 4.** SEM of second labial segment of *E. fullo*. (**A**) Ventral view; (**B**) Lateral view; (**C**) Dorsal view; (**D**) Enlarged view of outlined box of (**B**) showing sensilla basiconica 2 (Sb2); (**E**) Enlarged view of outlined box of (**A**) showing sensilla campaniformia 1 (Sca1), cuticular pores (Cpo), sensilla trichodea 1 (St1), sensilla coeloconica (Sco) and sensilla basiconica 3 (Sb3); (**F**) Sensillum basiconicum 3 (Sb3); (**G**) Enlarged view of surface of the second segment of outlined box of (**A**); (**H**) Enlarged view of outlined box of (**G**), showing sensilla basiconica 1 (Sb1), sensilla trichodea 1 (St1), cuticular pore (Cpo) and sensilla campaniformia 1 (Sca1).

The fourth segment (Lb4) is nearly conical, of uniform width from base to apical 1/4 then narrowing to the apex (Figure 6A–C). There are abundant sensilla distributed on this segment, including three types of sensilla trichodea (St1, St2, St3), three types of sensilla basiconica (Sb1, Sb2, Sb3), two types of sensilla campaniformia (Sca1, Sca2) and sensilla coeloconica (Sco) (Figures 6D–I and 7A–E). *E. fullo* has very long and numerous sensilla trichodea 3 (St3) covering the end of the labium giving it a brush-like appearance (Figure 7A–C). Several sensilla basiconica 5 (Sb5) are visible among these sensilla trichodea 3 (St3) (Figure 8A,B).

**Figure 5.** SEM of the third labial segment of *E. fullo*. (**A**) Ventral view; (**B**) Lateral view; (**C**) Dorsal view; (**D**) Enlarged view of outlined box of (**A**); (**E**) Enlarged view of outlined box of (**C**); (**F**) Enlarged view of surface of the third segment, showing sensilla basiconica 1 (Sb1), sensilla trichodea 1 (St1), and sensilla coeloconica (Sco); (**G**) Enlarged view of outlined box of (**F**), showing base pore (p).

**Figure 6.** SEM of the fourth labial segment of *E. fullo*. (**A**) Ventral view; (**B**) Lateral view; (**C**) Dorsal view; (**D**) Enlarged view of outlined box of (**C**), showing sensilla trichodea 2 (St2), sensilla coeloconica (Sco) and sensilla campaniformia 2 (Sca2); (**E**) Enlarged view of outlined box of (**D**), showing sensilla campaniformia 2 (Sca2); (**F**) Enlarged view of outlined box of (A), showing sensilla basiconica 2 (Sb2); (**G**) Sensillum trichodeum 2 (St2); (**H**) Sensillum basiconicum 2 (Sb2); (**I**) Enlarged view of surface of the fourth segment, showing sensilla campaniformia 1 (Sca1), sensilla trichodea 1 (St1), sensilla coeloconica (Sco) and sensillum basiconicum 1 (Sb1).

**Figure 7.** Proximal position the fourth labial segment of *E. fullo*. (**A**) Ventral view; (**B**) Lateral view; (**C**) Dorsal view; (**D**) Sensilla basiconica 4 (Sb4); (**E**) Enlarged view of outlined box of (**B**), showing sensillum trichodeum 2 (St2) and sensilla coeloconica (Sco).

**Figure 8.** Tip of labium of *E.fullo*. (**A**) Vertical view of labial tip showing sensilla basiconica 5 (Sb5) and sensilla trichodea 3 (St3); (**B**) Enlarged view of outlined box of (**A**), showing sensilla basiconica 5 (Sb5).

#### 3.1.3. Labial Sensilla Types and Their Arrangement

Based on their external morphology and distribution, eleven types (subtypes based on the length and shapes are distinguished) of distinct sensilla were observed on the surfaces of the labial segments. They were classified as: sensilla trichodea (St), sensilla campaniformia (Sca), sensilla coeloconica (Sco) and sensilla basiconica (Sb).

Sensilla trichodea (St) are hair-like sensilla. Their walls are smooth without any pores or grooves on the surface. Three subtypes of sensilla trichodea were distinguished. Sensilla trichodea 1 (St 1) are short (Table 2), aporous, smooth, with a slightly rounded tip and flexible sockets (Figure 2F). These sensilla are numerous and uniformly distributed on the labrum (Lm) and labium (Lb1–4) (Figures 2C and 3D). Sensilla trichodea 2 (St 2) are longer than sensilla trichodea 1 (Table 2), straight, with a smooth surface, a rounded tip and flexible sockets (Figure 6D,G). These sensilla are uniformly distributed on the fourth labial segment (Lb4). Sensilla trichodea 3 (St 3) are the longest sensilla (Table 2). These sensilla are curved at the tip and embedded in inflexible sockets. These sensilla are very numerous and located on the tip of the labium (Figure 8A,B).

Five subtypes of sensilla basiconica were distinguished. Sensilla basiconica 1 (Sb 1) are hair-like sensilla identical sensilla trichodea except for their smooth walls and blunt-tip. In the studied species, these sensilla are long (Table 2) ribbed and straight, slightly branched at the tip and arise from a cuticle with a flexible socket (Figure 2D,E). These sensilla are distributed on the labrum (Lm) and labium (Lb 2–4) (Figure 2B, Figure 3B,D,F, Figure 4G,H, Figure 5D–F and Figure 6I). Sensilla basiconica 2 (Sb 2) are cones with a smooth surface that arise from flexible sockets (Figures 4D and 6H). Three pairs of sensilla basiconica 2 are arranged at the junction of the first and second segment, two are present on each side of the junction of the third and fourth segments (Figures 4D and 6F). Sensilla basiconica 3 (Sb3) are short with a smooth surface, have a sharp tip and sit in a pit (Figure 4E,F). These sensilla are sparsely distributed on the ventral surface of the second segment. Sensilla basiconica 4 (Sb4) are peg-like with a smooth surface and have a rounded tip (Figure 7D). They are sparsely distributed on the lateral surface of the last segment (Figure 7B). Sensilla basiconica 5 (Sb 5) are present at the center of each distal lobe (Figure 7A). This type of sensillum is long, straight and with a smooth surface and a rounded tip, probably with a terminal pore (Figure 8B). Several of these sensilla basiconica are visible among the sensilla trichodea of the distal brush (Figure 7A,B).

Sensilla campaniformia (Sca) are flat, oval-shaped discs. Two subtypes of sensilla campaniformia are distinguished. Sensilla campaniformia 1 (Sca 1) are large (Table 2), numerous and present on the labrum (Lm) and labium (Lb 1–4) (Figure 2C, Figure 3E, Figure 4E,H and Figure 6I). Sensilla campaniformia 2 (Sca 2) are smaller than sensilla campaniformia 1 (Sca 1), fewer in number and located on the antero-lateral surface near the apical 1/3 (Figure 6D,E).

Sensilla coeloconica (Sco) consist of a small oval protuberance or cone inserted in a cuticular depression (Figure 2G). These are located on the labrum (Lm) and labium (Lb1–4) (Figure 2C, Figure 3C,D, Figure 4E, Figure 5F, Figure 6D,I and Figure 7E). These are without pores and have an inflexible socket.

**Table 2.** Distribution, morphometric data (mean ± SE), terminology and definition of sensilla used in the present paper. Data are mean ± SE values obtained from scanning electron microscopy. N = sample number; Lm, labrum; Lb, 1, 2, 3, 4, the first, second, third, fourth segmen<sup>t</sup> of labium; St 1–3, sensilla trichodea 1–3; Sb 1–5, sensilla basiconica 1–5; Sco, sensilla coeloconica; Sca 1–2, sensilla campaniformia 1–2; SF, sensory field on the labial tip; Wp, wall pore; Tp, tip pore.


#### *3.2. Stylet Fascicle*

The stylet fascicle is long, slender, and composed of two separated mandibular stylets and two interlocked maxillary stylets (Figure 9A), ensheathed by the labium at rest and extending from the opening of the labial tip during feeding.

**Figure 9.** SEM of stylet fascicle of *E. fullo*. (**A**) Stylet fascicle showing mandibular (Md) and maxillary stylets (Mx); (**B**)External view of mandibular stylet (Md) showing eleven short transverse ridges (tr); (**C**) Interior view showing small squamous textures (sst), bigger squamous textures (bst) and middle squamous textures (mst); (**D**) Lateral view showing two nodules (no); (**E**) Apices of interlocked maxillary stylets; (**F**) Apex of left maxillary stylet (LMx) showing food canal (Fc) and salivary canal (Sc); (**G**) Apex of right maxillary stylet (RMx) showing food canal (Fc) and salivary canal (Sc); (**H**) External view of left maxillary stylet (LMx); (**I**) External view of right maxillary stylet (RMx); dr, dorsal side; vr, ventral side.

The mandibular stylets, located on each side of the maxillary stylets, are crescent-shaped in cross-section, convex externally and slightly concave internally to form a groove enclosing the maxillary stylets. On the lateral surface of each mandibular stylet, a series of approximately parallel, curved serrate ridges or teeth (a regular series of longer transverse ridges and eleven shorter transverse ridges) extend over the most distal part (Figure 9B). The most obvious features observed on the mandibular stylets of this species are two nodules present on the dorsal margin of the convex external surface near the apex (Figure 9B,D). There are four rows of squamous structures regularly distributed on the inner surface of the mandibular stylet (Figure 9C). The first and third rows consist of small squamous textures

(sst), the second has bigger squamous textures (bst) and the fourth has medium-sized squamous textures (mst) with different cuticular spines.

The maxillary stylets (Mx) are interlocked by hook-like hinges and are not symmetrical (Figure 9E). The hook-like hinges include three joints from the cross-section, one of which is located at the center of the maxillary stylets and two of which are positioned at the lateral sides (Figure 10A,B). The external and internal surface of a maxillary stylet is smooth and the tip is sharp (Figure 9F–I). A row of nodes is present on the joint surface of the left stylet, which opposes the series of indentations on the right stylet (Figure 9G). A food canal (Fc) and salivary canal (Sc) are formed by the interlocked maxillary stylets, and the width of food canals is evenly distributed across the two stylets, while most of the salivary canal is housed in the right stylet (Figure 9F,G). The diameter of the central food canal is much greater than that of the salivary canal (Figure 10A,B). The cross-section of the stylet fascicle shows that each mandibular stylet has a dendritic canal, which is a large duct that runs the length of the stylet and is located centrally in the thickest portion of each structure (Figure 10A,B).

**Figure 10.** Cross-section of stylet fascicle of *E. fullo*. (**A**) Cross-section of stylet fascicle through middle of second and third segment showing food canal (Fc) and salivary canal (Sc); (**B**) Diagram of cross-section of stylet fascicle. LMd, left mandibular stylet; RMd, right mandibular stylet; LMx, left maxillary stylet; RMx, right maxillary stylet; Fc, food canal; Sc, salivary canal; Ic, interlocking canal; CN, dendritic canal; RPr, Right process of the maxilla; A, Straight; A', Hooked; B, Hooked; B', Straight; C, Straight; C', Hooked; D, T-shaped; D', Hooked; E, Hooked; E', Hooked; F, Straight.

#### *3.3. The Process of Feeding by E. fullo*

The adult feeding process involves several steps, including the exploring and puncturing of the plant epidermis, a probing phase, an engorgement phase, and removal of the mouthparts from the plant tissue. These processes vary slightly in mouthpart position and duration.

When the insect is at rest or not feeding, the rostrum is in contact with the ventral surface of the body from the front coxal base to the anterior part of the abdomen (Figure 11A). The proximal end of labial segment 2 articulates with the bucculae.

Insects feeding on the internal fluids of other organisms must first penetrate the plant tissues. After landing, an adult of *E. fullo* walks on its plant and explores for a suitable feeding location by probing. It then stops and remains still while the antennae swing up and down several times. After a few seconds, gripping the plant with its legs, the bug tilts the anterior part of its body upward at an angle to the surface, and the rostrum is then extended forward and used as a sense organ in conjunction with the eyes and antennae to examine the plant material for a suitable feeding spots. The labium first moves forward by swinging from its horizontal position of repose until it is perpendicular to the plant surface. The rostrum tip then taps the surface or slides over it. When the labium is moved

forward from its resting position, the stylet tip reaches the tip of the labium and may even extend a short distance beyond the top (Figure 11B).

**Figure 11.** Feeding on fruits and young stalks in adult *E. fullo* showing positions of the mouthparts. (**A**) At rest or not feeding; (**B**) Exploring suitable feeding location; (**C**) Feeding on grape; (**D**) Feeding on green stalk of grape; (**E**) Feeding on stalk of pear; (**F**) Feeding on pear.

Upon contact with a potential feeding site, the bug may probe with sensilla on the tip of the labium and penetrate the site with the stylets to test if this site is suitable for feeding. After selecting an appropriate feeding site, the insect then presses the tip of the labium onto the plant surface and inserts the feeding stylets. Then, the labium makes an elbow-like bend between the first and second segment, while the base of the stylet fascicle is held in the groove of the labrum (Figure 12A). The labium continues retracting to its maximum extent, at which the angle between the first and second segments is nearly 90◦ , allowing the head to be lowered as the stylet bundle penetrates the food tissue (Figure 12B,C), with the maxillary stylets lagging slightly behind the mandibular stylets. Stylet probing continues until a suitable tissue is found. It may take anywhere from five minutes to three hours from the beginning of probing until a feeding site is reached. The bug secretes viscous saliva as the stylets progress through the tissue.

**Figure 12.** Feeding stages on orange in adult *E. fullo* showing positions of the mouthparts. (**A**) Location of suitable feeding position by the labium; (**B**), (**C**) Puncture of orange by stylet fascicle showing elbow-like fold of proximal and second rostral segments and stylet penetration; (**D**) The bug lifts up the third and fourth segments of labium parallel to host surface and then feeds; (**E**) The bug gradual straightens the first and second labial segments; (**F**), (**G**) Termination of feeding showing retraction of stylets; (**H**) Use of forelegs to return stylet fascicle to labial groove; Lm, labrum; Lb, labium; Sf, stylet fascicle.

After a feeding site is reached, the bug bends the third and fourth segments of the labium backward, away from the inserted stylets until the distal section of the labium is parallel to the host surface, after which feeding can commence (Figure 12D). The bug then extracts and sucks host fluids repeatedly. Feeding may last from a few seconds to one hour at a time.

When finished feeding, the bug gradually straightens the first and second labial segments; meanwhile, the third and fourth labial segments rotate forward and contact the host surface (Figure 12E). Then the body gradually raises and pulls out the stylet fascicle (Figure 12F,G). The bug replaces the stylet fascicle into the labial groove with the help of the forelegs (Figure 12H). Finally, the rostrum rotates back to its resting position along the sternum.

The process of feeding on young shoots of the plant is similar to that observed for fruit feeding, except that the stylets are never fully retracted from the labium (Figure 11C–F).

#### **4. Discussion**

Substantial data are available on structure and function of mouthparts in Hemiptera. However, detail on the mechanics of feeding behavior, especially with respect to the sensory and motor feedback mechanisms, is lacking [45–47]. A study of the fine morphology of mouthparts allows us to interpret the function of the component parts of the feeding apparatus and improves our understanding of the actual feeding mechanism.

In this study, the feeding behavior of *E. fullo* is described. To our knowledge, this is the first time that the detailed mouthpart morphology and feeding performance in a member of Pentatomidae have been reported together. The modified mouthparts of *E. fullo* have a number of morphological similarities to those heteropteran species described previously [17,19,22,37,44,48–55], but our study revealed some new and interesting features that differ from those of other true bugs, and provide a better understanding of the feeding strategies and the sensory systems of *E. fullo*.

#### *4.1. Mouthpart Morphology and Their Adaptability to Feeding*

The labrum, a conspicuous anterior structure on the adult insect head, should play an important role in insect feeding. Recently the labrum was reinterpreted as fused paired appendages of an intercalary segment [56–58] and a few scholars have conducted detailed studies on its morphology in Heteroptera [52–55,59–61]. In previously published reports, the morphology of the labrum was used as a taxonomic feature of higher taxa of Heteroptera [59–63], but its structure varies according to feeding habits and mechanisms [61]. Spooner [59] recognized three basic types of labrum in Heteroptera: (1) a broad, flap-like labrum; (2) a long, narrow, triangular labrum; (3) a broad, flap-like sclerite with a long epipharyngeal projection. The labrum of *E. fullo* corresponds to the second group. This is similar to other true bugs, e.g., *Pyrrhocoris sibiricus* [52], *Cheilocapsus nigrescens* [53], and four species of Largidae (*Physopelta quadriguttata*, *Ph. gutta*, *Ph. cincticallis*, and *Macrocheraia grandis*) [55]. We observed regular wrinkles from base to end on the ventral surface of the *E. fullo* labrum. These wrinkles may function to add flexibility to the labrum, allowing deeper stylet penetration (Figure 12B,C). Such a long labrum (Table 1) may also be used to hold the basal part of the stylet fascicle in the labial groove during feeding (Figure 12D).

The labium of *E. fullo* has four segments as in most of other heteropteran bugs [27,52–55]. Usually, when insects are feeding, the second segment approaches the first segment, allowing the head to be lowered as the stylet fascicle penetrates the food tissue. Previous studies have reported that a band-like dorsal plate is present between the first and the second segment [27,52,54,55,64–66], while the base of the stylet fascicle is held in the groove of the labrum. However, in our study, there was no such structure (a band-like dorsal plate), and the distal part of the dorsum of the first segment is contracted inward (Figures 11 and 12). This is probably because, unlike *Pyrrhocoris sibiricus* [52], which moves the labium back to its abdomen, *E. fullo* bends the first and second segments for deeper feeding. Moreover, the first and second segment are stronger than the third and fourth segments. The first and second labial segments of *E. fullo* are presumably folded to support the head, allowing the stylet fascicle to penetrate the plant (Figure 12B).

Heteropteran stylets form a fascicle composed of two lateral mandibular stylets and two maxillary stylets; the former are armed with teeth or rasps and the latter interlock and forms the salivary and food canals [17,18,25]. As feeding and probing on host plants are responsible for the direct or indirect

damage to plants by phytophagous hemipteran insects, the stylets, including the shape and dentition of the tips, have been studied previously in several heteropterans [17,18,20,49,50,52–55,67–75]. In *E. fullo*, there are a series of squamous textures regularly distributed on the inner surface of the mandibular stylet and the left and right sides of the longitudinal groove are different. Similar structures are found in other phytophagous species [17,52,53,55]. Cobben [17] mentioned that the orientation of this parallel groove is such that the forward thrust of one mandible will cause considerable friction against the outer surface of the adjacent maxillary stylet contributing to its inward deviation. We also observed two nodules present on the dorsal margin of the external surface and a series of transverse ridges arranged on the outer surface. In different phytophagous Heteroptera, the nodules are slightly different, and the number of nodules also varies [49,52–55]. Depieri and Panizzai [49] observed 1 to 4 central teeth and 1–3 lateral teeth in *Dichelops melacanthus*, *Euschistus heros*, *Nezara viridula* and *Piezodorus guildinii*. Wang and Dai [52] found that mandibular stylets of *P. sibiricus* have three central teeth and two paired lateral teeth on the distal extremity, as well as five or six oblique parallel ridges on the subapex of the external convex region. In polyphagous species of Largidae (*Physopelta quadriguttata*, *Ph. gutta*, *Ph. cincticallis*, and *Macrocheraia grandis*), the serration pattern of the mandibles is 1–3 central teeth and 1–2 lateral teeth [55]. The teeth at the tip of the mandibular stylet may help to fix the stylets in host tissues [17,76].

Both mandibles together with the maxillary bundle function as a single plunging instrument [17]. Maxillary stylets are asymmetrical only in the internal positions of the longitudinal carinae and grooves. Their inner surfaces show traces of small, widely spaced notches arranged in longitudinal strips. As found by Cobben [17] in his study of *Graphosoma lineatum* L., we also found these grooves on the maxillary stylets of *E. fullo* to form a salivary canal (Sc) and a food canal (Fc). The maxillary stylets are longer than the mandibular stylets and the salivary canal is narrower than the food canal as in *Pyrrhocoris sibiricus* [52], *Cheilocapsus nigrescens* [53], *Stephanitis nashi* [54] and four species of Largidae (*Physopelta quadriguttata*, *Ph. gutta*, *Ph. cincticallis*, and *Macrocheraia grandis*) [55]. In *E. fullo*, the maxillary stylets are smooth externally but equipped with a longitudinal ridge that engages grooves in the mandibular stylets, causing it to curve inward during probing of plant tissue [17]. Moreover, the sharp ends of the maxillary stylet are specialized to pierce plant tissues while probing.

Bro ˙zek and Herczek [37] have studied the interlocking mechanisms of maxillae and mandibles in Heteroptera. Three locks between maxillae and mandibulae have been identified, i.e., dorsal, middle and ventral, similar to Fulgoroidea [3], in contrast with two locks in leafhoppers [6,8]. Our observation of the internal structure of *E. fullo* mouthparts based on the cross-section of the subapical segment of the rostrum reveals the same number of processes in each of the three locks. The food canal is oval and the salivary canal is smaller than that of the food canal, which is semicircular in cross-section. Both maxillary and mandibular stylets are flattened laterally; thus they are higher than wide in cross-section [37]. There are five upper processes on the right maxilla and six processes on the left maxilla, as found by Bro ˙zek and Herczek [37] in their study of other representatives of the Pentatomidae, e.g., *Acanthosoma haemorrhoidale* and *Elasmucha fieberi*.

Heteropteran insects have four feeding methods including stylet-sheath feeding, lacerate-and-flush feeding, macerate-and-flush feeding and osmotic pump feeding [17,28,31], and each is used on a different kind of host tissue. Miles [77] suggested that some pentatomomorphans can employ two types of feeding and that both phytophagous or carnivorous Pentatomorpha produce a stylet sheath. Generally, the polyphagous *E. fullo* primarily suck sap from the trunk, leaves, immature stems and fruits. Therefore, this species presumably employs the stylet-sheath feeding method when feeding from the phloem of the host plant, and employs lacerate-and-flush feeding when feeding on the fruit. In the stylet-sheath feeding method, the insect inserts the stylets into the feeding site (mainly phloem) and forms a salivary sheath around the stylets. In the lacerate-and-flush feeding type, these insects use their strong mandibular teeth to lacerate cells and the sharp ends of the maxillary stylet to pierce fruit for flush and suck feeding.

Usually, before feeding, heteropteran insects secrete some saliva on the surface of the host plant which is re-absorbed repeatedly to test the suitability of the feeding site [78]. In our observations, the labium lip of *E. fullo* has abundant sensilla trichodea (St3) and few sensilla basiconica (Sb5). We suspect that these large numbers of sensilla trichodea (St3) may be used to smear the saliva and the sensilla basiconica (Sb5) act as chemical sensors to taste the liquid.

#### *4.2. Labial Sensillar System*

Many previous authors have described rostral sensilla of Hemiptera and their possible function as chemoreceptors and mechanoreceptors [4,22,48,52–55,79,80]. Detailed morphological descriptions of Pentatomidae labial sensilla have never been previously reported. In this study, eleven types of sensilla were observed on the mouthparts of *E. fullo.*

The sensilla that cover the labial surface (except the labial tip) in *E. fullo* are evidently similar to those of most pentatomomorphan species, as well to other heteropteran species, as reported by several authors [22,44,52–55]. According to the inferred functions of the sensilla, we divided the sensilla on the labial surface into three categories: Thermo-hygroreceptive, proprioceptive and mechanosensory [43,81]. Mechanosensory sensilla include sensilla trichodea (St1, St2) and sensilla basiconica (Sb1), which have no pores or are uniporous and are embedded in flexible sockets. The proprioceptive sensilla include sensilla basiconica (Sb2), located on the junction between the first and second labial segment, and the third and fourth segment, and nonporous cupola (Sca1, Sca2) located on the surface of the cuticle or enclosed in a pit. The thermo-hygroreceptive sensilla include five types (Sb3, Sb4, Sco). Generally, all of the sensilla with this function are nonporous pegs (Sb3, Sb4, Sco).

The labial tip, which contacts the host surface during host selection and feeding, usually has poreless mechanosensory hairs and uniporous or multiporous pegs [34]. According to Rani [26], the carnivorous stinkbug *Eocanthecona furcellata* (Wolff) possesses numerous sensilla of different types at the tip of the labium, e.g., trichoid sensilla, long hairs with profusely branched shafts, an oval-shaped peg surrounded by sensory hairs with branched shafts and a short, stout peg encircled by a group of long hair-like sensilla. Six types of labial sensilla on the labium of phytophagous and predatory pentatomid species were described by Shama et al. [44]. Both studies found long cuticular projections and no sensory function on the labial tip. Nevertheless, in this study we observed in *E. fullo* many very long sensilla trichodea (St3) covering the labial tip, as well as a few sensilla basiconica (Sb5) on the central tip of the labium. Sensilla trichodea probably represent mechanosensilla as their morphology suggests, whereas sensilla basiconica are gustatory (chemosensitive sensilla). *E. fullo* is a polyphagous species sucking the sap from leaves, immature stems and fruits similar to other pentatomomorphan species. Feeding by this species causes yellowish brown spots to appear on the surface of the plant. Extensive injury results the leaf falling off. Damage to fruts includes which causes loss of edible value and yield loss [40]. So far, this is the only pentatomid species observed to have such long and numerous sensilla of the labial tip. Other studied polyphagous heteropteran species have fewer such sensilla (10 to 12 sensilla) and are more uniform in structure [22,44,52–55,82,83] in contrast to *E. fullo* in which the sensilla are much more numerous.

#### **5. Conclusions**

To sum up, the feeding structures in the few species of Pentatomidae studied so far seem similar to each other, presumably due to strong structural and functional constraints on their evolution. However, the mouthparts of *E. fullo* differ from those of previously studied stink bugs in the cross-sectional shape of the stylets, arrangement of labial sensilla and number of teeth of the mandibular stylets. This dissimilarity from other species of Pentatomidae and species of other hemipteran families so far described makes *Erthesina fullo* unique, particularly in its excessively long and numerous sensilla trichodea covering the end of labium. The structure and function of the mouthparts of this species are adapted for phytophagous feeding habits.

The adult feeding process involves several steps, including the exploring and puncturing of the host epidermis, a probing phase, an engorgement phase, and removal of the mouthparts from the host tissue. Studies of feeding behavior and mouthpart morphology of additional pentatomid species are needed to determine how much variation occurs in this diverse and economically important family.

**Author Contributions:** Data curation, W.D.; funding acquisition, W.D.; investigation, Y.W. and W.D.; project administration, W.D.; resources, W.D.; writing – original draft, Y.W. and W.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project was supported by the National Natural Science Foundation of China (Nos. 31772514, 31572306, 31272343) and the Program of the Ministry of Science and Technology of the People's Republic of China (2015FY210300).

**Acknowledgments:** We thank John Ri.chard Schrock (Emporia State University, Emporia, KS, USA) and Chris Dietrich (Illinois Natural History Survey, Champaign, IL, USA) for his comments on an earlier draft of this paper. We thank the Life Science Research Core Services of Northwest A&F University for providing scanning electron microscope.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
